Networked Society
Institute
Current State of
Automated Legal
Advice Tools
Discussion Paper 1
Current State of Automated Legal Advice Tools
Networked Society Institute Discussion Paper 1
April 2018
Authors
Judith Bennett, Tim Miller, Julian Webb, Rachelle Bosua, Adam Lodders
University of Melbourne
Scott Chamberlain
Australian National University
Contact
Julian Webb – julian.webb@unimelb.edu.au
ISBN
978-0-7340-5326-8
Licence
Creative Commons Attribution-ShareAlike – creativecommons.org/licenses/by-sa/4.0/
Networked Society Institute
The University of Melbourne’s Networked Society Institute catalyses interdisciplinary research to understand and create the
connected future. The Institute provides a focal point enabling the University of Melbourne to address the impact and realise
the opportunities of connectivity on society.
networkedsociety.unimelb.edu.au
Current State of Automated Legal Advice Tools
2
Contents
Executive Summary .................................................................................................................. 4
1
Introduction and Purpose ......................................................................................................... 7
1.1 Context of the RALAT Project ....................................................................................................... 7
1.2 The Team ...................................................................................................................................... 7
1.3 Scope of this Paper ....................................................................................................................... 8
2
Overview of ALATs and state of technologies............................................................................. 9
What are ALATs? .......................................................................................................................... 9
A brief history of technologies ................................................................................................... 11
Current state of technologies .................................................................................................... 13
3
Legal landscape ...................................................................................................................... 14
Australian legal industry............................................................................................................. 14
Definition of legal advice ............................................................................................................ 15
Legal advice or legal information? ............................................................................................. 16
Regulation of legal advice .......................................................................................................... 17
Regulation of quality and ethics ................................................................................................. 18
Regulation of process ................................................................................................................. 18
Regulation of capacity ................................................................................................................ 18
Adoption of technology by lawyers ........................................................................................... 21
4
Review of Examples of ALATs .................................................................................................. 22
Classification difficulties ............................................................................................................. 22
Classification by legal advice functions ...................................................................................... 22
Classification by technology ....................................................................................................... 23
Diversity...................................................................................................................................... 25
Specialised standalone technologies ......................................................................................... 26
Enablers of legal advice .............................................................................................................. 26
Further enablers of legal advice ................................................................................................. 27
Human-free smart contracts ...................................................................................................... 28
Sets of ALAT technologies .......................................................................................................... 29
5
Regulatory issues emerging from use of ALATs ........................................................................ 30
The impact of ALATs on regulation ............................................................................................ 30
Legal advice: just by lawyers? .................................................................................................... 30
Quality of legal advice? .............................................................................................................. 31
Substantive law? ........................................................................................................................ 32
Duty to be competent?............................................................................................................... 32
Wider professional duties? ........................................................................................................ 33
The Black Box problem: transparency versus explanation of legal reasoning?......................... 33
Timing? ....................................................................................................................................... 34
The scope and role of regulation? ............................................................................................. 35
6
Conclusions and next steps ..................................................................................................... 36
Appendix A: Examples of ALATs .............................................................................................. 37
Appendix B: Technology definitions......................................................................................... 58
Appendix C: References .......................................................................................................... 64
Current State of Automated Legal Advice Tools
3
Executive Summary
1. This paper is the first from the Regulating Automated Legal Advice Technologies (RALAT) project
supported by the University of Melbourne’s Networked Society Institute.
2. The project focuses on a cutting-edge development in legal technology: the automation of legal advice.
It seeks to understand the practice settings in which Automated Legal Advice Tools (ALATs) are being
adopted, issues regarding their effective management. It also explores the legal, regulatory, and ethical
risks and consequences, and how these will shape access to delivery of legal services.
3. This initial paper sets out to provide the following:
•
A working definition of ALATs;
•
An overview of the current technology landscape;
•
An overview of the Australian legal landscape;
•
A description and evaluation of the current state of ALATs; and
•
An overview of regulatory issues emerging from the use of ALATs.
Examples of current ALATs already in the market are set out in Appendix A.
4. The paper briefly places ALAT developments in the context of advances in technology, specifically the
emergence in the 2010s of a vigorous and qualitatively distinct ‘fourth wave’ of intelligent automation
research. This has arisen through significant investments in hardware platforms with large processing
power, new applications, increased system interconnectivity and enhanced data harvesting.
5. The Australian legal industry has not been immune from these trends. The paper first contextualises
developments by describing the current legal services market. It highlights the market segmentation
into corporate and private client sectors that affects patterns of investment in and take-up of new
automation technologies. At the same time, the potential for technology to disrupt the market is
highlighted including opening-up latent (currently uneconomic) markets and creating new areas and
forms of competition.
6. ALATs are identified as a critical technology in terms of market disruption. The giving of legal advice is a
central function of the legal profession. ALATs create opportunities, notably of commoditisation of
advice-giving. The potential for automated legal advice to reduce costs and open-up latent markets is
significant, particularly in the context of current debates around declining access to justice. But ALATs
also highlight challenges to market incumbents across the industry, for example, as technical legal
expertise becomes increasingly open to automation. US corporates like LegalZoom and Rocket Lawyer
are examples of ways non-lawyer entities may seek to enter and disrupt traditional markets for smaller
business and consumer legal services. Policy questions also arise as to the risks such disruptors may
pose to consumers, and how regulation should respond (if at all).
Current State of Automated Legal Advice Tools
4
7. The scope and scale of current ALAT development is discussed in section 4 with examples given in
Appendix A. The paper seeks to classify ALATs by reference both to function and intelligent capability.
Functionally, it identifies five subsets of technology:
•
Specialised standalone technologies, such as legal chatbots, apps and virtual assistants,
•
Enablers of legal advice such as legal automated drafting, legal document review and legal
algorithms,
•
Further enablers of legal advice such as legal data analytics and predictors, and legal artificial
intelligence,
•
Automation of legal advice with truly smart contracts, and
•
Sets of ALAT technologies enabling NewLaw business models and legal technology companies.
In terms of capability, a range of technologies are found to exist from simple non-AI tools relying solely
on hard-coded decisions through to “smarter” or “more intelligent” sophisticated technologies that use
deep learning and can parse text, learn causations and correlations from data, and reason about these
to make predictions.
8. We find that the market in ALATs is developing rapidly. Most applications have entered the market
since 2014, with the greatest activity in the US – likely reflecting the greater availability of venture
capital. ALATs are both new and at varying levels of sophistication, with the majority at the lower end
of the intelligence scale.
9. The challenge to legal services regulation posed by ALATs is explored in sections 3 and 5 of the paper.
Section 3 introduces the problem in terms of the wide definition of legal practice in Australia, which
reserves legal work to the legal profession. This section also explains how regulation demarcates the
provision of (unregulated) legal information from (regulated) legal advice-giving. This regulated
boundary between information and advice could prove to be a critical zone of engagement,
determining the impact of new market entrants, including unregulated disruptors. While controls on
advice-giving have consumer safety justifications, the development of automated intelligence
potentially changes the risk environment. To this extent, automation re-opens important questions
regarding the scope and proper function of legal services regulation.
Current State of Automated Legal Advice Tools
5
10. Section 5 details the ways in which ALATs create a number of specific issues for regulators and
government regarding: the usability of the existing information/advice distinction; how quality of
information and advice is to be assured; expectations of professional competence and related
standards; the need (if any) for explainability standards, and the timing of any reforms. The section
concludes by identifying three wider challenges highlighted by automation being uncertainties of the
law, the consistency, reach (functional and geographical) of regulation, and the purpose of regulation.
Most fundamentally, it suggests automation brings into stark relief the question whether the purpose
of legal services regulation is to facilitate access to lawyers or access to law.
11. In conclusion, this paper seeks to both inform and raise the extent of debate as to the potential for
ALATs to transform the delivery of legal services in Australia. It both complements and is distinct from
recent profession-centric contributions, such as the NSW Law Society’s FLIP Inquiry (2016), the Law
Institute of Victoria’s Disruption, Innovation and Change: The future of the legal profession (2015) and
Law Society of Western Australia’s The Future of the Legal Profession (2017).
12. Phase 2 of the project involving qualitative research will seek to further discussion and inform policymaking by fieldwork with a range of stakeholders: members of the profession, regulators, ALAT
developers and producers, and access to justice groups. In scoping that fieldwork, Section 5 highlights a
number of critical questions for discussion:
a. To what extent (if at all) is the legal information/advice distinction a barrier to legal services
innovation?
b. Conversely, is there a case for bringing ‘legal information’ substantively into legal services
regulation? If yes, how might that best be done?
c. What additional challenges to quality of advice are created by ALAT technologies? Are specific
new mechanisms or approaches necessary or adequate to regulate the quality of automated
legal advice?
d. Should the duty of professional competence be expressly extended to include an obligation to
stay up-to-date with relevant technologies? Are there other professional obligations that may
need to be revised in the context of increased automation of legal advice?
e. How (if at all) is automation of legal advice already impacting pricing and billing practices in the
profession? How might it impact them in the next five years?
f.
Should explainability standards be devised for ALATs using sophisticated (‘black box’)
automated intelligence? Why/why not?
g. Are there critical areas of automation of legal advice that create risks for consumers such as
where the separation between service provider liabilities under the general law (tort,
Australian Consumer Law, etc) and professional responsibility under the Uniform Law or
equivalent Legal Profession Acts?
h. How should or could regulation overcome the problem that online services may be delivered
from outside the regulator’s physical jurisdiction?
Current State of Automated Legal Advice Tools
6
1 Introduction and Purpose
Context of the RALAT Project
The Regulation of Automated Legal Advice Technology (RALAT) in Australia project seeks to further the
understanding of the practical and regulatory implications of specific new technologies supporting and
providing delivery of legal advice. We refer to such technologies as Automated Legal Advice Tools (ALATs).
There is currently widespread uncertainty regarding the potential of these advanced technologies, and
their likely impact on legal practice and service delivery over the medium to longer term.
The project’s objective is to explore the emergence and development of ALATs, their impact, limits, and
how they are transforming legal practice in Australia. The project particularly seeks to identify the issues
ALATs raise for the current regulatory landscape. This research also begins to address concerns that much
debate about the automation of legal work has not sufficiently engaged with the technical capabilities of
systems when discussing how different kinds of legal work may be more or less “amenable” to automation
(Remus & Levy, 2016).
To explore these themes and issues for ALATs, the project team is carrying out the following steps:
•
Mapping of current tools for automated legal advice giving (this paper);
•
Literature review to identify and review key themes and issues;
•
Interviews with legal practitioners, regulators and vendors to understand the scope,
operation and impact of ALATs;
•
Analysis of appropriateness of the current regulatory landscape; and
•
Making of recommendations.
The project seeks to better understand the technical development and practice settings for ALATs in order
to: inform policy on regulation, ascertain ALATs impact, support effective management, and how they may
shape access to delivery of legal services in an increasingly networked society. Additionally, the project will
contribute to the broader debate surrounding the future of professional services across the economy.
The Team
RALAT is an interdisciplinary project with a research team drawn from the Melbourne Law School (Legal
Professions Research Network), School of Computing and Information Systems, Department of
Management and Marketing, and the Networked Society Institute (NSI) all at the University of Melbourne,
together with ANU College of Law, Australian National University. The project has been financially
supported through the Networked Society Institute’s Seed Funding Program.
Current State of Automated Legal Advice Tools
7
Scope of this Paper
This paper is the first in the RALAT Project series. It offers an introduction to the technological and
regulatory issues that are to be explored, delivers a preliminary evaluation of current ALAT developments,
and provides a launch-point for discussions in the second, empirical, phase of the project.
In the following sections, the paper provides:
•
A working definition of ALATs;
•
An overview of the technology landscape;
•
An overview of the Australian legal landscape;
•
A description and evaluation of the current state of ALATs; and
•
An overview of regulatory issues emerging from the use of ALATs.
It concludes with three Appendices providing the following information:
A. Examples of current ALATs;
B. Technology terms and definitions; and
C. References.
Current State of Automated Legal Advice Tools
8
2
Overview of ALATs and state of technologies
What are ALATs?
This paper focuses on the current state of automated legal advice tools (ALATs); that is, technologies whose
major purpose is “giving legal advice” as regulated by the legal profession. The rather circular quality of this
definition reflects, first, that “legal advice giving” is itself a fuzzy concept. It lacks clear definition, even
within legal professional regulation (see section 3.3). We recognise that technology can change not just the
efficiency with which legal advice is delivered, but the way in which it may be bundled with other services
(so that boundaries, such as between legal and business advice become more blurred), and, even more
fundamentally, that technology may perform certain advice-giving tasks quite differently from human
advisers. This of course reflects the extent to which advising, above a relatively basic level, requires
capabilities that stretch, or are currently beyond the reach of software: for example, the capacity to factorin complex considerations of interests, tactics and values, or the ability to engage in unstructured
communication with others (Remus & Levy, 2016: 40, 65ff).
The emphasis on automation in our definition reflects the focus on technologies “by which a process or
procedure is performed without human assistance” (Groover, 2014). This limits the scope of the study to
the use of advanced cognitive tools and emerging technologies rather than existing process-based tools.
While process automation “has been maturing quietly over the last decade and is now used for enterprisescale deployments” that can assist in giving legal advice, “… intelligent automation, while still nascent,
promises hugely transformative potential in the near future” (Deloitte, 2017).
In sum, by emphasising automated advice-giving, we include ALATs that use legal analysis, legal reasoning,
and prediction functions:
•
To give legal advice on their own;
•
To give advice supervised or reviewed by a lawyer;
•
To assist or augment legal advice given by a lawyer; and
•
To offer limited or partial legal advice by unbundling transactions into smaller discrete tasks.
Advice-giving ALATs are currently at the leading edge of new information technologies and are still at a
relatively early stage of adoption by lawyers and the legal industry. ALATs rely on advances in big data,
interconnectivity and processing power combined with logic techniques variously known as artificial
intelligence, intelligent automation, cognitive computing, natural language processing and machine
learning. Examples are given in Figure 1 below. The scope and potential scale of ALAT use is uncertain as
the underlying technologies are rapidly and often exponentially changing. We thus find a range of ALATs at
different stages of sophistication (Appendix A).
Current State of Automated Legal Advice Tools
9
Figure 1: Areas in the broad AI ecosystem (NarrativeScience.com).
Key terms used in this paper are explained here, others are in Appendix B. Artificial Intelligence includes
several sets of technologies, applications and consequences. It is used here as an umbrella concept to
encompass the AI ecosystem. Artificial refers to machines, and Intelligence is also used in a broad sense as
to “learn, reason and act in a rational way” (Miller in Webb, Miller, Bosua, & Bennett, 2017). In this context
it is important not to read too much into terms like intelligence or learning. As both Turing and Marvin have
observed, intelligence is simply a name for any problem-solving mental process that we do not yet
understand. Like the concept “unexplored regions of Africa: it disappears as soon as we discover it”
(Copeland, 2017). Similarly, learning in a machine context does not necessarily imply a capacity for higher
order cognition; it may simply indicate the functional ability of a system to improve its own performance of
a specified task, over time (compare Surden, 2014).
Using artificial intelligence in this broad sense, this paper also examines ALATs as to their positioning on
various dimensions that range from smart and smarter to the more intelligent programs that can teach
themselves to learn, plan, and act “when exposed to new data in the right quantities” (Huff Eckert, Curran,
& Bhardwaj, 2016).
Current State of Automated Legal Advice Tools
10
A brief history of technologies
Artificial intelligence is not new. However, a review of the history of technology development shows the
extent to which current advances are exponential rather than linear.
The 1950s saw the first wave of AI with McCarthy defining artificial intelligence as “the science and
engineering of making intelligent machines” (McCarthy, Minsky, Rochester, & Shannon, 1955). Much
research in this era took the view that computers could ultimately be made as intelligent as humans by
mimicking and replicating the functions and processing structures of the human brain. This overly
ambitious premise, and the limits of processing power, largely account for the limited advances in this
period. Consequent frustration with lack of success, lack of scale and failure of assumptions led to a lull in
AI research and funding during the 1970s. Interest in legal computing began in the 1960s, with the first
strands of research into AI and the law in the 1970s (Nissan, 2017).
The 1980s saw a second wave of AI research with the success of expert systems designed as rule-based
conditional logic operations using hand-coded knowledge. This approach was facilitated by the rise of the
PC and client-server model. Workable legal expert systems were developed, with examples including
Richard Susskind’s PhD (Susskind, 1987), his collaboration with Capper on Latent Damage Law, and BenchCapon’s project creating tools to assist with frontline social security benefit assessments (Bench-Capon,
1991). Yet due to various failures from high ambitions (Leith, 2010) to non-scaleable hand coding, research
and funding again stalled, and developers entered the “AI winter” of the late 1980s to mid 1990s.
The increased availability of processing power and use of the internet led to a small third wave at the end
of the 1990s and early 2000s. However, progress in AI was dwarfed by other features of the information
technology revolution, and particularly the growth of process-based automation across business and
government.
