Design Guidelines for Negotiation Support Systems: An
Expert Perspective Using Scenarios
Alina Pommeranz , Willem-Paul Brinkman, Pascal Wiggers, Joost Broekens, Catholijn M.
Jonker
Section Man-Machine Interaction, Delft University of Technology
Mekelweg 4, 2628CD, Delft,
The Netherlands
{a.pommeranz, w.p.brinkman, p.wiggers, j.broekens, c.m.jonker}@tudelft.nl
ABSTRACT
Negotiation support systems (NSS) can enhance humans’
performance in negotiations. Much research in this area focuses
on finding optimal bids. However, there is little research on
human factors in technological negotiation support. We believe
an in-depth analysis of the task involving experts and users is
needed to build a new generation of NSS focusing on manmachine collaboration. We describe a scenario-based approach
to gathering requirements for such a system. We wrote five
scenarios containing part of the envisioned functionality in the
most important use situations, e.g. face-to-face negotiation, on
the phone, collaborative or mobile preparation. We used claims
analysis to clarify our design decisions. To evaluate our claims
we organized focus groups including six general and six job
negotiation experts. The filmed scenarios were used together
with two claims each to guide the discussion. Based on the data
analysis we constructed 12 design guidelines for NSS.
Keywords
Decision making and problem solving, Decision aiding,
Human-system interaction, Design approaches.
ACM Classification Keywords
H.4.2 [Types of Systems] Decision support.
1. INTRODUCTION
1.1 Background
Existing Negotiation Support Systems (NSS) can enhance the
human performance in negotiations and increase the number of
win-win outcomes if the negotiation space is well-understood
[9, 10]. This is because computers are good at coping with the
computational complexity involved in calculating offers.
However, there are a number of issues inherent in real life
negotiations that are difficult to deal with using classical
Artificial Intelligence (AI) approaches alone. These issues
mostly relate to the interpretation of the social setting.
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies
are not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. To copy
otherwise, or republish, to post on servers or to redistribute to lists,
requires prior specific permission and/or a fee.
ECCE 2009, September 30 – October 2, 2009, Helsinki, Finland.
Copyright 2009 ACM 1-58113-000-0/00/0004…$5.00.
Therefore, NSS are required that take into consideration the
strengths of both the machine and the human.
Current research on technological negotiation support is carried
out in different areas, primarily in management science,
electronic commerce and Artificial Intelligence [1, 14, 21]. It is
hard to find studies in this area that include human factors [11],
which is surprising since NSS are a type of interactive systems
that offer rich possibilities for researching and designing
human-computer interaction. However, different problems such
as reaching optimal solutions and bids, formal descriptions of
negotiations, the extraction of user preferences or problem
representations remain the major foci of research efforts.
As formulated in [2, 19], NSS research concentrates on
technological solutions, but the social problems that they intend
to solve are secondary or even completely neglected. More in
particular Swaab and colleagues [19] argue for a careful
analysis of social and psychological processes in order to design
good NSS. They claim that the success of an NSS is not only
dependent on technological feasibility but also on the
understanding of the activity that the system will support. These
authors attempt to inform the design process of NSS.
However, they primarily look at two aspects that influence the
outcomes of negotiations positively, namely common (cultural)
identity and shared cognition. In this sense NSS can help by
providing information to the opposing parties to establish a
common understanding of the problem and possible solutions.
Their studies show that the nature and representation of the
information can influence negotiation outcomes. Another effort
to emphasize the importance of social and also emotional issues
in negotiation and their consideration for NSS has been made by
Bui [2]. In his article the author points out problems that evolve
from the fact that empirical research focuses only on the rational
aspects of negotiation. For instance, the negotiation models that
are implemented in NSS assuming strict economic rationalization
ignore that people also take decisions based on social
acceptability of different means to achieve a deal. Adding
reasoning based on ethical and social norms to negotiation
models will allow them to better represent the real life
negotiation processes. Bui explores socio-emotional aspects
such as conflict awareness, thoughts, emotions, intensions, trust
and norms and their impact on negotiation. He creates a general
list of aspects that NSS should help users with, such as
identifying controversy, clarifying issues/criteria, equalizing
parties or finding solutions and simulating impacts of potential
decision. These can be seen as more generic guidelines for the
functionality and design of NSS. Both cases [2, 19] refer to
shared NSS used either collaboratively by all parties or as
mediators. This is only one type of
requirements.
NSS
with special
An interesting research area where social aspects are actually
considered is group decision support [e.g. 13]. However, also in
those cases the focus is on collaboration and verbal
communication between the participants rather than aspects like
thoughts, emotions, trust etc.
In this paper we focus on the challenges of developing a system
that is used only by one party in the negotiation and takes into
account the human factors occurring in negotiations. Our aim is
to extract detailed guidelines for this negotiation setting that
extend the generic guidelines presented in [1].
We explain our scenario-based approach for finding such design
guidelines. The second section presents a thorough description
of our qualitative method using scenarios, claims analysis,
videos and expert focus groups. The data analysis is presented
in the third section, followed by the results including the design
guidelines in section four. Finally, we give a conclusion about
our work in section five.