The current and fourth wave has seen rapid adoption of AI technologies in new software applications
advancing pattern recognition capabilities or software agency, that is, the ability of software to behave like
real actors. This is being driven by third platform technologies that are AI friendly (such as Alphabet’s
Tensor Processing Unit, Nvidia’s Volta GPU and IBM’s TrueNorth neuromorphic computing platform),
extensive gains in computer processing power and speed, increased network interconnectivity, greater
scalability, and massive proliferation of the digital data necessary to create the large data sets that make
intelligent automation possible (IDC, 2018). The development by governments and corporations of
innovation accelerator projects has also significantly assisted innovators and disruptors to achieve
commerciality, saleability and a faster move to market.
The legal industry has not been immune from these trends. There is increasing market demand plus a push
from large well-financed companies to leverage the potential of new technologies (iScoop n.d.). AI
techniques such as text mining, knowledge based self-learning, machine learning and natural language
processing (are coming into play as a means not just of enhancing accuracy and efficiency of existing
services, but in creating new value-added services, such as automated legal prediction. They can analyse
huge amounts of data with descriptive, diagnostic, predictive and prescriptive analytics tools (iScoop n.d.).
Current State of Automated Legal Advice Tools
11
Recent advances see even more sophisticated AI technologies with even more complex machine learning
and artificial intelligence including technologies such as neural networks, natural language generation and
social intelligence solutions often layered together. Blockchain technologies add further dimension to
automation, as they provide for transactions to be completed in a decentralised, distributed manner with
no intermediaries or human involvement (Bacina, 2017).
Earlier innovations, like word processing, email and the internet have significantly transformed legal
practice. It would be fundamentally unwise to assume that current technological innovations will have
effects that are ultimately any less profound. At the same time, however, the hype (and fear) around
automation is loud. Some champions of change argue that “AI and machine learning have reached a critical
tipping point” and will increasingly augment and extend virtually every technology enabled service, thing or
application” (Gartner, 2017). Yet as the hype cycle curve in Figure 2 indicates, while AI and machine
learning are at the tip of the hype cycle, they are still at least some two to ten years away from significant
adoption.
Figure 2: Gartner Hype Cycle for Emerging Technologies (Gartner, 2017)
Current State of Automated Legal Advice Tools
12
Current state of technologies
Legal researchers have highlighted areas of the law where AI technology will dramatically change what
lawyers currently provide: discovery, legal search, document generation, advice generation, and prediction
of case outcomes (McGinnis & Pearce, 2014). For this paper, a critical issue is that the cutting edge of
fourth wave applications is fundamentally different from second wave expert systems and third wave
automated process technology that used more causal, defined logic. A key difference is that the rules by
which machine learning technology recommends decisions are not explicitly programmed by a human
(Lodewyke, 2017), rather the machines “self-learn” from data using statistical reasoning. The machines are
provided with gigabytes of data from selected databases and use algorithms to find concepts and patterns
in the data, form and test hypotheses, and develop recommendations with analysis of that data.
The more intelligent systems can learn, adapt and potentially act autonomously rather than simply execute
pre-defined instructions. The capacity for self-learning in turn enables these systems to build more
complex, dynamic and adaptive models, and “improve” their performance over time on specific tasks
(Surden, 2014). However, these machine systems do not “understand” their hypotheses in any sense of
human understanding, nor do they reason about the causality in a strong or human-like way (Pearl, 2018).
They also do not aim to mimic human intelligence. One vision of AI is to have systems that can learn the
causal rules, using for example, inductive learning. However, such technology does not scale nor generalise
as well as current statistically-based methods (T. Miller, 2018).
Will these new and accelerating technological advances see ALATs attain “superintelligence” in the near
future? Superintelligence refers to intelligence greater than the smartest humans (Bostrum, 1997).
Currently superintelligence is hypothetical. If superintelligence occurs, such machines would be able to
create new machines more intelligent than themselves, ad infinitum (Bostrum, 2014). In other words,
machines that “can learn and change future behaviour, leading to the creation of more intelligent devices
and programs” (Gartner, 2017). While IT experts differ as to the speed with which AI will attain
“superintelligence”, this is not likely to be in the next 5 or even 20 years. Predictions currently range
between 25 to 80 years, with an average of 45 years (Etzioni, 2016). However, it is important to note that
similar predications made in the past have not been fulfilled. In this paper our focus is limited to those
technologies that are currently in or planned for production and available in the market for use by lawyers
and the community.
Current State of Automated Legal Advice Tools
13
3 Legal landscape
Australian legal industry
The Australian legal industry context is important as the industry is currently facing multiple external and
internal pressures, accelerated by new technologies. The Australian legal profession is a diverse industry
with a multitude of players carrying out a range of legal advisory work. The majority, 69% of 70,000 plus
solicitors work in private practice, with 16% working in-house, 10% in government and 5% in other sectors
such as community and legal assistance (Urbis, 2017).
While forming only 3% of Australian law firms, the large and increasingly international corporate law firms,
along with the Big Four accounting firms and niche legal practices are likely to be a driving force in the
development of ALATs, initially for routine and commoditisable work. This market segment has largely
corporate, in-house and government clients seeking legal advice – much of which is likely to be relatively
complex and specialised, and with relatively high profit margins. They are increasingly pressured by clients
for greater efficiencies and reduced costs and have the resources to invest in technology to improve the
value they provide.
Smaller firms with 1 to 4 partners representing 92% of lawyers in private practice. Of these, 73% are sole
practitioners (Urbis, 2017). While this sector includes a growing number of boutique and ‘NewLaw’ entities
that are adept at leveraging technology for corporate clients, this segment as a whole tends to have a client
profile of smaller business and individual clients. These firms give legal advice that ranges from simple to
complex yet more likely to be high volume and routine with lower profit margins. They tend to have less
resources to invest in technology and may also see less pressure to change.
Community legal centres and legal aid services are a small sector with more limited resources and a diverse
range of legal advice functions for their disadvantaged clients. This sector is seeing uptake in ALATs –
especially those in niche and discrete areas of law to improve access and outcomes for clients.
There is also a potential market in the context of large-scale unmet legal need, as many individuals and
SMEs experience legal problems, yet do not qualify for legal aid and cannot afford lawyers (Coumarelos et
al., 2012). This “missing middle” constitutes a “latent market” should cheaper, quicker legal advice become
available (Productivity Commission, 2014; Susskind & Susskind, 2015). Automation is, obviously, one way in
which the access to justice gap may be narrowed. Automation may reduce existing cost barriers and make
economic work that is not cost-effective to lawyers under existing service and pricing models. The
reduction of geographical and time constraints in moving online also creates potential competition from
differently or non-regulated overseas providers despite jurisdictional barriers to market entry. These
providers may operate on larger scale than an individual law firm and have different cost models and
margins allowing them to compete within the market. This is being seen as a potentially significant market,
and possibly consumer, risk by the Australian legal profession (see, Law Society of NSW, 2017).
This overview shows how the market for legal advice and therefore for ALATs is significantly differentiated
by law firm and client segments, and also the content and kind of legal advice sought.
Current State of Automated Legal Advice Tools
14
Definition of legal advice
This project focuses on the work of legal advice and its automation. Advice-giving is a core lawyering
function and at the heart of legal services regulation. The potential of technology to support the advicegiving function or even replicate and commoditise human advice-giving is central to the revival of interest
of automation in law. Therefore, the question of what is involved in giving legal advice is important both
conceptually and in regulatory terms.
Conceptually, it is generally agreed that legal advice concerns a legal problem or question specific to a
client and “is taken as being tailored to the individual circumstances of the ‘client’” (Giddings & Robertson,
2001). Typical examples of legal advice include:
•
Selecting, drafting, or completing legal forms, documents or agreements that affect a person’s legal
rights;
•
Representing a person before a court or other decision-making body;
•
Negotiating legal rights or responsibilities on behalf of a person; and
•
Predicting an outcome of a legal dispute.
Three questions arise from this definition with implications for system design, regulation, and the future
development of the legal services market. First the question of (irreducible) complexity, second the
separability of the legal component of advice, and third the more technical question of what distinguishes
legal advice from ‘mere’ legal information.
Simplicity is relative. What is simple to an expert in the field will often seem difficult or complicated to a lay
person. A distinction needs to be made between complicated and genuinely complex. Sometimes legal
advice giving is straightforward, drawing on rules and principles that are relatively simple and clearly
defined, such as whether or not a parking ticket was issued correctly, or whether you can legally change
your name. Other times it may be more complicated such as whether or not a complex taxation ruling was
issued correctly, or whether someone else can legally change your name. Simple, routine, advice-giving is
can easily be automated, for example where rules are clearly expressed, readily applicable and reducible to
simple decision trees.
There is, however, a genuine and possibly irreducible complexity to more bespoke legal advice giving. This
requires the interplay of a broad range of knowledge, skills and professional attributes, such as capacity
and skills in legal research, evidence gathering and weighting, legal analysis, the exercise of professional
and ethical judgment, prediction as to the range of outcomes and assessment of those most likely, and the
negotiation and representation of client interests. More expansively, it might also include elements of
client counselling and relationship management, along with business decisions such as pricing/costing the
service. This complexity sets a limit to what is replicable by technology. While technology is playing a
growing role in assistance – particularly for research, analysis and prediction – it cannot presently replicate
the whole package. Whether it will ever be able to do so is still moot.
Current State of Automated Legal Advice Tools
15
Secondly, should the deep technical expertise associated with legal professional advice-giving become
more readily replicable via machine technologies, this has major implications for the market. It potentially
speeds up the blurring of boundaries between legal and other forms of commercial advice and consultancy.
If technology can do the legal component as well or better than a human, what becomes of the legal
function? What is distinctively legal? What will lawyers offer in the future that constitutes their unique
selling position? And can the separation of law from other parts of business consulting still be justified?
Thirdly, the legal information/advice distinction matters both for system design and regulation. Legal
information can be conceived of as a public good that should be readily and openly accessible. This
principle sets some limits on both the regulation and commercialisation of pure legal information.
Alternatively, legal advice has long been treated as a quasi-market good subject to significant protections
and market closure. Legal technology has the potential to exploit and disrupt these boundaries and, at the
least, create a set of new questions and challenges for regulation. This is a matter we discuss further in the
next Section 3.3 and in Section 5.
Legal advice or legal information?
The concept of legal advice can be difficult to define and, specifically, to distinguish from the activity
of providing legal information, which may be separate from, and also often a precursor to, as well as
part of, giving legal advice. Commonly, “[l]egal information is described as generic, not addressing the
particular circumstances of the individual. Legal advice is more tailored and specific to the needs of the
consumer” (Giddings & Robertson, 2001). Examples that are not legal advice include:
•
Legal information and self-help forms obtained from free online legal websites, including law firms,
legal assistance and government departments;
•
Advice from friends, family members, or previous clients of a lawyer;
•
Information heard on the radio, read on social media websites or seen in newspapers; and
•
Printed materials listed in a how to guide.
In theory there is a general distinction made between the giving of legal advice and the giving of legal
information. However, in practice, there is not always a clear line between these.
The problem is apparent when we consider the development of technology-based and technologyenhanced document generation services, where there is a range of choices enabling a degree of
customisation and selection. At what point does the technological assistance move from information to
advice?
This distinction matters for regulation as discussed in the next section. For example, the information/advice
distinction has been a key issue in the US, where new service providers like LegalZoom, Rocket Lawyer and
Access Legal have developed new and sophisticated infomediary business models.
Current State of Automated Legal Advice Tools
16
Regulation of legal advice
This section describes the ways in which legal advice is presently regulated. It also provides a background to
Section 5 that explores how current regulatory arrangements impact automated legal advice.
The Australian legal profession is regulated in its practice by legislation, common law duties, and codes of
conduct. The current regulation of legal advice operates through this interplay of general principles of law
and specific professional regulation. In Australia, most of the regulations are set out in formal professional
standards having the status of secondary legislation while some are supplemented by soft law guidance or
norms from regulators or professional bodies. An example, showing how these interact, is the general duty
to provide competent advice. If a solicitor does not act competently, they can be found guilty of both
common law negligence, and of a breach of a fundamental professional obligation (Legal Profession
Uniform Law Australian Solicitors Conduct Rules 2015, rule 4.1.3). However, competent advice is not
defined in the professional rules leaving much of its scope and standard to be assessed by the common law.
Regulatory arrangements in Australia are complicated by the fact that the legal profession is organised on a
State and Territory basis. New legislation, consistent with a Model Law was enacted in all Australian
jurisdictions between 2004 and 2008, except South Australia. Since July 2015, the legal professions in
Victoria and New South Wales have been governed by a fundamentally common regime under the Legal
Profession Uniform Law (the Uniform Law). Despite this, regulations still vary. Although it is likely that
outcomes with concerning the legal advice and information distinction would be substantially similar across
Australia. (A comparative survey is in Beames, 2017, p298-9).
These laws and regulations govern three main features of legal advice relevant for ALATs:
•
The intrinsic quality of legal advice;
•
The process of advice-giving; and
•
The capacity to deliver legal advice.
Current State of Automated Legal Advice Tools
17
Regulation of quality and ethics
Quality controls include lawyers’ duties to the court and to clients, such as to avoid professional negligence,
to provide standards of service consistent with the Australian Consumer Law, and professional duties to be
competent. Lawyers also have duties to maintain high ethical and professional conduct standards. These
may also be seen as an intrinsic part of the quality of professional legal advice.
Regulation of process
There is a less well-established and less extensive system of process regulation. While there are no legal
rules directly governing the process of how to give legal advice to a client, there are regulations that
indirectly impact that process. For example, process rules around information standards for cost disclosure
require a description of the kind of legal work being carried out (Legal Profession Uniform Law Division 3).
This also includes description of the scope of retainer and disclosure of certain risk information such as
conflicts of interest or commissions. Other forms of process control exist in general legislation, including,
for example, privacy and data protection requirements and consumer law.
Regulation of capacity
As legal advice-giving is a controlled activity, the category of capacity rules covers what is legal advice and
legal practice, and who can lawfully give legal advice. The nature and extent of capacity regulation has
significant consequences for market segmentation and access to justice. By defining who can give legal
advice, it sets direct controls on supply-side access to the market for legal services. While such controls
have consumer protection justifications, in the context of access to justice, policy questions are being
raised as to where and how to draw capacity lines. It is recognised that both regulation and technology can
be used as disruptive tools, which have the potential to separate access to justice from access to lawyers.
In this section we look at the way common law systems in general, and Australian systems in particular,
regulate capacity, and therefore access, and address its consequences for the technological disruption of
the legal services market. Most common law systems seek to regulate the supply of legal services, in one of
two ways. Australia, Canada and the US treat the supply of legal services broadly as an activity reserved to
certain professional titles, such as a solicitor or barrister. In contrast, England and Wales treat the supply of
legal services by reserving specific types of legal work to those with a professional title.
Current State of Automated Legal Advice Tools
18
In England and Wales, legal services regulation focuses on six broadly reserved activities. These are the
exercise of a right of audience, the preparation and conduct of litigation, preparation of a contract for the
sale or transfer of land, probate and notarial activities, and the administration of oaths (Legal Services Act
2007, s.12). Solicitors and barristers in England are regulated for all their work. This includes those reserved
activities listed above but also much business and employment work, and large areas of consumer and
social welfare law. As a result, there is a growing unregulated market for unreserved legal services. These
services may be provided by non-lawyers, or by lawyers without practising certificates, for example in
human resource departments, debt recovery businesses, or as self-employed professional litigation
assistants and lay advisers (Legal Services Board (E&W), 2016).
In the US and Canada, the legal services markets are less liberalised with ‘legal advice’ used as a term of art
to distinguish core activities reserved to the regulated legal profession. It is illegal to provide legal advice
other than as a licensed lawyer since this constitutes the unauthorised or unqualified practice of law
(known as UPL). The history of UPL controls reveals mixed motivations ranging from the private interests of
bar associations restricting competition from non-lawyers (Denckla, 1999) to the public interest and
consumer protection arguments for access to justice and assuring quality of legal advice-giving.
As noted above, in the US, new service providers like LegalZoom, Rocket Lawyer and Access Legal have
developed sophisticated infomediary business models (see Appendix A). The US has seen legal industry
stakeholders challenge such ALATs, alleging they are engaging in “unauthorised practice of law” often using
the information/advice distinction to challenge such disruptors (Campbell, 2012; Brescia, McCarthy,
McDonald, Potts & Rivais, 2015). For example, LegalZoom has had cases filed against it from California to
North Carolina asserting that LegalZoom’s process goes beyond providing legal documents and enters the
role of legal adviser. These challenges had early success in some states but have also seen some failures.
Overall, the reasoning and outcomes in the courts have varied widely. Reversing the trend, LegalZoom filed
a US$10.5m antitrust suit against the North Carolina State Bar for denying it permission to sell legal services
in June 2015 (IBA, 2016). The case ultimately settled with a consent agreement which permitted the
company to continue operating in that jurisdiction. Notwithstanding the various challenges, LegalZoom
continues to operate “filling what is seen as a previously unmet gap in the [US] legal services market and
serving to date over three million customers” (Beames, 2017, p298), and expanding its operations
overseas.
Does this mean that going forward the new disruptors have a green light? Not necessarily. The US litigation
on UPL focuses on the nature of the technology. For example, one US State court found there was no UPL
because the software did not ultimately deliver advice (Medlock v LegalZoom, 2013), finding that:
LegalZoom’s software acts at the specific instruction of the customer and records the customer’s
original information verbatim, exactly as it is provided by the customer. The software does not
exercise any judgment or discretion but operates automatically in the same fashion as a ‘mail merge’
program” (Ambrogi, 2014).
However potentially, as software becomes more sophisticated and intelligent, the advice/information
distinction may surface again.
Current State of Automated Legal Advice Tools
19
The information/advice distinction operates somewhat obliquely within Australian legal services regulation.
This reflects the historical tendency of legislation to define the scope of regulated legal services as
equivalent to “engaging in legal practice”, while leaving the meaning and scope of legal practice itself
undefined (Bartlett & Burrell, 2013). This tendency is continued in the new Uniform Law. The Legal
Profession Uniform Law prohibits “unqualified entities” from engaging in legal practice (Section 10). It then
defines “legal practice” non-exhaustively and in a manner that appears somewhat circular (see also
Beames, 2017, p298). To ‘‘engage in legal practice” includes to ‘‘practise law or provide legal services but
does not include engage in policy work (which, without limitation, includes developing and commenting on
legal policy)” (Legal Profession Uniform Law, s6). In turn, legal services are defined as “work done, or
business transacted, in the ordinary course of legal practice” (Legal Profession Uniform Law, s6 note).
The Uniform Law’s emphasis on work done in the “ordinary course” of practice echoes earlier tests
adopted by the courts in determining UPL, which made material the distinction between information and
advice. The widely approved starting point is Justice Phillips decision in Cornall v Nagle [1995] 2 VR 188.
Section 90 of the then Legal Profession Practice Act 1958 (Vic) prohibited any unqualified person from
“acting or practising as a solicitor”. His Honour identified giving legal advice per se (at least for reward) as
lying “at or near the very centre of the practice of law.” The control over legal advice was justified in classic
public interest terms; it involves “doing something which, in order that the public may be adequately
protected, is required to be done only by those who have the necessary training and expertise in the law”
(Cornall v Nagle (1995) p20).
Later cases have shown that any tailoring of legal information to specific circumstances and to “ensure a
specific result” is likely to involve legal advice and therefore be seen as (regulated) legal practice (Barristers’
Board v Palm Management Pty Ltd [1984] WAR 101; followed in Attorney-General v Quill Wills Ltd [1990]
WASC 604, Legal Practice Board v Giraudo [2010] WASC 4, and Legal Services Commissioner v Walter [2011]
QSC 132). This also has the effect of leaving informational legal work that is merely clerical (outside the
reach of regulation (Attorney-General v Quill Wills Ltd [1990] WASC 604).
Australian academics and professional bodies are recognising the regulatory significance and effect of UPL
rules in the context of new technologies. Recently two professional bodies indicated some unease at the
current scope of regulation. The Law Society of NSW’s FLIP Inquiry (2017) explored the role of new
information providers and concluded by recommending that “the Law Society investigate bringing legal
information within the regulatory fold” (Recommendation 16). In contrast the Law Society of Western
Australia (2017) has recently called for the creation of a regulated market in ALATs stating “An entity that is
not a qualified entity may not in this jurisdiction give to another entity a product or thing that provides, or
is capable of providing, legal services unless the second entity is a qualified entity” (p12).
In conclusion, the Australian jurisprudence aligns more closely with the US and Canadian approaches yet
retains some of its own features and complexities. There is also the risk of inconsistency across state and
territory jurisdictions. Therefore, continuing conversation is required about how regulation should evolve..
Current State of Automated Legal Advice Tools
20
Adoption of technology by lawyers
Richard Susskind’s book The End of Lawyers? suggested a future where lawyers adapted to new
technologies or were gradually replaced by increasingly capable machines (Susskind, 2010). While adoption
of technologies has been slower in law than in other service industries (Wallace, 2017), this is changing.
Technology is now core to the practice of law for most lawyers with widespread use of email, online legal
research and electronic court filing. Process-oriented technologies are gradually increasing with online legal
information, simple document assembly, e-discovery, workflow and project management. New
technologies are adding front-end client-facing services and business development (J. Goodman, 2016).
Lawyers increasingly seek to use technologies to give or assist them in giving legal advice, acknowledging
the benefits of productivity gains and cost savings.
The literature is increasingly predicting the rise of artificial intelligence for lawyers and legal work (for
example Beaton, 2014; K. Miller, 2015; Law Society of NSW, 2017; Susskind & Susskind, 2015). Some
authors predict the rise of robot lawyers (Koebler, 2017) and technological unemployment (Mangan, 2017).
Our review has found that this is not yet the case with ALATs and legal advice. Other authors are more
restrained seeing that AI can “make life easier for lawyers” (Lodewyke, 2017) and can “remove lawyers
from routine tasks that computers can handle, allowing lawyers to focus on tasks that truly add value, and
things that computers really cannot do or do well” (S. Miller, 2017a).
The new and emerging ALATs that are the focus of this project are sometimes simple. Yet they are
becoming increasingly smart, smarter and intelligent. They are built on artificial intelligence, natural
language processing, and machine learning tailored for legal services. They have access to vastly increasing
amounts of digitised data such as court cases, legal documents and pleadings, legal journals and research,
legal websites, and a range of legal subject matters.
The next section explores the current state of ALATs and the key trends in the automated giving of legal
advice.
Current State of Automated Legal Advice Tools
21
4 Review of Examples of ALATs
This paper has collated and reviewed a selection of examples of current state of ALATs as at February 2018
from the UK, US and Australia. The selection includes new business models that arise from the use of these
ALAT technologies. These are set out in Appendix A. This section examines the trends and findings arising
from the development, use and uptake of ALATs.
Classification difficulties
Classifying ALATs is difficult. There are multiple dimensions to legal advice and to ALAT technologies for
even very task-specific AI. From this, a key finding is that there is no necessary correlation between the
complexity of the underlying technologies and the complexity of legal advice. Both ALAT technologies and
legal advice have far more dimensions. This paper is a starting point.
Classification by legal advice functions
Our primary classification of ALATs is by their function in relation to legal advice giving. As seen in
Appendix A, these are:
•
Specialised standalone technologies, such as legal chatbots, apps and virtual assistants;
•
Enablers of legal advice such as legal automated drafting, legal document review and legal
algorithms;
•
Further enablers of legal advice such as legal data analytics and predictors, and legal
artificial intelligence;
•
Automation of legal advice with truly smart contracts; and
•
Sets of ALAT technologies enabling NewLaw business models and legal technology
companies.
Current State of Automated Legal Advice Tools
22
Classification by technology
This project aims to engage with the technical capabilities of ALATs and so understand how legal advice
may or may not align with existing technologies.
With the ALATs, there are scales of “smart” in the underlying technologies. Some ALATs are simple, being
non-AI tools relying solely on hard-coded decisions. More ALATs integrate some concepts of AI, with most
being “smart”, using simple pre-programmed causal rules and automated reasoning to make decisions.
Smarter ALATs add machine learning where a computer can make decisions with minimal programming,
that is, rather than following pre-set rules about how to interpret a set of data, the computer uses learning
algorithms for solving particular problems that allow it to determine the rules itself. Deep learning uses
more advanced algorithms to perform more abstract tasks. With experience, data, and feedback,
computers become better at the relevant tasks. “Smarter” to “more intelligent” sophisticated technologies
can parse text, learn causations and correlations from data, and reason about these to make predictions.
The examples of ALATs contained in Appendix A are primarily classified by their functions in relation to the
giving of legal advice. For many of the ALATs, it is not always possible to know the technology underlying
them and additionally these may be private or commercial-in-confidence. We have made educated guesses
and will aim to improve these in later stages of the research.
While we recognise that there are multiple dimensions, this project uses the dimensions and rating scale
for ALATs in Appendix A set out in Figure 3 below. This scale ranges from simple to intelligent with the
intermediate steps of smart and smarter (T Miller, 2017).
Rating
Simple
Smart
Smarter
Intelligent
Symbol
Figure 3: ALAT Rating System
We seek to explore the technical state of automated legal technology using a broad umbrella perspective
of artificial intelligence. As a basic model, we place each technology into one of four categories based on
the ‘level’ of intelligence. These categories are derived from two dimensions, each with two values:
•
The first dimension is whether the knowledge in the underlying tool model has been hand-coded
based on human expertise or has been learnt via data, that is, machine learning, and
•
The second dimension is whether the underlying model consists of causal information or merely
correlations. As noted by Pearl (2018), statistical, or ‘model-free’ learning techniques answer
questions about associations between variables, but their lack of causal models means they cannot
answer questions about interventions (‘What if we did X?’) or counterfactuals (‘Why did X
happen?’).
Current State of Automated Legal Advice Tools
23
Figure 4 shows the four quadrants from these two dimensions and how they map to the rating system.
Statistical Reasoning
Hand-coded
Knowledge
Hand-coded by
people
Reasoning is by association only,
and knowledge is limited to what
the human can encode
Reasoning is reasonably sophisticated in
that the intelligence reasons about cause
and effect between variables to derive
outcomes and can answer questions about
interventions and counterfactuals.
Reasoning is by association only;
predictions are based on patterns
found in data. May extract
relationships seen by human
encoder
Combines the power of causal reasoning
with ability to learn from data
Data-learnt
Knowledge
Learnt from data
Causal Reasoning
Figure 4: ALAT Rating System Categories
Hand-coded models can often appear more sophisticated than learnt models, but deriving and maintaining
hand-coded models is time-consuming and expensive. Such models may miss relationships that a machine
learning algorithm could extract but are not clear to a human. The ultimate is to have learnt, causal models;
the weakest is hand-coded correlative models. The other two are in-between and not really comparable on
an intelligence scale per se, as their sophistication depends mostly on the underlying data or hand-coding
respectively. However, in the current environment, machine learning has certainly been used much more
successfully in many domains, including law. As such it is placed higher on the scale than hand-coded causal
approaches.
While some technologies (and some artificial intelligence techniques) can be categorised into more than
one quadrant, we believe that this simple model is suitable for the purposes of ALATs. Typically, most
artificial intelligence techniques and the technologies built on them fall into either the top-right or bottomleft quadrant. Models describing associations are not as powerful as those describing causation, so most
techniques that rely on hand-coded models explicitly encode causal relationships, meaning that
technologies in the top-left are rare. It is difficult to extract true causal relationships from data alone,
meaning that the bottom-right quadrant is also rare. Techniques to learn causal modelling exist (Pearl,
2018), but research in this area is in its infancy compared to statistical machine learning. One could
categorise recent techniques such as deep reinforcement learning into this category as at a basic level, they
have been used to capture relationships between actions in applications such as in Alpha Go (Silver at al
2017), although they are far from the sophistication found in many other causal-based techniques.
Current State of Automated Legal Advice Tools
24
Diversity
Our review has found a range and diversity of ALATs. This diversity offers opportunities for different and
even changed ways of providing “legal advice”. ALATs are at different stages of capability to provide legal
advice, from giving legal advice on their own, to being supervised or reviewed by a lawyer, to assisting and
augmenting legal advice, or unbundling legal advice into smaller discrete tasks again.
The review also found that ALATs are still new, with the majority dating from 2015 to 2017. Most are found
in the US rather than the UK and Australia, perhaps due to a greater amount of venture capital (Furlong,
2014) and a large number of ALATs are start-ups.
There is an active debate as to the type and amount of legal work that can be automated. It is argued that
different kinds of legal work are more or less “amenable” to automation with predictions ranging from 40%
to under 10% (compare for example, Deloitte, 2016; NSW Law Society, 2017; Susskind & Susskind, 2015).
When examining the task of legal advice and its multiple dimensions, different types of automation may be
able to carry out some or many subtasks. Routine aspects of the task are most likely to see automation.
With rapid technological developments more complex tasks in legal services are also being automated.
Often the path of simplicity to complexity is assumed – as it correlates with the level of complexity of
technology. However, as discussed above and shown in table 4 automation based upon a number of
factors. Amenability to automation will turn on a variety of factors, including the specific area of law, legal
subject matter and context particularly as to ambiguity, conflicting rights, open-endedness and unique facts
(Pasquale & Cashwell, 2015). The next sections of this chapter examine the categories of ALATs in greater
detail.
Current State of Automated Legal Advice Tools
25
Specialised standalone technologies
One set of technologies tend to be specialised standalone technologies that offer legal advisory outcomes
to both consumers and to lawyers. Chatbots, or legal bots, are mobile or web-based services powered by
rules using conditional and casual decision logic trees and sometimes more sophisticated AI techniques.
From 2015 early commercial and free chatbots provided legal support and advice, particularly facilitating
access to justice. Now dozens of chatbots deliver legal services mainly on a single legal issue, both for
consumers and small businesses, some to assist lawyers (Ambrogi, 2017a; Coade, 2016).
Legal apps or applications carry out analysis and apply legal logic, meaning that they have the capacity to
(often) give legal advice. While often very specific to an area of law or legal practice, is there a need for a
lawyer where such an app exists? An early adopter of legal apps was the legal aid and legal assistance
sector as a means of extending resources (K. Miller, 2015). Legal apps are also used by commercial lawyers,
for example, one corporate legal team “is using around 30 bespoke legal apps supporting a wide variety of
work, including mergers and acquisitions, contract negotiation, litigation, e-billing, and digital signatures”
(Walker quoted in Law Society of New South Wales, 2017). Others are more focused on broadening access
to the law by assisting lawyers and also creating self-help tools to deliver services directly to users. As
introduced by Georgetown University Law Center, some Australian law schools are now running electives
such as Law Apps courses and #hackathons (for example, Melbourne, Deakin, UNSW, UTS and QUT).
Virtual Assistants are an additional class of chat bots. These extend chat bots to work within law firms. The
virtual assistants provide information and advice to lawyers helping to make internal processes more
efficient and manageable.
Enablers of legal advice
ALATs are also enablers of legal advice when assisting lawyers to do their work better with intelligent
automation, as well as when targeted directly to clients. The current iteration of legal automated drafting
ALATs sees largely automated tailoring through checklists and self-help guides. Currently many smaller
offerings are limited to niche areas of law (Gardner, 2017) including a range of will drafting options (K
Miller, 2017). Other tools targeting larger law firms tend to be broader, assisting lawyers to advise on and
draft a range of simple to complex documents.
Similarly, ALATs assist lawyers to review and advise on documents. ALATs may read, review and analyse
contracts as to the risks and obligations and highlight legal issues to consider. More advanced versions use
AI and machine learning to learn from new contracts processed, such as identifying missing clauses or
clauses that are unusual. One ALAT example that reviews contracts has a disclaimer that states it does not
give legal advice.
Underlying many ALATs are algorithms or mathematical instructions that are both explicitly programmed to
calculate outcomes or, when more advanced, are designed to allow computers to learn on their own and so
create their own algorithms. These algorithms are also used to assist with sentencing or bail decisions in
court or to allocate assets in family law.
Current State of Automated Legal Advice Tools
26
A key issue is whether or not the algorithms are transparent and explainable. For example, as a commercial
service widely used in the US to assist with bail determinations, Compass Core’s Northpointe does not
make its algorithms public. Research has questioned the equity of the outcomes when using such
algorithms. Some research has shown that software helping judges decide on bail can cut crime and reduce
racial disparities when tested on over 100,000 cases (Simonite, 2017). In contrast ProPublica’s study of
7000 offenders found that Compas Core appeared to have a strong bias against black defendants (Angwin,
Larson, Mattu, & Kirchner, 2016).
Further enablers of legal advice
Two more sets of ALATs provide data legal analytics and predictive analytics. These tend to use
technologies that are in our terms smarter and more intelligent. Legal analytics ALATs use advanced
technologies such as machine learning and natural language processing to mine the volumes of data in the
litigation ecosystem. Legal analytics reveal trends and patterns in past litigation that inform legal strategy
and anticipate and predict outcomes in current cases, offering data-based answers to key questions such
as: What are our odds of winning? What are the preferences of Judge X? What tactics have defence lawyers
used in similar cases?
One significant form of ALAT uses advanced legal artificial intelligence technologies to facilitate legal
research functions. Lawyers develop advice by carrying out legal research, finding the relevant law to apply
to facts of a specific problem using legal reasoning. A leading example is the work of Ross Intelligence, built
on IBM’s Watson platform. It enables users to ask questions in natural language and will analyse the
question. The system will undertake legal research by mining the body of law – legislation, cases and
secondary materials to provide specific answers. The result takes on average a few seconds to produce
research that might take human lawyers several hours (Rinaldi, 2017). Ailira, an Australian legal artificial
intelligence also uses natural language processing and scaled machine learning to provide automated
research assistance. Focussed initially on taxation law, Ailira is expanding into other areas of work to
develop what its founder, Adrian Cartland describes as “the law firm without lawyers” (Bindman, 2017).
Current State of Automated Legal Advice Tools
27
Human-free smart contracts
Another aspect of ALATs is via the use of smart contracts to automatically form and execute contractual
relationships. Smart contracts are automated computer programs that self-execute based upon a specific
input. These contracts are contained in a decentralised distributed ledger known as the blockchain (Jehl,
2018).
Trust is the key problem that blockchain technology addresses (Würst & Gervais, 2017). This is reflected in
the distributed nature of the technology enabling the exchange of information in trustless environments.