1.2 Research goals
Our project goal is to build an NSS that supports one party in a
dyad negotiation thorough all phases of the process
(preparation, begin, analysis, bidding, closure). Since all
negotiations differ and the domain of the negotiation has a
major influence on the process, our project focusses on two
example application domains: housing and job negotiations.
To achieve a good interaction between our system and the user
it is essential that both explicitly share a generic task model. In
order to implement such a model, we need to find out what task
we will support and understand that task in depth. Furthermore,
as pointed out in the background it is important to consider
social, psychological and emotional aspects in the design of
NSS. However, only a few researchers [2, 19] investigated such
aspects and they focussed on broad social science concepts and
their implications on negotiation. Therefore, the design
guidelines that can be extracted from their research are generic.
We believe additional in-depth analysis of negotiations is
necessary to create specific guidelines that will inform the
design process of the next generation of NSS. In addition to
studying theoretical research from negotiation and social
science literature, it is important to explore negotiation practice
in the real world. We used a scenario-based design approach [6]
involving general negotiation experts and job negotiation
experts. Our primary research goal is to construct a number of
design guidelines for NSS considering real life practice from an
expert perspective. To reach that goal we have a sub-goal of
understanding the users’ task in context, as well as their
behaviours and needs.
2. SCENARIO-BASED APPROACH
Overall we follow a user-centered design approach within this
project. Our target users are people with different negotiation
skill levels. As shown in the literature [20] most people are not
very good negotiators although it is an almost daily task.
Whereas users might not be able to explain themselves,
negotiation experts and trainers have a good grasp of the
common practice, mistakes and support that is needed. As
pointed out by other researchers, a participatory design
methodology making the user a co-designer from the beginning
might not be sensible when the user knows little about the
domain and is supposed to be taught about it by the software.
We therefore adopted the informant design framework
suggested by Scaife and his colleagues [15] that proposes to
involve various participants at different stages in the design
process. By this we can maximize their input and advance the
development. The participants can be users as well as experts
depending on the kind of data needed in certain design stages.
At this early stage we will aim at gaining knowledge about
negotiation from domain experts. In order to get a structured
overview of the situations and the ways our NSS should support
users, we decided to organize a number of focus groups with
experts. As will be explained in more detail below focus groups
offer an interesting way to discuss first ideas due to their
dynamic group element.
To be able to gather useful data we decided to guide the group
discussions by using filmed scenarios of use situations of an
NSS. Not only did we want to show our ideas about parts of the
envisioned functionality but also get feedback on possible use
contexts. Use contexts are determined by the different phases of
a negotiation as well as the situation and conditions in which
the negotiation takes place. A first brainstorming session with
seven researchers of our project group helped to envision
related functionality and situations in which different phases of
the negotiation take place. This group of researchers is
interdisciplinary consisting of HCI researchers with foci on
psychology, emotions and user-system interaction as well as
researchers from the fields of computer science and Artificial
Intelligence. All researchers are familiar with classical
negotiation literature. A first selection of feasible ideas was
written separately on post-it notes which we clustered. Based on
those clusters we generated, in close collaboration with a
negotiation coach, five distinct use contexts that together cover
all negotiation phases and types of system use (e.g. open,
hidden etc.). The contexts are: face-to-face negotiation, remote
negotiation (phone, internet etc.), preparing collaboratively for
an upcoming negotiation and preparing for a negotiation with
time constraints while being mobile (e.g. on the train). Each of
these situations offers characteristics that influence the
acceptability as well as the functionality of the NSS.
2.1 Scenarios
Scenarios are useful in the design process since they capture the
consequences and trade-offs of designs [5]. The narrative nature
of scenarios enables users, experts as well as designers to
imagine the use situations and contexts of new or existing
technology. For each of the five use contexts we wrote a
scenario presented here in summary. Italic text is taken from the
original texts of the scenarios. We chose to write two scenarios
illustrating a job negotiation, two with real estate content and
one about buying a car. We included one scenario set outside of
our application domains in order to investigate how a
completely different domain can influence the devices’ role and
functionality. All scenarios were checked by a professional
negotiation coach to make sure that they were sufficiently
realistic. Each scenario is briefly discussed below.
Mobile Preparation with Time Constraints. Preparation is
one of the negotiation phases stressed in the literature, e.g. [8].
In this scenario we describe a preparation situation with special
constraints. The job applicant Martin is already on his way to
the interview. Therefore he has limited time to prepare himself.
In addition, the mobile setting constitutes another constraint,
namely limited resources. Both constraints require special
regard when it comes to the functionality of the device. Just
before getting on the train to the interview Martin has received
a NSS on a pocket device. He uses the device’s speed
preparation function to prepare himself in the short time he has
left. Among other functions the device allows him to receive
knowledge about the job negotiation domain. He wonders how
much money he could ask for. He chooses „expert opinion‟ on
the interface and types in „salary‟. The PN suggests a website
that has a forum where you can discuss current average
salaries for IT consultants with an expert in the field. After
reading through the forum Martin has a quite good idea what
he can ask for with his kind of educational background and
experience. With that knowledge he feels more secure and
relieved.