Transactions are stored in blocks that comprise an encrypted ledger of some transactions. Blocks are
produced by computers that facilitate the transactions by performing advanced cryptographic calculations.
The blocks are collated together into chains, hence the term blockchain. Once a transaction is deemed valid
based on the consensus model the block is written to the blockchain. The blockchain is then updated across
all nodes with the new block. This provides a record of all the transactions. Changing the blockchain is
difficult as it requires the involvement of the majority nodes. Therefore, should an individual seek to rewrite the transaction it will not align with cryptographic signature of chain and therefore will be invalid
(Würst & Gervais, 2017). There is not one canonical blockchain. Instances of blockchains can be established
between nodes of computers and as a ledger can store information for a diverse range of use cases such
across finance, Internet of Things, document management, and insurance. (Crosby, Nachiappan, Verma, &
Kalyanaraman, 2016).
Smart contracts are computer programs that build upon the blockchain. One such technology is Ethereum,
which allows for the creation of applications using a complete programming language, solidity, that are
then distributed across the blockchain (Buterin, 2018). Applications are built on top of this foundation to
manage and control in a decentralised manner. Smart contracts they have more in common with computer
programs than contemporary legal contracts. They allow the execution of specified code in response to an
input. A basic example is assessment of the purported status of goods that have been transfer. If they meet
the desired quality, the transaction completes automatically.
Smart contracts are not an advice-giving technology as such, however, as they become widespread, they
may enact “a fundamental shift in the role of legal advice” (Bacina, 2017). For example, lawyers may need
to pay additional attention to the establishment of contracts that are to be encoded upon a blockchain, as
given their immutable nature deviation once created they are extremely difficult to undo. This is likely to
create new risks and responsibilities intrinsic to smart contracts, which involve different considerations and
advice, from the more conventional contracting process.
Current State of Automated Legal Advice Tools
28
Two key elements of smart contracts are relevant to legal regulation. First, following the establishment of
the agreement there is no human involvement – lawyer or client – in performance. Instead, the automated
code is designed to execute based on specific predetermined parameters. This is done without reference to
the contracting parties’ later intentions or desires. Second, decentralised blockchains remove or reduce the
need for a trusted third party (Bacina, 2017). This raises issues relating to liability relating incorrect inputs
triggering a contract action, errors and mistakes. Smart contracts are proposed for a number of industries,
with many focusing on supply chain management. In Australia, agtech company, AgriDigital is piloting of
the world’s first wheat sale using a smart contract on a blockchain ledger.
Sets of ALAT technologies
ALATs are emerging against a backdrop of broader technological disruption in the legal market. There are
now mash-ups of sets of ALAT technologies enabling NewLaw business models such as Legal Zoom and
Rocket Lawyer who use automation technology for high-volume work. Since 2017 Legal Zoom has been
using more intelligent AI for marketing in a partnership with Veritone One. At the corporate law end of the
market, Riverview Law has partnered with Liverpool University AI’s expertise. In Australia, developments
seem slower and less advanced.
A number of ALATs are developed by technology companies that work in the legal services market. Some
provide technology platforms providing multiple tools, others are more specialised. As seen in examples in
Appendix A, many of these technology companies are key players and partners in developing smart apps
and other ALATs that “encapsulate legal knowledge, reasoning and judgement to provide self-service real
time legal advice” (J. Goodman, 2016).
Current State of Automated Legal Advice Tools
29
5 Regulatory issues emerging from use of
ALATs
The impact of ALATs on regulation
With its focus on the current state of ALATs, this paper can only begin to raise some of the regulatory issues
that emerge from using ALATs for legal advice. Thus, it does not explore more general legal questions of
privacy, confidentiality or cybersecurity. These will be considered later in this project.
As can be seen in the examples in Appendix A, the current state of ALATs may create multiple challenges
for regulators of legal advice: reframing or extending paradigms, and even creating new ones. It seems that
technology is racing ahead and has the potential or capability to give automated legal advice at least in
simple specialised areas of law and, for more complex cases, at least for some of the tasks involved. Yet
legal professional regulation does not seem to be keeping pace. What consideration around regulatory risks
and barriers to access to justice and the impact of the use of ALATs has been carried out specifically for the
legal industry?
While it has been recognised that “artificial intelligence raises regulatory and ethical issues that require
investigation and guidance for solicitors” (Law Society of NSW, 2017 p4), there seems to have been limited
exploration of such issues.
This paper finds that technological innovation may both extend known regulatory issues, and also create
new regulatory concerns. Together these raise issues for oversight of ALATs, that is, for the role of
regulators.
Legal advice: just by lawyers?
A fundamental and controversial question is the role of lawyers in giving legal advice. Are lawyers enablers
of access to the law or seekers of monopoly rents? Is the current regulatory monopoly of lawyers over legal
advice justified in this technology-driven world? Underlying these crucial questions are a number of others.
Regulatory restrictions as to who can provide legal advice creates fuzzy distinctions between “legal advice”
and “legal information”. Various ALATs claim they only provide generic legal information such as draft
letters of advice (for example the DoNotPay chatbot) or have disclaimers that free legal documents are not
legal advice. Should legal industry stakeholders be able to challenge ALATs for engaging in “unauthorised
practice of law”?
The difference between advice and information is not merely technical, it is also important for public and
regulatory policy. The law, particularly case law, statutes and regulations, is a matter of public record, and
access to such legal information can be defined as a public good in its own right. While public interest
arguments can be raised against unqualified practice of law (UPL) controls, it is also not in the public
interest that access to public legal information be circumscribed in support of professional monopoly
privileges. These arguments may of course apply with less force where such information is to be embedded
Current State of Automated Legal Advice Tools
30
in commercial rather than free systems.
The Australian regime is susceptible to many of the criticisms directed at its US counterparts (see eg,
Denckla, 1999; Rhode and Ricca, 2014). For example, it could be said to have adopted and maintained a
paternalistic approach to consumer choice, a reliance on broad prescriptions that are potentially anticompetitive in effect and, perhaps, a failure to engage with the question of what regulation in the public
interest requires in a climate of declining access to law.
Should there be a withdrawing or perhaps an extension of UPL regulation, or should “the key focus” shift
from “blocking these innovations from the market, [to] using regulation to ensure that the public’s interests
are met” (Rhode and Ricca, 2014 pp 2607-8)? That in turn is a question for the empirical research that will
follow and flow out of this paper.
A recent report (IBA, 2016) has suggested that the evolution of legal services from bespoke to
commoditised and standardised or packaged services with the aid of ALATs “are likely to yield significant
benefits for consumers in terms of cost, quality and access to justice” (IBA, 2016, p5). This may require
lawyers to “overcome the conservative, risk-adverse culture that seems to pervade the profession and may
need to deconstruct their structure and pricing models” (IBA, 2016, p5).
Quality of legal advice?
One regulatory issue raised is around the quality of legal advice provided by ALATs. Whether the legal
advice is of the same quality is often debated. Some argue that technology carries out certain tasks better
than humans, particularly searching for patterns in large volumes of data (Nissan, 2017) and predictive
analysis. One recent study recorded legal experts as providing 66% accuracy compared to 70% for the
computer (Katz, Bommarito II, & Blackman, 2017). Some see access to “good enough” law as desirable
when faced with a choice between nothing and an unaffordable service (Susskind & Susskind, 2015). Others
see the risks to consumers as too significant (Law Society of NSW, 2017). How best do we balance the
competing interests involved?
Current State of Automated Legal Advice Tools
31
Substantive law?
Technological issues can also arise around ensuring the substantive law applied by the ALAT is accurate.
Who is translating the law into code that decides the output of advice? Are they legally trained? Who
assesses how often complex and labyrinthine law is accurately translated into code? (Hogan-Doran, 2017).
Who ensures the law contained within the ALAT incorporates the latest legal developments and is up-todate? Who decides and how is it decided as to the relevant context, content or weighting of different
factors? If framing decisions are made at the time of coding, how is the discretion exercised at the time of
giving the advice? (McCalman, 2017). And:
[A]lgorithms can make systems smarter, but without adding a little common sense into the equation
they can still produce some pretty bizarre results (DeAngelis, 2014).
More difficult and philosophical questions arise around advice on meaning when there is ambiguity and
uncertainty in the law. Dworkin used the metaphor of the imaginary ideal judge Hercules.
To resolve hard cases, Hercules invoked his entire ethical being, his knowledge of the principles
underpinning legal institutions in a democracy, his familiarity with legal policy, a sense of the need for
coherent doctrine and finally, his knowledge of the specific statute and case law, to approach matters
for which there was no precedent. How do we code for Hercules? (Law Society of NSW, 2017, p41).
Duty to be competent?
There is a distinction between being comfortable with technology, or a digital native, and being conversant
with how technology is used in legal advice. What does the duty to deliver legal advice competently now
require? For example, does a lawyer who provides a legal service supported by an ALAT need to
understand how that technology works, or who uses an artificially intelligent algorithm need to understand
the workings of the algorithm and the integrity of the data used (Law Society of NSW, 2017)?
The US profession has recommended extending the duty of competence incorporate some degree of
technological competence. In 2012, the American Bar Association changed the Model Rules of Professional
Conduct to recommend that a lawyer’s duty of competence include staying up-to-date with changes in
relevant technologies. At least 25 states have adopted that change with many also mandating technologyspecific learning in continuing professional development (Law Society of NSW, 2017, p41). Are we moving
to a point where, as the reliability of technologically-assisted outcomes starts to supersede human levels of
accuracy mean that failure to use technology will itself become a failure of lawyer competency?
Current State of Automated Legal Advice Tools
32
Wider professional duties?
As the examples in Appendix A show, using ALATs for assisting and giving legal advice are powerful tools.
ALATs potentially enable citizens to access quicker and cheaper legal advice, and so increase access to the
law, including access for the middle market discussed above. They can also free up lawyers from routine
and basic transactional tasks, giving lawyers more time to think through problems and advise clients. This
points to potential increases in the quality of advice, and the efficiency with which it is delivered. These
potentialities, of themselves, have larger ethical implications. If technology reduces costs, could or should
the normalisation of its use shape assessments of what is fair and reasonable charging? Longer term,
should lawyers have any ethical obligation to use ALATs if this would significantly reduce the cost and
increase efficiency and accessibility of legal service delivery?
The Black Box problem: transparency versus explanation
of legal reasoning?
A new regulatory issue for legal advice is created by the “black box problem”. This refers to the fact that
legal decisions, or support for them, may be provided by an algorithm that does not provide any reason or
explanation for this decision. The ability to give reasons is critical to sophisticated advice-giving by human
lawyers. Decisions made by opaque algorithms is “analogous to evidence offered by an anonymous expert,
whom one cannot cross-examine” (Brooks, 2017, quoting Pasquale).
The issue is that extracting and presenting reasons for AI-based decisions is a challenging task. Where logic
is causal and structural, such as with process automation, this can allow lawyers and clients to see the
reasoning and assess how the technology delivered the legal advice. However, an explanation is not as
simple as extracting a chain of causal reasoning: it needs to be presented to a person (Miller, Howe, &
Sonenberg, 2017), answer the specific question that the person has, and select the most pertinent causes
(Miller, 2018).
However, the problem becomes even more difficult with many machine learning techniques. First, most of
these techniques learn associations using statistical methods, while people present and evaluate
explanations using causality (Miller, 2018). Second, many of these techniques, such as neural networks,
learn models that are difficult even for experts to understand. Making these models transparent to nonexpert users would be a pointless exercise. Instead that they would require post-hoc explanation that
justifies decisions based on input parameters.
With smarter and more intelligent automation where machines are increasingly learning using big legal
data, and dynamically so, this is more complex, leaving big questions: How can we trace legal reasoning
logic? What legal data has been considered, seen as relevant, and how was it sourced? How did learning
occur and were there any biases? And further, deeper questions arise. What values lie within the logic?
What conscious or unconscious assumptions have been made that are not explicated? Why has the data
chosen been so chosen?
Current State of Automated Legal Advice Tools
33
This becomes more complex where the ALAT is a commercial service and the owners seek to protect their
intellectual property by keeping the logic and data confidential.
This issue was raised in 2016 in the US:
Problem: AI and automated decision-making systems are often deployed as a background process,
unknown and unseen by those they impact. Even when they are seen, they may provide
assessments and guide decisions without being fully understood or evaluated. Visible or not, as AI
systems proliferate through social domains there are few established means to validate AI systems’
fairness, and to contest or rectify wrong or harmful decisions or impacts.
Recommendation: Support research to develop the means of measuring and assessing AI systems’
accuracy and fairness during the design and development stage (Crawford & Whittaker, 2016).
This is considered such an important problem that in 2016, the European Union introduced the General
Data Protection Regulation (GDPR). This has been described by many as mandating the “right to
explanation” for any decision made about an individual where the decision can “significantly affect users”
(Goodman & Flaxman, 2016). A recent article has questioned this arguing “the GDPR lacks precise language
as well as explicit and well-defined rights and safeguards against automated decision-making, and
therefore runs the risk of being toothless” (Wachter, Mittelstadt & Floridi 2017, p 76). In order for such a
right to be useful it needs to be legally binding, either through amendment of the GDPR or through nation
states acting to provide additional protections (Wachter, et al., 2017). A central question facing ALATs is do
we need to create explainability standards and, if so, what should these look like?
Timing?
While the hype is loud as to the future potential of AI technologies, there are real concerns as to the scope
and scale of regulatory, ethical, and security issues that can, and should, be addressed now. For example, it
is foreseeable that as technology becomes increasingly ubiquitous throughout the market for legal services,
the question how digital divides within the legal industry can be managed will become more pressing.
There are also issues relating to values and ethics designed into the ALATs. These concern matters such as:
•
The values encoded into the logic;
•
Larger assumptions regarding what constitutes ‘appropriate’ professional services (such as whether
advice should routinely go beyond technical accuracy and account for competing interests and
values of clients and other interested parties); and
•
Whether low levels of explainability could actually lead to a diminished and highly instrumentalised
understanding and valuation of law itself.
Current State of Automated Legal Advice Tools
34
If these debates are not had soon, the solutions are more likely to be both ad hoc and hard-wired into the
technology. Is the first obligation of designers and adopters to “Let these debates and the rich diversity of
human values remain human” (De Clerck & De Wit, 2017; Tucker, 2016)?
The scope and role of regulation?
A further question is who should carry out the regulation and for what purpose? Currently in Australia,
legal services regulation is fragmented and sub-divided between the professional regulation of lawyers and
bodies of general law and regulation that aim to protect consumers, such as consumer protection law, tort
liability, data use and privacy. Already there are, inevitably, overlaps and discontinuities between and
within such protection and redress systems. Whether, and if so how, regulatory responsibilities and access
to consumer redress could be streamlined is a much larger question than that explored in this project.
However, it is likely that increased professional engagement with technology will create new boundary
questions for regulation.
These all point to legal services regulators as having a role and responsibility for the oversight of technology
and ALATs. More specifically, ALATs create at least three significant problems for legal services regulators:
Consistency: Currently different supervisory regimes exist for supervising automated legal information
outside the legal industry. and for automated legal information and advice within the regulated legal
industry. Is this variance appropriate?
Reach: There are many questions about the reach of regulation. For example, should there be oversight
of the technology, or is it adequate to assess only the quality of service outcomes, however achieved?
Also, how should or could regulation overcome the problem that online services may be delivered from
outside the regulator’s physical jurisdiction?
Purpose: In the context of a relatively profession-centric, as opposed to more widely market-based,
model of legal services regulation in Australia, the disruptive potential of ALAT technology raises a
fundamental challenge whether the primary function of regulation is to enable access to lawyers, or
access to the law? (Webb et al., 2017). It has been suggested that professional regulation is perceived
by some as primarily protecting lawyers’ monopoly (Law Society of NSW 2017).
Since regulators in Australia are required to “ensure efficient, effective, targeted and proportionate
regulation” (Legal Services Commissioner, 2017), we suggest that an appropriate regulatory response to all
of these is likely to require a fine balancing act between the competing interests of consumers, the legal
market, the legal profession and access to justice (Wallace, 2017).
Current State of Automated Legal Advice Tools
35
6 Conclusions and next steps
This paper has sought to identify the current state of ALAT development, to consider how ALATs are
disrupting the legal industry and creating challenges for regulators.
ALATs use a range of technologies to offer different ways of providing legal advice. ALATs exist at different
stages: some give legal advice autonomously or supervised, some assist or enable legal advice, and some
unbundle and change legal advice. Indeed, the emergent technologies such as blockchain are being used to
code legal outcomes in smart contracts that have the limited involvement humans as either lawyers or
clients.
ALATs are being greatly enabled by accelerating technology however progress is not as fast as the hype
would have it appear. Most ALATs tend to be smart and smarter rather than truly more (artificially)
intelligent. As ALATs continue to evolve and the numbers grow, this is likely to change.
The study so far points to the existence of a digital divide. Currently the majority of ALATs operate in niche
and specialised areas of law. Their use seems segmented across the legal services industry, linked to the
ability to invest. ‘Enabling’ ALATs are more concentrated within the large corporate hemisphere while the
legal assistance sector has seen a greater proliferation of lower tech apps and chatbots. The proprietary
and commercial nature of some ‘in-house’ ALATs is likely to keep costs high, and, for the present,
diminishes the potential for smaller firms to benefit from economies of scope and scale. This may change as
and when more ‘white label’ applications enter the market.
As for regulation, this paper suggests that we are now approaching a tipping point in terms of regulatory
action. ALATs are raising many regulatory issues that have not adequately been explored. These include
questions about the way technology affects the duty of competence, and how the use of technology raises
new implementation and supervision risks that need specific answers. The analysis in this paper also points
to the need to revisit the legal information/legal advice distinction. This may actually prove the greatest of
the challenges identified, raising large questions about the public interest, access to justice, and the
potentially disruptive role not just of technology but of legal services regulation itself.
This paper seeks to contribute to the emerging debate from an independent, academic perspective on
regulation and policy-making. To that extent, it both complements and is distinct from recent professioncentric contributions, such as the Law Society of NSW FLIP Inquiry (2017), the Law Institute of Victoria’s
Disruption, Innovation and Change (2015) and the Law Society of Western Australia’s The Future of the
Legal Profession (2017).
The next phase of this research is to gather the views from a range of stakeholders – members of the
profession, regulators, ALAT developers and producers and access to justice groups about the impact of
ALATs across the legal services industry. This second phase of this research will add a useful cross-section of
views exploring emerging issues, key problems, and their potential solutions.
Current State of Automated Legal Advice Tools
36
Appendix A: Examples of ALATs
This Appendix provides a selection of current examples of ALATs from the UK, US and Australia. The list is
not comprehensive but aims to capture the breadth and diversity of ALATs in the market.
Please contact the research team if you seek to have your automated legal advice tool included in any
future iterations of this paper.
The format for each Example is:
Name (Country)
Name
Country
Date created
UK, US, Australia
Website
Tech rank
Owner/Vendor
Cost
(estimate)
Free and/or $ (Commercial)
Users
Public
Lawyers
Types of tech
Examples include: natural language processing, machine learning, blockchain, decision tree
About
A short description of the tool, its purpose and objectives
Current State of Automated Legal Advice Tools
37
Legal chatbots
DoNotPay (US, UK)
Name
DoNotPay
Date created
Sep 2015
Country
US, UK
Owner/Vendor
Joshua Browder, Stanford University
student
Website
Donotpay.co.uk
Cost
Free
Tech rank
Users
Types of tech
Natural language processing, statistical machine learning, decision tree to guide the interaction.
About
The most successful and well known “robolawyer” was developed by Joshua Browder as a
student at Stanford. Called DoNotPay, it is a free chatbot that fights parking tickets. It asks a
series of questions about your case, such as: Were the signs clearly marked? Were you parked
illegally because of a medical emergency? and then the website’s algorithm generates a legal
letter that can be filed with the appropriate agency. As at June 2016, the bot had helped more
than 250,000 people challenge traffic and parking tickets in London, New York, and Seattle with
a 40% success rate.
Browder recently added new functions including to help people demand compensation from
airlines for delayed flights and file paperwork for government housing assistance. He later
expanded it to provide free legal aid to the homeless and to help refugees seeking asylum in the
US and Canada. More are on the way (Koebler, 2017).
Browder’s motivation is enabling access to the law, stating “I think it is a huge shame that those
most likely to make a mistake and get a ticket are the most vulnerable members of society —
the elderly and disabled. That’s why this is a free service” (Weather & Hale, 2016). Browder said
that he ultimately hopes to replace “25,000 exploitative lawyers” with robots that can respond
to questions with appropriate human emotions powered by artificial intelligence (Boyce, 2016).
Source: home.bt.com/tech-gadgets/tech-news/need-to-fight-a-parking-ticket-give-do-not- paysrobot-lawyer-a-call-11364041299723
LISA (UK)
Name
Robot Lawyer LISA
Date created
2017
Country
UK
Owner/Vendor
Chrissie Lightfoot, Adam Duthie,
AI Tech Support.
Website
robotlawyerlisa.com
Cost
Free
Tech rank
Users
Types of tech
Use the Neota Logic AI platform for creating code-free apps
About
A free legal tool that allows users to create legally binding Non-Disclosure Agreements (NDAs) in
less than 7 minutes at no cost. The robot, named LISA, intelligently drafts the documents while
helping them understand the legal and commercial principles on which it is based.
“Our goal is to make every day basic legal services accessible and affordable to the masses of
students, consumers and business people who are unhappy with, or overly reliant on, human
lawyers and law firms”.
Current State of Automated Legal Advice Tools
38
Lexi (Australia)
Name
Lexi the Legal Bot
Date created
Mid 2016
Country
UK, US, Australia
Owner/Vendor
LawPath
Website
try.lawpath.com.au/privacybot/
Cost
Free and/or $ (Commercial)
Tech rank
Users
Types of tech
Natural language processing, statistical machine learning, decision tree to guide the interaction,
interaction is using a form of rules, probably an ‘expert system’, to process user information that
is filtered through hundreds of rules and logical causal connections.
About
LawPath, a technology provider of online legal services for businesses, recently released Lexi as
an intelligent legal bot to “better help clients seeking customised legal documents”. As an
experimental prototype in automated delivery, Lexi provides privacy law information and
generates a free privacy policy or non-disclosure agreement tailored by user input to that user’s
needs.
The information and documents are delivered through online interactive chat. The software
combines machine learning and natural language processing principles to process user
information that is filtered through rules and logical causal connections. The end result is a
document that matches user queries. “The problem with forms, traditionally, is that they really
haven’t been able to give insight and educate as a bot does, whilst delivering the outcome as
well,” Mr Andreasen said (www.lawyersweekly.com.au/news/19105-chatbot-explores- frontiersof-legal-service),
Current State of Automated Legal Advice Tools
39
Others in the US, UK and Australia
In 2017 Stanford’s 2017 Codex FutureLaw conference ran a session on “The rise of the legal chatbots”.
Another author, Ambrogi has listed some legal chat bots in abovethelaw.com (Ambrogi, 2017b). Chatbots
designed for consumers and small businesses include:
Coralie (US), a virtual assistant chatbot that helps survivors of military sexual trauma
connect with services and resources. It recently won the Tech for Justice hackathon during
the ABA Techshow.
Docubot, a chatbot that works through lawyers’ websites to help consumers generate legal
documents and that also performs client intake.
LawDroid, a bot that helps users incorporate a business for free on a smartphone.
LawGeex LawBot, a chatbot that can be added to Slack (a cloud-based collection of
proprietary team collaboration tools and services), and then sends legal contracts for
analysis (https://blog.lawgeex.com/lawgeex-launches-first-ever-law-bot-slack/).
RentersUnion (UK) is a chatbot that provides legal advice on housing issues for residents of
London. The bot analyzes a user’s tenancy agreement and then helps generate letters or
recommend appropriate action.
Speak with Scout (Australia) is a chatbot that works through AI and also humans via
Facebook Messenger to provide legal guidance and references to a lawyer.
Other chatbots are being developed to “make lawyers’ lives easier”
Termi (UK) owned by business intelligence and analytics company Helm360 is an AI
assistant for lawyers. As Joanna Goodman explains in The Law Society Gazette, Termi
interrogates the Thomson Reuters Elite legal practice management system to request
billing and other management information.
Some law firms are developing their own chatbots.
Conveybot: a chatbot owned by UK conveyancing firm Convey Law (UK) claimed to be “the
first fully automated chatbot that can engage with conveyancing clients”, provide instant
fee quotes and then arrange a follow-up conversation with a member of the firm.
Current State of Automated Legal Advice Tools
40
Legal apps
Legal Aid NSW (Australia)
Name
Legal Aid NSW app
Date created
2015
Country
Australia
Owner/Vendor
Legal Aid NSW
Website
www.legalaid.nsw.gov.au/getlegal-help/legal-aid-nsw-app
Cost
Free
Tech rank
Users
Types of tech
Natural language processing, statistical machine learning, decision tree to guide the interaction
About
The Legal Aid NSW app gives you easy access to information about Legal Aid NSW services and
the law. It covers searching for a Legal Aid NSW service, videos about the law, workshop about
the law and access to factsheets and resources. Users can find out how to get a grant of legal aid
and what to do if they are not eligible. Plus pay their contribution towards the legal costs of their
case. Lawyers and community workers can use it to search for a private lawyer who carries out
legal aid work (www.legalaid.nsw.gov.au)
Picture It Settled (US)
Name
Picture It Settled
Date created
2014
Country
US
Owner/Vendor
Don Philbin,
Southwest Research Institute
Website
www.pictureitsettled.com
Cost
Free
Tech rank
Users
Types of tech
Big data, predictive analytics
About
Picture It Settled® helps you visualize the negotiation dance and calculate your next steps.
The Picture It Settled app is “Moneyball for negotiation”. The behavioral software has learned
negotiating patterns from parties to thousands of litigated cases in a wide variety of
jurisdictions and claim types. It uses that intelligence to make accurate predictions of where a
negotiating round is headed in time for parties to act on it using the program’s planning tools.
The planning tools allow users to fine-tune their target settlement and project what impact a
particular move might have on the round before making it. The result is more settlements on
more advantageous terms (Crunchbase).
Current State of Automated Legal Advice Tools
41
Robot Lawyers Australia (Australia)
Name
Robot Lawyers Australia
Date created
2014
Country
Australia
Owner/Vendor
Bill Doogue, Andrew George of Doogue +
George Defence Lawyers
Website
www.robot-lawyers.com.au
Cost
Free
Tech rank
Users
Types of tech
TBA
About
A free online service for people who are pleading guilty to theft, driving, assault, drug or drink/
drug driving charges before the Magistrates’ Court of Victoria. The service aims to assist users to
let the court know about circumstances relevant to sentencing. The user completes an online
interview about issues relevant to their sentence. The user may be referred to a lawyer, eg, if an
answer suggests that imprisonment is a risk. The interview includes questions asked by a lawyer
both about the office and also about the person such as whether the user has “learned anything
positive from [their] offending”. Then “robot lawyer” emails the user a document that can be
used by the magistrate, as well as copies of the answers (a Personal Instructions document) and
a Character Reference guide (K. Miller, 2017)
Neota Logic and Hive Legal Superannuation App (Australia)
Name
Hive Legal Super App
Date created
2016
Country
Australia
Owner/Vendor
Hive Legal (founded 2014)
With Neota Logic
Website
hivelegal.com.au/tools/
Cost
$
Tech rank
Users
Regulated superannuation funds
Types of tech
Formal logic and weighted criteria decision making
About
NewLaw firm Hive Legal, established in February 2014 with a focus on using technology to
provide its clients with efficient solutions, has developed the Hive Legal Super App in
conjunction with Neota Logic as a legal techspert combination. “It is part of Hive Legal’s DNA to
combine its highly sophisticated legal expertise with the power of intelligent technology like
Neota Logic” (Jodie Baker, managing director, Hive Legal) (Bullock, 2016).
Current State of Automated Legal Advice Tools
42
Helper (Australia)
Name
Legal Aid Case-Helper
Date created
2017
Country
Australia
Owner/Vendor
Bryon White, Joseph O’Neill, Rachel
Hovenden and Jessy Xie,
Melbourne Law School with Neota Logic
Website
TBA care of
www.neotalogic.com
Cost
Free
Tech rank
Users
Legal Aid Clients
Types of tech
Formal logic and weighted criteria decision making
About
Legal Aid Case-Helper is an example of a winning app developed by four law students at
Melbourne Law School. It assists clients of Victoria Legal Aid (VLA) to provide lawyer
services free of charge when the clients need help navigating the court system. The
technology was provided by Neota Logic. “It’s exciting to see how law students apply
Neota Logic’s technology to address legal and business issues in novel ways” (Julian
Uebergang, Managing Director Asia Pacific, Neota Logic).
Virtual assistants
Riverview Law (UK)
Name
Riverview Law Virtual Assistants
Date created
April 2016
Country
UK
Owner/Vendor
Riverview Law/KIM
Website
www.riverviewlaw.com/virtualassistants-2
Cost
$
Tech rank
Users
Types of tech
Natural language processing, statistical machine learning, decision tree to guide the interaction.
Interaction is using form of rules, probably an ‘expert system’ to process user information that is
filtered through hundreds of rules and logical causal connections.
About
Riverview Law has been a leading technology adapter for legal services.
In April 2016 it released two versions of Virtual Assistant that automate taking
instructions, triaging and case management processes – with a standard Foundation VAs
and a Professional VAs that is more customisable for law firms. The Virtual Assistants are
powered by KIM Technologies, a “next generation provider of software that applies
artificial intelligence capabilities to knowledge automation”.
Current State of Automated Legal Advice Tools
43
Termi (UK)
Name
Termi
Date created
April 2017
Country
UK/US
Owner/Vendor
Helm 360
Website
http://www.helm360.com/
news/introducing-termichatbot/ and
www. youtube.com
/watch?v=Tx0Vkj-Ffb4
Cost
$
Tech rank
Users
Types of tech
Natural language processing, statistical machine learning, and voice.
About
Business intelligence and analytics company Helm360 has launched Termi, as part of its business
solution offerings. Termi is a voice-activated tool that appears on the screen as an avatar. It is
focused on internal processes rather than client services.
It is designed to turn management information into actionable content, “remove complexity and
give users mobile access to relevant information without logging into multiple platforms” (Dave
quoted in Goodman, 2017).
It is built on Microsoft’s cognitive services platform that interrogates Thomson Reuters
legal practice management systems via, eg, Skype, Team or a web browser (Goodman,
2017).
FTA Portal (Australia)
Name
FTA Portal
Date created
April 2017
Country
Australia
Owner/Vendor
Data61 with Australian Department of
Foreign Affairs and Trade
Website
ftaportal.dfat.gov.au
Cost
$
Tech rank
Users
Export Farmers
Types of tech
Natural language processing, Information retrieval (probably with statistical machine learning)
About
In Australia, DFAT has worked with data innovation company, Data 61, to make Free Trade
Agreements accessible via a pilot Free Trade Agreement Portal. Hosted on the DFAT website, it is
a service that guides farmers on the 900-page legal regulations around exporting goods from
Australia.
“You can type it in plain English and the system will give you an answer. It’s not the same
service you’d get from a $500-an-hour lawyer, but it’s a starting point. It tells you
enough”. This means you don’t need to know that beef is officially called ‘meat of bovine
animals’ to get you started (www.data61.csiro.au).
Current State of Automated Legal Advice Tools
44
Legal document automation
Exari DocGen (US)
Name
Exari DocGen
Date created
1999
Country
Australia (also US, UK, EU)
Owner/Vendor
Justin Lipton, Jamie Wodetzki of Exari
Systems
Website
www.exari.com
Cost
Tech rank
Users
Types of tech
To advise
About
Exari Systems is an automated document assembly and contract automation software
company. Exari software uses a web-browser interface to enable business people to create
their own contracts using the templates created by their legal department. This approach
accelerates the contracts process and reduces the burden on busy legal departments.
Exari DocGen uses an intuitive, browser-based interview to generate any type of contract. To
gather the necessary contract information, users simply answer a dynamic set of questions,
instinctively grouped by topic. This information then populates the contract template, creating a
compliant, accurate contract in record time. And with Exari’s authoring tools, it’s never been
easier for companies to create the contract templates that power the DocGen engine.
Clerky (US)
Name
Clerky
Date created
2011
Country
US
Owner/Vendor
Chris Field, Darby Wong
Website
www.clerky.com
Cost
$
Tech rank
Users
Types of tech
To advise
About
Offering automated legal paperwork for start-ups, Clerky currently provides company
incorporation documents and is privately trialling automated fundraising and commercial
documents (such as non-disclosure agreements and employment agreements). Currently only
US (Gardner, 2017).
Clerky privides the easiest way for startups to get legal paperwork done safely. We’re 100%
focused on helping startups get legal paperwork done safely, going far beyond simply providing
forms. Get your legal paperwork done with confidence, so you can get back to building your
company.
Current State of Automated Legal Advice Tools
45
Legal document review
Kira Systems (Canada)
Name
Kira Systems
Date created
2015
Country
Canada
Owner/Vendor
Alexander Hudek, Noah Waisberg
Alliance with Deloitte, KPMG, etc.
Website
kirasystems.com
Cost
$
Tech rank
Users
Types of tech
Combines natural language processing and machine-learning. The system responds to queries
without requiring particular terminology, and its output becomes increasingly accurate as it
learns from experience and feedback.
About
Kira Systems undertakes mergers and acquisitions, due diligence and – for this category - contract
analysis.
Seal (US)
Name
Seal
Date created
2010
Country
US
Owner/Vendor
Kevin Gidney, Ulf Zetterberg
Website
www.seal-software.com
Cost
$
Tech rank
Users
Types of tech
Uses machine learning, deep learning and AI to find contracts and extract data.
About
Seal is a leading provider of Contract Discovery and Analytics. Contract Discovery and Analytics
software is a class of artificial intelligence technology purpose-built and extensively trained to
discover and extract critical data from legal documents.
Having a clear vision into your contractual relationships is critical. Knowing the risks, obligations,
and opportunities with third parties is required for dealing with business, legal, or regulatory
changes. This visibility is also a valuable and untapped source of intelligence for improving overall
business performance. Our software tells you where all your contracts are and what they contain,
at scale.
Current State of Automated Legal Advice Tools
46
Contract Probe (Australia)
Name
Contract Probe
Date created
2017
Country
Australia
Owner/Vendor
Michael Pattison Funded by Mills Oakley
Website
www.contractprobe.com
Cost
$
Tech rank
Users
Types of tech
Algorithm, machine learning, and AI.
About
ContractProbe reviews a contract and generates a report within 15 seconds, providing a quality
“score”. Clients could use the score as a benchmark to pass through some contracts without
human oversight at all. It has been “trained” on thousands of executed non-disclosure
agreements, consultancy agreements, employment agreements and technology licences.
It also has an artificial intelligence “front-end” capable of learning from new contracts it
processes, so that it can identify missing clauses or unusual clauses. “It goes so far as to suggest a
sample clause where it knows one is missing, but that just goes into the report rather than the
document itself, we don’t want to cross the line into providing legal advice,” said Mr Pattison.
Early testing by lawyers revealed a 40% increase in efficiency by those using ContractProbe
against a control group (Bailey, 2017).
LawGeex LawBot (US)
Name
LawGeex LawBot
Date created
2017
Country
US
Owner/Vendor
Noory Bechor and Ilan Admon
Website
www.lawgeex.com
Cost
$
Tech rank
Users
Types of tech
Machine learning, text analysis and natural language processing.
About
The LawGeex Artificial Intelligence engine reviews, reads and analyses incoming contracts, and
highlights any issues suggesting edits based on a company’s pre-defined legal policies.
Contracts that meet these policies can be automatically approved within an hour. Contracts that
don’t align with policies are escalated to a human for guided editing and approval.
Current State of Automated Legal Advice Tools
47
Legal artificial intelligence
ROSS Intelligence (Canada/US)
Name
ROSS Intelligence
Date created
2014
Country
Canada/US
Owner/Vendor
Jimoh Ovbiagele, Andrew Arruda, Pargles
Dall’Oglio, Akash Venkat
Website
www.rossintelligence.com
Cost
$
Tech rank
Users
Types of tech
Natural language processing + statistical machine learning. Built on IBM’s Watson artificial
intelligence that uses cloud-based predictive analytics and cognitive computing services
embedded in Watson’s analytics engine.
About
Ross Intelligence does more than humanly possible. The tool supercharges lawyers with artificial
intelligence.
ROSS, promoted as the world’s first artificially intelligent lawyer”, is built on IBM’s Watson
artificial intelligence. It is designed to perform legal research while approximating the experience
of working with a human lawyer. It can understand questions asked in natural language, analyses
the questions, then goes through the body of law to provide specific, analytic answers. On
average, it takes a few seconds (Rinaldi, 2017). In the past year, more than 10 major law firms
have “hired” Ross (see the Ross Website).
Ailira (Australia)
Name
Ailira
Date created
201?
Country
Australia
Owner/Vendor
Adrian Cartland, Cartland Law
Website
www.Ailira.com
www.CartlandLaw.com/Ailira
Cost
$
Tech rank
Users
Types of tech
Natural Language processing + statistical machine learning
About
Ailira is developed in Australia and named Ailira after the Artificially Intelligent Legal Information
Research Assistant. Ailira automates legal advice (for consumers) and automates legal research
(for lawyers). Currently the focus is Australian federal tax research, extending to assist victims of
domestic violence having won the D3 challenge (see the Ailira website).
From November 2017, Ailira is being used to assist a Darwin, Northern Territory law firm to draft
virtual wills (Marks, 2017).
Current State of Automated Legal Advice Tools
48
Ravn ACE (UK)
Name
Applied Cognitive Engine (ACE)
Date created
2010, AI in 2015
Country
UK
Owner/Vendor
Peter Wallqvist
Website
www.ravn.co.uk
Cost
$
Tech rank
Users
Types of tech
Natural Language processing and Machine Learning areas of artificial intelligence
About
“We derive structure from chaos”.
Ravn uses artificial intelligence analytical ACE platform technology which reads, organises,
discovers and summarises unstructured data such as legal documents. It adds specific business
solutions on top of the ACE platform such as compliance.
In 2017 Ravn launched an analytical tool to predict and forecast outcomes by analysing historical
data. A typical use is the ability to predict the cost of legal matters and other types of projects
using AI algorithms on data surfaced using the ACE platform (AI Business, 2017). Note: RAVN’s
Extract is a plug-and-play version of the ACE product.
Legal algorithms
Compas Core (US)
Name
COMPAS CORE
Date created
?
Country
US
Owner/Vendor
Northpointe (owned by Equivant)
Website
www.equivant.com/
solutions/inmateclassification
Cost
$
Tech rank
Types of tech
Users
Judges
This company keeps the details of their algorithm private as commercial.
Must be using statistical machine learning to do these predictions.
About
In many US states, judges use software called COMPAS CORE to help with setting bail and
deciding whether to grant parole for offenders recently removed from or currently in the
community e.g. jail, probation, community corrections, etc. (www.equivant.com).
A software algorithm develops a score to predict the risk an offender will commit a new violent
crime, be likely to re-offend or be a flight risk. The software uses information from a survey with
over 137 questions ranging from informing demographics such as gender, age, criminal history,
and personal relationships (although not race), asking questions such as “Was one of your
parents ever sent to jail or prison?” “How many of your friends/acquaintances are taking drugs
illegally?” and “How often did you get in fights at school?” and requests you to agree/disagree
with statements such as “A hungry person has a right to steal” (Angwin et al., 2016).
Current State of Automated Legal Advice Tools
49
Divorce Right (Australia)
Name
Divorce Right
Date created
2015
Country
Australia
Owner/Vendor
Anne-Marie Cade, owner Victorian
online legal firm Daniel Lew Le Mercier &
Co
Website
divorceright.com.au
Cost
$
Tech rank
Users
Divorce clients, divorce lawyers
Types of tech
About
Australian startup DivorceRight focuses on the divorce process, aiming to make it as easy as
possible and keep families out of court. It does this by completing as much of the divorce
process online. DivorceRight uses an algorithm to assist in the division of property. The
algorithm takes into account how each party ranks a list of non-allocated marital property from
most to least preferred and divides it (Baldassarre, 2015).
Legal data analytics and prediction
Lex Machina (US)
Name
Lex Machina
Date created
2009
Country
US
Owner/Vendor
Founded by George Gregory & Joshua
Walker (ex Stanford Uni)
Purchased by LexisNexis in 2015
Website
lexmachina.com/legal-analytics
Tech rank
Cost
Users
$
Law firms, leading companies,
consultants, public interest users
Types of tech
Natural language processing and statistical machine learning.
About
Lex Machina offers “moneyball lawyering” (Salian, 2017). It applies natural-language processing
technology developed at Stanford University to the big data of millions of court decisions and
other legal information to find strategic insights on judges, lawyers, parties, and more. This
allows law firms and companies to predict the behaviours and outcomes that different legal
strategies will produce, enabling them to win cases and close business (see the website).
Originally using intellectual property litigation data and predictive analytics, now owned by
LexisNexis, it has broadened to delivery as Software-as a-Service. As an example, its case
analysis tool can show which judges tend to favour plaintiffs, summarize the legal strategies of
opposing lawyers based on their case histories, and determine the arguments most likely to
convince specific judges.
Current State of Automated Legal Advice Tools
50
Ravel (US)
Name
Ravel
Date created
May 2012
Country
US
Owner/Vendor
Founded by Nik Reed, Daniel Lewis (ex
Stanford Uni)
In 2012, Ravel spun out of Stanford
University’s Law School, Computer
Science Department, and d.school, with
the support of CodeX (Stanford’s Center
for Legal Informatic
Purchased by LexisNexis 2017
Website
ravellaw.com/
Tech rank
Cost
$
Users
Law firms, law students
Types of tech
Natural language processing and statistical machine learning.
About
Ravel Law is a legal search, analytics, and visualization platform. Ravel assists lawyers to find,
contextualize, and interpret information that turns legal data into legal insights. Ravel’s sets of
tools include data-driven, interactive visualisations and analytics, enabling lawyers to better
understand trends, the law and prepare for litigation (http://ravellaw.com/next-step-in-ourjourney).
Premonition (US)
Name
Premonition
Date created
March 2014
Country
US, UK
Owner/Vendor
Toby Unwin, Guy Kurlandski
Website
www.premonition.ai
Cost
$
Tech rank
Users
Lawyers, Courts
Types of tech
Statistical machine learning
About
“Which attorneys win before which judges? Premonition knows”.
Premonition also uses predictive legal technology, accessing data from litigation databases. It
can predict the winner of a case before it goes to court, based on statistical analyses of verdicts
in similar cases.
It’s obvious, but nobody does it. People pick attorneys based on recommendations from friends,
online reviews, because they’re friends, friends of friends, went to a particular law school, have
nice offices, work for a well-known firm, saw an advertisement, their name was first in the
phone book, etc. Years ago, we started collecting lists of “the best” attorneys in various
specialties. Recently, using the Premonition system, we took a look at the leading lists, people
who are recognized by their peers as being “the best”. It turns out they’re average. The only way
that they stood out was a disproportionate number of appeals and re-opened cases, i.e. they’re
good at dragging out litigation (see the website).
Current State of Automated Legal Advice Tools
51
Human-free smart contracts
Agridigital (Australia)
Name
AgriDigital
Date created
2017
Country
Australia
Owner/Vendor
Emma Weston, Bob McKay, Ben Reid
Website
www.agridigital.io
Cost
Free and/or $ (Commercial)
Tech rank
Users
Types of tech
Blockchain, statistical machine learning, and a decision tree to guide the interaction.
About
“AgriDigital software solutions simplify commodity management, revolutionise supply chain
finance, and bring traceability to your agribusiness”.
AgriDigital ran a pilot of the world’s first wheat sale using a pilot blockchain ledger and smart
contract code (Bacina, 2017).
AgriDigital software platforms are designed to assist in the transaction and settlement of
agricultural commodities and to manage supply chain risk. Through applied blockchain
technologies, distributed ledgers and smart contracts, AgriDigital provides “real time payment to
growers, increased efficiencies for brokers, flexible supply chain for buyers and financiers, and
paddock to plate transparency for consumers”.
Current State of Automated Legal Advice Tools
52
NewLaw business models
Legal Zoom (US)
Name
Legal Zoom
Date created
Dec 1999/2001(?)
Country
US & other countries
Owner/Vendor
Eddie Hartman, Brian Liu, Brian Lee
Website
www.legalzoom.com/country/
au (Australia)
Cost
Affordable
Tech rank
Users
Individuals, families,
and small businesses
Types of tech
Automation technology
About
“LegalZoom is a trusted technology platform giving access to professional legal advice, so
people can protect what matters most”.
We didn’t start out to be disruptive,” says John Suh, LegalZoom’s CEO. “We were set up to fix a
problem. The legal system was broken, and too many people were frozen out of it” (Brescia).
LegalZoom provides high quality online legal document services and affordable legal plans with
access to experienced attorneys to individuals, families and small businesses. Its goal is to
provide access to the legal system for the 84% of Americans and the vast majority of small
businesses who can’t afford an attorney and do not qualify for free legal services.
LegalZoom is now in its third “chapter”. The LegalZoom model specialises in identified areas of
law where written information, forms, and legal advice can serve a large number of clients,
meaning they can build economies of scale and bring the cost of services down considerably,
“often 80% lower than what a lawyer would charge” (Brescia p762). Suh sums up the LegalZoom
business model as involving:
“Scale (including in mass market advertising), lots of volume within a specialized area, high
repetition, leveraging lawyers that are extremely well-versed in that particular field and
developers that know how to codify the law within the technology, so we can deliver the same
experience each and every time … We’re really about technology-enabled lawyers. And we think
that’s the future” IBA 54.
Current State of Automated Legal Advice Tools
53
Rocket Lawyer (US)
Name
Rocket Lawyer
Date created
August 2008
Country
US
Owner/Vendor
Charley Moore
Website
www.RocketLawyer.com
Cost
Free to $
Tech rank
Users
Types of tech
Automation technology
About
Legal made simple.
Founded in August 2008, Rocket Lawyer aims to provides simple and affordable online legal
services for everyone. Since then, the company has helped over 20 million families and small
businesses take care of their legal matters. Users can get free legal documents, seek legal advice,
create legal documents online, have documents reviewed for packaged and discounted rates
with attorneys (www.crunchbase.com/organization/rocketlawyer).
At Rocket Lawyer, we want to change things by making legal services affordable, simple and
available to more people than ever before. We combine free legal documents and free legal
information with access to affordable representation by licensed attorneys.
Riverview Law (UK)
Name
Riverview Law
Date created
2012
Country
UK
Owner/Vendor
Riverview Law
Website
www.riverviewlaw.com
Cost
$ For profit
Tech rank
Types of tech
Users
Multiple using the KIM (Knowledge Information Meaning) platform.
As for AI, Riverview Law has partnered with Liverpool University to leverage its AI expertise since
2015. Again, in 2015 it purchased CliXLEX, an AI company based in the US, to develop its AI
virtual assistants and KIM platform.
About
In 2012 the company entered the market with a bang, promising to be a “legal disruptor” that
would do things differently.
Riverview Law has a fixed priced model for services to provide customers with budget certainty.
The business model is built from the customer up, not the law firm partner down, with the
strapline – ‘Legal input. Business output.’ – which tells clients what they can expect from us;
high quality legal advice and support provided in the context of their business, risk appetite and
their tactical and strategic commercial drivers. All of which are underpinned by talented people,
effective processes, scalable technology and a customer-centric culture.
KIM has several levels as an AI platform based on IBM Watson. It builds workflows being a
service software (as compared to Software-as-a-Service), manages workload and, on an
enterprise level, supports expert decision-making (Goodman, 2016).
Current State of Automated Legal Advice Tools
54
Legal Vision (Australia)
Name
Legal Vision
Date created
2012
Country
Australia
Owner/Vendor
Lachlan McKnight, Ursula Hogben, Evan
Tait-Styles
Part-owner Gilbert & Tobin in 2016
(20%)
Website
legalvision.com.au
Tech rank
Cost
$ fixed-fee
Users
Originally small and medium businesses,
also corporations and in-house legal
teams
Types of tech
Machine learning and automated document technology.
About
LegalVision was born in 2012 as an online legal documents business that enabled users to build
their own documents online. Over 40 documents are free at the website, more complex and
tailored versions see fixed-fee legal costs.
In 2014 Legal Vision adapted its model to become an incorporated legal practice (ILP) that is a
“full service law firm that uses technological innovations to deliver legal services to clients in an
efficient, cost effective way”. It continues to have a strong technology strategy that includes
investing in machine learning, document technology, various apps and blockchain (Law Society of
NSW, 2017 p49/51).
“We don’t subscribe to the view that technology is going to replace lawyers. Rather, we think it
will free up time for lawyers to focus on the ‘value add’ aspects of their role”.
Current State of Automated Legal Advice Tools
55
Legal technology companies
More examples of legal technology companies serving the legal sector will be added in due course.
Neota Logic Systems (US)
Name
Neota Logic System (NLS)
Date created
2010
Country
US, UK, Australia
Owner/Vendor
Co-founder Michael Mills
APAC Julian Uebergang
Website
www.neotalogic.com
Tech rank
Types of tech
Cost
$
Users
Formal logic and weighted criteria decision-making.
NLS consists of an AI-powered platform and toolset.
About
Re-imagine Professional Services with Artificial Intelligence
NLS consists of an AI-powered platform and comprehensive toolset that allows professionals to
rapidly build and deploy application solutions that automate their expertise, increasing
productivity, improving client satisfaction and creating new business opportunities.
Neota Logic is a global provider of intelligent software for the legal and compliance industries. It
has developed a number of smart applications by combining rules, reasoning, decision
management, and document automation, thus enabling business solutions that deliver process
improvements, reduce risk, and ensure compliance.
The heart of NLS is our unique, proprietary hybrid reasoning engine. The Engine combines
Boolean rules of any complexity with mathematical reasoning (formulas, Excel spreadsheets),
multi-factor reasoning (weighted scorings), and a wide range of external tools, as illustrated. The
Engine automatically integrates all these forms of reasoning—calling upon them as needed to
drive an application and solve a problem. Authors can build very large and complex applications
in individual, small, comprehensible segments that are easier to create and much easier to
maintain. The Engine is designed and configured at Amazon Web Services for security, fault
tolerance, high availability, and easy scaling to assure good performance at high levels of use
(www.neotalogic.com/platform/reasoning-engine).
Current State of Automated Legal Advice Tools
56
Data 61 (Australia)
Name
Data61
Date created
2016
Country
Australia
Owner/Vendor
CSIRO
Website
www.data61.csiro.au
Cost
Free & $
Tech rank
Users
Types of tech
TBA
About
Data61 is focused on “creating our data-driven future”. It combines the CSIRO Digital
Productivity and NICTA teams to build a data-focused research and innovation powerhouse. It
provides a network of capabilities, addressing key growth areas for a data-focused world
including: Autonomous systems, Computer vision, Data analytics, Digital economy, Machine
learning, Mobile systems, Optimisation, Software systems, Wireless and networks
Data61 is working on exemplar applications, to help cut through dense laws and regulations that
contribute to the $250 billion annual cost of red tape, with various organisations including the
Australian Taxation Office and the Australian Building Codes Board. One app developed with
PwC compares staff pay to awards and agreements to rapidly highlight discrepancies (Walsh,
2017).
A key project is Regulation as a Platform as a proof-of-concept project that aims to maximise the
value of regulation, being the key data set of government. “We’re re-imagining regulation as an
open platform based on digital logic to help support a growing ecosystem of digital regulation
tools and services.” Under the National Innovation and Science Agenda: Platforms for Open Data
framework, the Australian Government is exploring opportunities to maximise the value of
public data for the benefit of all Australians. Data61 is working with government stakeholders to
transform their rules into digital logic. It is a multi-stage process. It starts with parsing large
amounts of legal text, automatically converting as much as possible into machine-readable logic.
Then policy experts and regulators provide oversight of the digital logic to ensure the intent of
the law is accurately represented. After quality checking, the rules are endorsed for publication
by regulators and made publicly available on the Regulation as a Platform prototype. The
ultimate aim is to provide free and open access to legislation and regulation via public APIs,
which will allow users to access the database of endorsed logic rules and a reasoning engine to
process rules and data into accessible digital logic.
Current State of Automated Legal Advice Tools
57
Appendix B: Technology definitions
Key definitions and terms used in this technological arena are explained in this section.
Terminology
Definitions & Source
Artificial Intelligence
AI is an “umbrella” concept that is made up of numerous technologies. Used in a
general sense, it refers to the development of programs that can teach themselves to
learn, understand, reason, plan, and act (that is, become more “intelligent”) when
exposed to new data in the right quantities (Huff Eckert et al., 2016).
Artificial Intelligence –
Soft
Also called ‘weak-AI’ or ‘narrow-AI’, is AI built for a specific domain with the capability
of intelligent decisions of that context.
Artificial Intelligence –
Hard
The most difficult AI problems are referred to as AI-complete or AI-hard implying the
difficulty level of these computational problems is equivalent to solving the central AI
problem of making computers as intelligent as people (AKA strong AI). To call a
problem AI-complete means it would not be solved by a simple specific algorithm
(en.wikipedia.org/wiki/AI-complete).
Artificial Intelligence –
Strong
Artificial intelligence where computers are as intelligent as people and that aims to
duplicate human intellectual abilities (Copeland, 2017).
Algorithms
Algorithms are mathematical instructions. An algorithm “is a step-by-step procedure
for calculations” (Deangelis, 2014). That is, they take some values as an input and
produce values, as output. An algorithm is “a sequence of computational steps that
transform the input into the output” (Cormen, Leiserson, Rivest & Stein, 2009, p5). It
is an unambiguous specification of how to solve a class of problems.
Algorithms can perform calculation, data processing and automated reasoning tasks.
An algorithm is an effective method that can be expressed within a finite amount of
space and time and in a well-defined formal language for calculating a function.
Algorithms are instructions that a computer uses to transform a set, or sets of data,
into an output. Algorithms are the basic techniques
However now, rather than follow only explicitly programmed instructions, some
computer algorithms are designed to allow computers to learn on their own (i.e.,
facilitate machine learning) or to construct small programs on their own that work
just for a specific context, are run once, and discarded immediately. Uses for machine
learning include data mining and pattern recognition (Deangelis, 2014).
Current State of Automated Legal Advice Tools
58
Terminology
Definitions & Source
Analytics – Descriptive to
Prescriptive
This diagram shows how analytics are developing from descriptive and diagnostic to
predictive and even prescriptive analytical techniques. There are four types of big
data analytics in a chain of evolution from descriptive to diagnostic to predictive, and
culminating with prescriptive (Gartner, 2017):
•
Descriptive – What is happening now based on incoming data.
•
Diagnostic – A look at past performance to determine what happened and why.
•
Predictive – An analysis of likely scenarios of what might happen.
•
Prescriptive – This type of analysis reveals what actions should be taken. Many
organizations are still in the descriptive stage.
(Anadiots 2016).
App (abbreviation for
Application)
Big Data
A small, specialised software program that is designed to perform a specific function
for the user. It can run on the internet, your computer or as a download on mobile
devices.
Big data is a term for data sets that are so large or complex that traditional data
processing application software is inadequate to deal with them. Big data challenges
include capturing data, data storage, data analysis, search, sharing, transfer,
visualisation, querying, updating and information privacy. Lately, the term “big data”
tends to refer to the use of predictive analytics, user behaviour analytics, or certain
other advanced data analytics methods that extract value from data, and seldom to a
particular size of data set (Mayer-Schönberger & Cukier 2013).
Current State of Automated Legal Advice Tools
59
Terminology
Definitions & Source
Blockchain
A blockchain is a decentralised distributed ledger (Jehl, 2018). The blockchain records
transactions between two parties in a verifiable and permanent manner. When used
as a distributed ledger, a blockchain is typically deployed in a distributed manner.
With the network collectively adhering to a protocol for validating new blocks. Once
validated these blocks are added to the chain. Once recorded, the data in any given
block cannot be altered retroactively without the alteration of all subsequent blocks,
which requires collusion of the network majority.
The first distributed blockchain was conceptualised in 2008 by an anonymous person/
group known as Satoshi Nakamoto and implemented in 2009 as a core component of
bitcoin where it serves as the public ledger for all transactions. (Tapscott & Tapscott,
2016)
Blockchains are potentially suitable for many records management activities, such as
identity management, transaction processing, documenting provenance, or food
traceability. Tapscott sees that blockchain technology has the potential to
revolutionize the world economy (McKinsey & Company, 2016; Tapscott & Tapscott,
2016).
Bot / Chat Bot
A chatbot (also known as a talkbot, chatterbot, Bot, IM bot, interactive agent, or
Artificial Conversational Entity) is a computer program which conducts a conversation
via auditory or textual methods. Such programs are often designed to convincingly
simulate how a human would behave as a conversational partner, thereby passing the
Turing test. Chatbots are typically used in dialog systems for various practical
purposes including customer service or information acquisition. Some chatterbots use
sophisticated natural language processing systems, but many simpler systems scan
for keywords within the input, then pull a reply with the most matching keywords, or
the most similar wording pattern, from a database.
The term “ChatterBot” was coined by Michael Mauldin in 1994 to describe these
conversational programs (Mauldin, 1994).
Cognitive Computing
At present, there is no widely agreed upon definition for cognitive computing in
academia or industry.
Cognitive computing (CC) is a term used by IBM to refer to new hardware and/or
software that mimics the functioning of the human brain (2004) and helps to improve
human decision-making. In this sense, it is a new type of computing with the goal of
more accurate models of how the human brain/mind senses, reasons, and responds
to stimulus. CC applications link data analysis and adaptive page displays (AUI) to
adjust content for a particular type of audience. As such, CC hardware and
applications strive to be more affective and more influential by design (Kelly, 2016).
Data visualisation
The ability to visually display and interact with date for analysis and communication
(Smith 2016).
Current State of Automated Legal Advice Tools
60
Terminology
Definitions & Source
Deep Learning
Deep learning is part of a broader family of machine learning methods based on
learning data representations, as opposed to task-specific algorithms. Learning can be
supervised, partially supervised or unsupervised.
also known as Deep
structured learning or
Hierarchical learning
Some representations are loosely based on interpretation of information processing
and communication patterns in a biological nervous system, such as neural
coding that attempts to define a relationship between various stimuli and associated
neuronal responses in the brain. Research attempts to create efficient systems to
learn these representations from large-scale, unlabelled data sets.
Deep learning architectures such as deep neural networks, deep belief networks and
recurrent neural networks have been applied to fields including computer vision,
speech recognition, natural language processing, audio recognition, social network
filtering, machine translation, bioinformatics and drug design where they produced
results comparable to and in some cases superior to human experts (Bengio, LeCun,
and Hinton, 2015).
Distributed ledger
A distributed electronic ledger uses software algorithms to record and confirm
transactions with reliability and anonymity. The record of events is shared between
many parties and information once entered cannot be altered, as the downstream
chain reinforces upstream transactions (Huff Eckert et al., 2016).
Intelligence automation
Intelligent automation—the combination of artificial intelligence and automation.
Intelligent automation systems sense and synthesize vast amounts of information and
can automate entire processes or workflows, learning and adapting as they go.
Applications range from the routine to the revolutionary: from collecting, analysing,
and making decisions about textual information to guiding autonomous vehicles and
advanced robots (Schatsky & Mahidar, 2014).
Internet of Things (IoT).
Network of objects — devices, vehicles, etc. — embedded with sensors, software,
network connectivity, and computing capability, that can collect and exchange data
over the Internet. The term IoT has come to represent any device that is now
“connected” and accessible via a network connection (Huff Eckert et al., 2016).
IBM Watson
Referred to as a ‘supercomputer’, Watson’s underlying cognitive computing
technology combines sophisticated analytical software and artificial intelligence to
answer questions posed in natural language (IBM, 2018).
Machine learning
Machine learning is a field of computer science that gives computers the ability to
learn without being explicitly programmed. This learning in turn enables them to
improve their performance over time on specific tasks (Surden, 2014).
Machine learning focuses on the development of programs that can teach themselves
to learn, understand, reason, plan, and act (i.e., become more “intelligent”) when
exposed to new data in the right quantities.
Machine learning is a method of data analysis that automates analytical model
building. Using algorithms that iteratively learn from data, machine learning allows
computers to find hidden insights without being explicitly programmed where to look
(SAS, 2018).
Current State of Automated Legal Advice Tools
61
Terminology
Definitions & Source
Natural language
processing
Natural Language Processing (NLP) is a field of artificial intelligence, computational
linguistics and computer science that concerns the interactions between computers
and human or “natural” languages. NLP involves computer programs that effectively
process the large database of natural languages to understand human
communication and intention, such as how to answer questions asked as if to a
person (Manning & Schütze, 1999).
(Artificial) Neural
networks
Artificial neural networks (ANNs), a form of connectionism, are computing systems
inspired by the biological neural networks that constitute animal brains. Such systems
learn (progressively improve performance), to do tasks by considering examples,
generally without task-specific programming. For example, in image recognition, they
might learn to identify images that contain cats by analysing example images that
have been manually labelled as “cat” or “no cat” and using the analytic results to
identify cats in other images. They have found most use in applications difficult to
express in a traditional computer algorithm using rule-based programming (Graupe,
2013).
Robots
Electro-mechanical machines or virtual agents that automate, augment or assist
human activities, autonomously or according to set instructions — often a computer
program (Huff Eckert et al., 2016).
Smart contracts
Computer scientist Nick Szabo first posited the idea of smart contracts in 1994, as ‘a
computerized transaction protocol that executes the terms of a contract’ (Tapscott &
Tapscott, 2016).
Smart contracts facilitate and permit trusted transactions and agreements to be
carried out among anonymous, disparate parties without the need to have central
authority, a legal system or external enforcement mechanisms. As computer
protocols that facilitate, verify or enforce the negotiation or performance of a
contract, they allow traceable, transparent and irreversible transactions. The code
and agreements in smart contracts exist across a distributed, decentralized network
called a blockchain.
Superintelligence
By a “superintelligence” we mean an intellect that is much smarter than the best
human brains in practically every field, including scientific creativity, general wisdom
and social skill.
Superintelligence is currently hypothetical. If superintelligence occurred, it would see
machines that are able to create new machines more intelligent than themselves, ad
infinitum (Bostrum, 2014).
Third platform
The third platform is a term used by IDC to describe the four sets of interconnected
technologies – mobility, big data analytics, cloud and social media - that also connect
with AI (IDC, 2018).
Current State of Automated Legal Advice Tools
62
Terminology
Definitions & Source
Turing test
Developed by Alan Turing in 1950 in his paper “Computing Machinery and
Intelligence” (Turing, 1950), this is a test of a machine’s ability to exhibit intelligent
behaviour equivalent to, or indistinguishable from that of a human. Turing proposed
that a human evaluator would judge natural language conversations between a
human and a machine designed to generate human-like responses. The evaluator
would be aware that one of the two partners in conversation is a machine, and all
participants would be separated from one another. The conversation would be
limited to a text-only channel such as a computer keyboard and screen, so the result
would not depend on the machine’s ability to render words as speech. If the
evaluator cannot reliably tell the machine from the human, the machine is said to
have passed the test. The test does not check the ability to give correct answers to
questions, only how closely answers resemble those a human would give.
Turing’s paper opens with the words: “I propose to consider the question, ‘Can
machines think?’” Because “thinking” is difficult to define, Turing chooses to “replace
the question by another, which is closely related to it and is expressed in relatively
unambiguous words.” Turing’s new question is: “Are there imaginable digital
computers which would do well in the imitation game?” This question, Turing
believed, is one that can actually be answered. In the remainder of the paper, he
argued against all the major objections to the proposition that “machines can think”
(Turing, 1950).
Current State of Automated Legal Advice Tools
63
Appendix C: References
AI Business (2017). How is RAVN ACE’s AI Bringing Order to the Legal Sector?. Retrieved from aibusiness.com/ravnaces-ai-bringing-order-legal-sector/
Ambrogi, R. (2017, April 17). This week in legal tech: Everyones talking about chatbots. Above The Law. Retrieved from
abovethelaw.com/2017/04/this-week-in-legal-tech-everyones-talking-about-chatbots
Anadiotis, G. (2016, November 23). Data to analytics to AI: From descriptive to predictive analytics. ZD Net. Retrieved
from www.zdnet.com/article/data-to-analytics-to-ai-from-descriptive-to-predictive-analytics/
Angwin, J., Larson, J., Mattu, S., & Kirchner, L. (2016, May 23). Machine Bias. ProPublica. Retrieved from
www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing
Attorney-General v Quill Wills Ltd [1990] WASC 604
Bacina, M. (2017). Smart contracts in Australia: Just how clever are they? Bulletin (Law Society of South Australia),
39(10) 22-23. Retrieved from search.informit.com.au/documentSummary;dn=301846892418302;res=IELHSS
Bailey, M. (2017, March 30). Ex-Allens partner gets Mills Oakley backing for legal startup. AFR. Retrieved from
www.afr.com/business/legal/ex-allens-partner-gets-mills-oakley-backing-for-legal-startup-20170330-gv9oyi
Baldassarre, G. (2015, September 10). Melbourne startup DivorceRight. StartupDaily. Retrieved from www.
startupdaily.net/2015/09/melbourne-startup-divorceright-wants-to-make-divorce-cheaper-more-efficient-andmore-amicable/
Barristers’ Board v Palm Management Pty Ltd [1984] WAR 101
Bartlett, F. & Burrell, R. (2013). Understanding the "safe harbour": The prohibition on engaging in legal practice and its
application to patent and trade marks attorneys in Australia. Australian Intellectual Property Journal, 24(2), 7493.
Beames, E. (2017). Technology-based legal document generation services and the regulation of legal practice in
Australia. Alternative Law Journal 42(4), 297–303. doi:10.1177/1037969X17732709
Beaton, G. (2014). NewLaw, New Rules: A conversation about the future of the legal services industry. Retrieved from
www.beatoncapital.com.
Bengio, Y., LeCun, Y., & Hinton, G. (2015). Deep Learning. Nature, 521(7553): 436–444. doi:10.1038/nature14539
Bench-Capon, T. (1991). Knowledge-Based Systems and Legal Applications. (T. Bench-Capon, Ed.). Academic Press.
Bindman, D, (2017, November 30). ‘Chatbot-based “firm without lawyers” launched. Legal Futures. Retrieved from
www.legalfutures.co.uk/latest-news/chatbot-based-firm-without-lawyers-launched
Bostrum, N. (1997). Superintelligence. Retrieved from nickbostrom.com/superintelligence.html
Bostrum, N. (2014). Superintelligence: Paths, dangers, strategies. Oxford: Oxford University Press.
Boyce, L. (2016, January 13). Plucky student entrepreneur who created parking ticket appeal website now launches
robot to give Britons free legal help. The Daily Mail. Retrieved from www.dailymail.co.uk/money/news/article3394093/Joshua-Browder-created-DoNotPay-launches-robot-Britons-free-legal-help.html
Brescia, R.H., McCarthy, W., McDonald, A., Potts, K. & Rivais, C. (2015). Embracing Disruption: How Technological
Change in the Delivery of Legal Services Can Improve Access to Justice. Albany Law Review 78(2), 553–621.
Brooks, M. (2017). Artificial Ignorance. New Scientist, 238(3146), 28–33. doi:10.1016/S0262-4079(17)31972-3
Copeland, B.J. (2017). Artificial intelligence. In Encyclopædia Britannica. Retrieved from
www.britannica.com/technology/artificial-intelligence
Bullock, L. (2016, April 28). NewLaw firm partners with AI provider. Lawyers Weekly. Retrieved from
www.lawyersweekly.com.au/news/18467-newlaw-firm-partners-with-ai-provider
Current State of Automated Legal Advice Tools
64
Campbell, R. (2012). Rethinking Regulation and Innovation in the U.S. Legal Services Market. New York University
Journal of Law and Business 9(1) 1–70. Retrieved from www.nyujlb.org/copy-of-8-2
Centre for Innovative Justice. (2013). Affordable Justice – A pragmatic path to greater flexibility and access in the
private legal services market. Retrieved from www.rmit.edu.au/about/our-education/academicschools/graduate-school-of-business-and-law/research/centre-for-innovative-justice/what-we-do/pastresearch/affordable-justice
Coade, M. (2016, July 20). Chatbot explores frontiers-of legal service. Lawyers Weekly. Retrieved from
www.lawyersweekly.com.au/news/19105-chatbot-explores-frontiers-of-legal-service
Coumarelos, C., Macourt, D., People, J., Mcdonald, H. M., Wei, Z., Iriana, R., & Ramsey, S. (2012). Legal Australia-Wide
Survey: Legal Need in Australia. Law and Justice Foundation of New South Wales. Retrieved from
www.lawfoundation.net.au/ljf/app/6DDF12F188975AC9CA257A910006089D.html
Cornall v Nagle (1995) 2 VR 188
Crawford, K., & Whittaker, M. (2016). The AI Now Report: The Social and Economic Implications of Artificial
Intelligence Technologies in the Near-Term. Retrieved from artificialintelligencenow.com/media/
documents/AINowSummaryReport_3_RpmwKHu.pdf
iScoop. (n.d.). Artificial intelligence (AI) and cognitive computing: what, why and where. Retrieved from www.iscoop.eu/artificial-intelligence-cognitive-computing
De Clerck, J. P., & De Wit, R. (2017b). Understanding and debating the fears about artificial intelligence.
Denckla, D.A. (1999) The Use of Nonlawyers. Fordham Law Rev 67, 1813. Retrieved From:
ir.lawnet.fordham.edu/flr/vol67/iss5/4
DeAngelis, S. (2014, September). Artificial Intelligence: How Algorithms Make Systems Smart. Wired. Retrieved from
www.wired.com/insights/2014/09/artificial-intelligence-algorithms-2
Deloitte. (2016). Developing legal talent Stepping into the future law firm, (February).
Deloitte. (2017). A guide to robotic process automation and intelligent automation. Perspectives. Retrieved from
www2.deloitte.com/us/en/pages/operations/articles/a-guide-to-robotic-process-automation- and-intelligentautomation.html
Etzioni, O. (2016, September 20). No, the Experts Don’t Think Superintelligent AI is a Threat to Humanity. MIT
Technology Review. Retrieved from www.technologyreview.com/s/602410/no-the-experts-dont-thinksuperintelligent-ai-is-a-threat-to-humanity
Furlong, J. (2014, May 13). An incomplete inventory of new law. Retrieved from www.law21.ca/2014/05/incompleteinventory-newlaw/
Gardner, J. (2017, March 14). Still drafting documents the old fashioned way? Digital Lawyer. Retrieved from
digitallawyer.news/articles/still-drafting-documents-the-old-fashioned-way
Gartner. (2017). Top 10 Strategic Technology Trends for 2017. Retrieved from
www.gartner.com/smarterwithgartner/gartners-top-10-technology-trends-2017/
Gold, N., Mackie, K., & Twining, W. (1989). Learning Lawyers Skills. Butterworths, Commonwealth Legal Education
Association.
Goodman, B., & Flaxman, S. (2016). European Union regulations on algorithmic decision-making and a “right to
explanation.” Arxiv.org. doi.org/10.2139/ssrn.2903469
Goodman, J. (2016, June 26). Firms Must Embrace AI or Risk Being Left Behind. Raconteur.net, Retrieved from
www.raconteur.net/business/firms-must-embrace-ai-or-risk-being-left-behind
Goodman, J. (2017, March 20). Legal technology: the rise of the chatbots. Law Gazette UK. Retrieved from
www.lawgazette.co.uk/features/legal-technology-the-rise-of-the-chatbots/5060310.article
Graupe, D. (2013). Principles of Artificial Neural Networks. 3rd Edition, World Scientific Publishing Co Pte Ltd.
Groover, M. P. (2014). Fundamentals of Modern Manufacturing: Materials, Processes, and Systems. Wiley.
Current State of Automated Legal Advice Tools
65
Heinz, J., Laumann, E., Nelson, R., & Michelson, E. (1995). The Changing Character of Lawyers’ Work: Chicago in 1975
and 1995. Law & Society Review, 32(4), 751–776.
Hogan-Doran, D. (2017). Accountability mechanisms: Part III Automated decision making. Retrieved from
static1.squarespace.com/static/568c9f234bf1182258eb9fbc/t/58b803cf37c58149faf5a5a7/1488454608349/Acc
ountability+Mechanisms+Beyond+Merits+Review.pdf
Huff Eckert, V., Curran, C., & Bhardwaj, S. C. (2016). Tech breakthroughs megatrend: How to prepare for its impact.
Retrieved from www.pwc.com/techmegatrend
IBA. (2016). “Times are a-changin”: Disruptive innovation and the legal profession. Retrieved from
www.ibanet.org/Document/Default.aspx?DocumentUid=2C42BEFA-DDC4-4EF5-BDD5-41FA502B987B.
IBM. (2017). Big Data Analytics. Retrieved from www.ibm.com/analytics/us/en/big-data
IBM. (2018). Watson. Retrieved from www.ibm.com/watson/
IDC. (2018). The 3rd Platform is Evolving. Retrieved from https://www.idc.com/promo/thirdplatform
iScoop. (n.d.). Artificial intelligence (AI) and cognitive computing: what, why and where. Retrieved from www.iscoop.eu/artificial-intelligence-cognitive-computing
Katz, D., Bommarito II, M., & Blackman, J. (2017). A general approach for predicting the behaviour of the Supreme
Court of the United States. Plos.org, April (12). Retrieved from journals.plos.org/plosone/
article?id=10.1371/journal.pone.0174698
Kelly, J.E. (2015), Computing, cognition and the future of knowing: How humans and machines are forging a new age
of understanding, IBM, Retrieved From:
www.research.ibm.com/software/IBMResearch/multimedia/Computing_Cognition_WhitePaper.pdf
Koebler, J. (2017). Rise of the RoboLawyers. The Atlantic. Retrieved from
www.theatlantic.com/magazine/archive/2017/04/rise-of-the-robolawyers/517794/
Law Society of New South Wales. (2017). FLIP: The Future of Law and Innovation in the Profession. Retrieved from
www.lawsociety.com.au/ForSolictors/Education/ThoughtLeadership/flip/Onlinereport/index.htm
Law Society of Western Australia (2017). The Future of the Legal Profession. Retrieved From
https://www.lawsocietywa.asn.au/wp-content/uploads/2015/10/2017DEC12-Law-Society-Future-of-the-LegalProfession.pdf
Legal Practice Board v Giraudo [2010] WASC 4
Legal Profession Practice Act 1958 (Vic)
Legal Profession Uniform Law Application Act 2014 (NSW)
Legal Profession Uniform Law Application Act 2014 (Vic)
Legal Profession Uniform Law, Schedule 1 of Legal Profession Uniform Law Application Act 2014 (Vic) and Legal
Profession Uniform Law Application Act 2014 (NSW)
Legal Profession Uniform Law Australian Solicitors’ Conduct Rules 2015
Legal Services Act 2007 (England & Wales)
Legal Services Board (England & Wales), (2016, June 28) Mapping of for profit unregulated legal services providers,
Retrieved from research.legalservicesboard.org.uk/?attachment_id=3695
Legal Services Commissioner. (2017). Uniform framework: Uniform Law. Retrieved from
www.legalservicescouncil.org.au/Pages/uniform-framework/uniform-law.aspx
Legal Services Commissioner v Walter [2011] QSC 132
Leith, P. (2010). The rise and fall of the legal expert system. European Journal of Law and Technology, 1(1). Retrieved
from ejlt.org/issue/view/1
Lodewyke, T. (2017, March 19). AI to make life easier for lawyers. Lawyers Weekly. Retrieved from
Current State of Automated Legal Advice Tools
66
www.lawyersweekly.com.au/corporate-counsel/20838-ai-to-make-life-easier-for-lawyers
Mangan, D. (2017, February 17). Lawyers could be the next profession to be replaced by computers. CNBC. Retrieved
from www.cnbc.com/2017/02/17/lawyers-could-be-replaced-by-artificial-intelligence.html
Manning, C.D. & Schütze, H. (1999). Foundations of Statistical Natural Language Processing. The MIT Press.
Marks, L. (2017, November 14). Artificial intelligence law firm aims to roll out in remote, low socio-economic
communities. ABC News. Retrieved from www.abc.net.au/news/2017-11-13/artificial-intelligence-law- firmwithout-lawyers-in-darwin/9146332
Mauldin, M. (1994). ChatterBots, TinyMuds, and the Turing Test: Entering the Loebner Prize Competition. Proceedings
of the Eleventh National Conference on Artificial Intelligence. AAAI Press.
Mayer-Schönberger, V. & Cukier, H. (2013). Big Data: A Revolution that will Transform how we live, work, and think.
Houghton Mifflin Harcourt.
McCalman, L. (2017, October 18). Ethics by numbers: how to build machine learning that cares. The Conversation.
Retrieved from theconversation.com/ethics-by-numbers-how-to-build-machine-learning-that-cares-85399
McCarthy, J., Minsky, M., Rochester, N., & Shannon, C. E. (1955). A Proposal for the Dartmouth Summer Research
Project on Artificial Intelligence.
McGinnis, J., & Pearce, R. (2014). The great disruption: How machine intelligence will transform the role of lawyers in
the delivery of legal services. Fordham Law Review, 82(6), 3041–3066. Retrieved from
ir.lawnet.fordham.edu/cgi/viewcontent.cgi
McKinsey & Company. (2016). How blockchains could change the world: Interview with Don Tapscott. Retrieved from
www.mckinsey.com/industries/high-tech/our-insights/how-blockchains-could-change-the-world
Medlock v LegalZoom, 2013, Special Referee of South Carolina’s Supreme Court, unreported settlement agreement
Michalski, Ryszard S., Jaime G. Carbonell, and Tom M. Mitchell, eds. Machine learning: An artificial intelligence
approach. Springer Science & Business Media, 2013.
Miller, K. (2015). Disruption, Innovation and Change: The future of the legal profession. Law Institute of Victoria.
Retrieved from www.liv.asn.au/Staying-Informed/Presidents-Blog/LIVPresBlog2015/December-2015/The-futureof-the-legal-profession--Tools-for-lawy
Miller, K. (2017). Opportunity knocks. Law Institute Journal, 9.
Miller, S. (2017a). Part II: The Future of Artificial Intelligence. Thomson Reuters. Retrieved from
legalsolutions.thomsonreuters.com/law-products/news-views/corporate-counsel/legal- department2025/artificial-intelligence/future-of-artificial-intelligence-robot-lawyer-army-or-not
Miller, S. (2017b). Part IV: AI Adoption and Ethical Considerations for Inhouse Counsel. Thomson Reuters. Miller, T.
(2017). Email.
Miller, T. (2018). Explanation in Artificial Intelligence: Insights from the Social Sciences. ArXiv e-prints, 1706.07269.
arxiv.org/abs/1706.07269
Miller, T., Howe, P., & Sonenberg, L. (2017). Explainable AI: Beware the Inmates Running the Asylum; or How I Learnt
to Stop Worrying and and Love the Social and Behavioural Sciences. In IJCAI 2017 Workshop on Explainable AI
(XAI). arxiv.org/abs/1712.00547
Nissan, E. (2017). Digital technologies and artificial intelligence’s present and foreseeable impact on lawyering,
judging, policing and law enforcement. AI and Society, 32(3), 441–464. doi.org/10.1007/s00146-015-0596-5
Pasquale, F., & Cashwell, G. (2015). Four Futures of Legal Automation. UCLA Law Review Discourse 63 26–48.
Retrieved from www.uclalawreview.org/wp-content/uploads/2015/06/Final-ALL.pdf
Productivity Commission. (2014). Access to Justice Arrangements: Inquiry Report Volume 2 (Vol. 2). Canberra.
Pearl, J. (2018). Theoretical Impediments to Machine Learning with Seven Sparks from the Causal Revolution. arXiv
preprint. Retrieved from arxiv.org/abs/1801.04016
Current State of Automated Legal Advice Tools
67
Remus, D., & Levy, F. S. (2016). Can Robots Be Lawyers? Computers, Lawyers, and the Practice of Law. SSRN. Retrieved
from papers.ssrn.com/sol3/Papers.cfm?abstract_id=2701092
Rhode, D.L. & Ricca, L.B. (2014). Protecting the Profession or the Public? Rethinking Unauthorized-Practice
Enforcement, Fordham Law Review, 82, 2587. Retrieved from ir.lawnet.fordham.edu/flr/vol82/iss6/2
Rinaldi, L. (2017, September 18). Ross the robot is the city’s best legal mind. Toronto Life. Retrieved from
torontolife.com/tech/ross-robot-citys-best-legal-mind
Giddings, J. & Robertson M. (2001). Informed litigants with nowhere to go: Self-help legal aid services in Australia.
Alternative Law Journal, 26(4), 184–190.
SAS. (2018). Machine Learning: What is it and why it matters. Retrieved from:
www.sas.com/en_au/insights/analytics/machine-learning.html
Schatsky, David and Vikram Mahidhar, (2014, January 22) Intelligent automation: A new era of innovation, Deloitte
Insights Retrieved from dupress.deloitte.com/dup-us-en/focus/signals-for-strategists/intelligent-automation-anew-era-of-innovation.html
Salian, I. (2017, July 14). “Moneyball” legal analytics helps lawyers assess judges. San Francisco Chronicle. Retrieved
from www.sfchronicle.com/business/article/Moneyball-legal-analytics-helps-lawyers-11289892.php
Silver, David, et al. (2017) Mastering the game of go without human knowledge. Nature 550 354–359 (2017),
doi:10.1038/nature24270
Simonite, T. (2017, March 6). How to Upgrade Judges with Machine Learning. Technology Review. Retrieved from
www.technologyreview.com/s/603763/how-to-upgrade-judges-with-machine-learning
Surden, H. (2014). Machine Learning and Law. Washington Law Review, 89(1), 87–115.
Susskind, R. (1987). Expert Systems in Law: A Jurisprudential Inquiry. Oxford University Press.
Susskind, R. (2010). The End of Lawyers? (Revised). Oxford University Press.
Susskind, R., & Susskind, D. (2015). The Future of The Professions: How Technology will Transform the Work of Human
Experts. Oxford University Press.
Tapscott, D., & Tapscott, A. (2016). Don Tapscott and Alex Tapscott, The Blockchain Revolution: How the Technology
Behind Bitcoin is Changing Money, Business, and the World. Brilliance Audio.
The world’s most valuable resource is no longer oil, but data. (2017, May 6) The Economist. Retrieved from
www.economist.com/news/leaders/21721656-data-economy-demands-new-approach- antitrust-rules-worldsmost-valuable-resource
Tucker, I. (2016, November 27). Genevieve Bell: “Humanity”s greatest fear is about being irrelevant’. The Guardian.
Retrieved from www.theguardian.com/technology/2016/nov/27/genevieve-bell-ai- robotics-anthropologistrobots
Turing, A. (1950). Computing Machinery and Intelligence. Mind, LIX(236), 433–460. doi:10.1093/mind/LIX.236.433
Urbis. (2017). National Profile of Solicitors 2016. Retrieved from www.lawsociety.com.au/cs/groups/public/
documents/internetcontent/1378059.pdf
Wallace, C. (2017). Competition, Growth and Consumer Outcomes: Challenges for Policy. In Westminster Legal Policy
Forum. Retrieved from www.westminsterforumprojects.co.uk/publications/westminster_legal_policy_forum
Walsh, K. (2017, September 21). Lawyers need to “wake up”: next disrupters could be supermarkets. AFR. Retrieved
from www.afr.com/business/legal/lawyers-need-to-wake-up-next-disrupters-could-be-supermarkets-20170918gyjoid
Wachter, S., Mittelstadt, B. & Floridi, L. (2017) Why a Right to Explanation of Automated Decision-Making Does Not
Exist in the General Data Protection Regulation International Data Privacy Law, 7(2), 76–99.
Weather, H., & Hale, B. (2016, January 2). Meet the boy genius who can get you off a parking fine. The Daily Mail.
Retrieved from www.dailymail.co.uk/news/article-3381510/Meet-boy-genius-parking-fine-Joshua-18-fedgetting-tickets-set-website-help-escape-penalties.html#ixzz4wb2zqs00
Current State of Automated Legal Advice Tools
68
Webb, J., Miller, T., Bosua, R., & Bennett, J. (2017). Regulating automated legal advice tools. Networked Society
Symposium, University of Melbourne.
Wüst, K. & Gervais, A. (2017). Do you need a Blockchain?" IACR Cryptology ePrint Archive 2017, 375. Retrieved from
eprint.iacr.org/2017/375
Current State of Automated Legal Advice Tools
View publication stats
69
Keep reading this paper — and 50 million others — with a free Academia account
Used by leading Academics
Hector Olasolo
Universidad del Rosario
Mauro Grondona
University of Genova
Asiah Bidin
Universiti Sultan Zainal Abidin, Terengganu, Malaysia
Jose Cretella
Escola Paulista de Direito - EPD