Later in the scenario Martin makes use of the training module of
the NSS which enables him to go through a simulated interview.
He receives on-the-fly advice about his and the opponents’
actions. The scenario ends with a more relaxed applicant, who
knows what to expect in the upcoming negotiation.
Face-to-Face Negotiation, Secret Use. The situation described
in this scenario is a negotiation between an employee, Bianca,
and her boss. Bianca is using a pocket device with a NSS. She is
hiding the fact that she has such support by telling her boss he
is using her device to take notes.
Bianca has been working for a big telecommunication company
in The Hague for 2 years now. Today her annual evaluation
with her boss is due. Her boss is known as a quite friendly
person, who hardly ever becomes aggressive or ill-tempered.
However, he is very worried about his department‟s
performance and likes to know exactly what his employees are
doing. Bianca wants to take this meeting as an opportunity to
re-negotiate some parts of her contract. Since her husband got
a new job in another city, they decided to move further away.
Therefore she wants to discuss with her boss about
opportunities to handle the new situation. She knows that she
worked hard and well in the last year and should get what she
wants, but she does not consider herself a good negotiator.
Therefore, she recently got the NSS and prepared herself for this
negotiation with the device.
Throughout the negotiation described in the scenario Bianca
receives help from the device. Several functions are described in
this scenario including e.g. affect management, generating new
options, and giving behavioral advice.
Bianca presses the button „opponent concerned‟. The NSS
advices her to uncover the reasons for Mr. Smith‟s worries and
show sympathy. Bianca asks: “May I ask you what your
concern is?” Mr. Smith replies: “We always have a lot of
spontaneous meetings to decide on how to proceed, which you
will be missing if you were not here and since you are one of
the main developers I think you should attend such meetings.”
“I really understand your worries, Mr. Smith. However, the
welfare of my family is very important to me. But I am sure we
can find a solution that considers both our concerns.” The
scenario ends with a deal in which both parties gain something
and are satisfied with.
Collaborative Preparation. Negotiation involves a lot of
emotions on both sides of the bargaining table, but also within a
party, e.g. between a couple buying a house together. In this
case the first step is to merge the demands and preferences of
both partners before starting a negotiation with the opponent
side. Our scenario describes a couple that is planning to buy a
house together and uses the NSS during the preparation to sort
out their preferences and to download domain knowledge about
real estate. They both sit close to each other on the sofa and
look at the screen together. Mary starts the NSS and a virtual
agent (VA) welcomes her. “What would you like to do?” he
asks. Mary types in „merge my partner‟s preferences with
mine‟. The „collaborative preparation‟ module starts up. After
a short introduction the NSS asks each of them to put in their
preferences for a house separately. Since they also have the NSS
software installed on their laptop they put in their preferences
in parallel. From both preference profiles the NSS creates a
matching profile and shows the clashes of their preferences. It
advices the couple discussing the clashes and trying to find
trade-offs between them that suit both. During this process of
compromising the couple gets into a quarrel in which both insist
on their own wishes without even communicating the
underlying reasons in detail. In this case our device takes on a
proactive role and interrupts the couple to give advice on how
to handle the conflict. The NSS senses the noise and the angry
voices in the room and assumes an argument. On the screen the
VA appears and says “it became very loud in the room. Are you
arguing?” Since the device does not get any attention a red
LED starts blinking and a beep sounds. Both Mary and Piet
stop talking and look at the NSS. Mary answers the NSS‟s
question with yes. The NSS suggests calming down […and…]
prompts them to put in an emotional value on a scale from „I
don‟t care at all‟ to „I would die for this‟ for each variable they
have different preferences on.”
After having sorted out all their preferences they start looking
for houses. In the last scene of the scenario the couple visits a
house and takes advantage of the NSS’s feature of taking
pictures and storing them together with other information about
the house in a database.
Negotiation on the phone. A negotiation in which both parties
are not situated in a face-to-face setting, but are distant from
each other offers different design challenges for a NSS. First of
all one party does not see the other party and therefore the use
of a NSS can take place without each others’ notice. Especially
in real estate situations, e.g. when buying a house another
aspect to consider is that the negotiation is split into a number
of phone calls. This gives the user time in between the calls to
use the system in each step of the negotiation. In our scenario a
couple has decided to buy a house. Before the wife starts the
negotiations with the real estate agent of the seller, the couple
decides on a price. They use the NSS to download information
about prices of similar objects in the same region to know what
to expect. Furthermore, the PN has downloaded housing
domain knowledge, such as contracts and legal issues and the
prices of similar houses in the neighborhood to take into
account. Before Mary came to work this morning she had
decided with Piet to set a first bid around 450.000 Euro.
At work Mary calls the agent and starts negotiating. Before and
during the phone calls she uses the NSS on her laptop to receive
advice about different steps in the negotiation, e.g. the NSS
advices her to not start the negotiation with offering a price,
but instead talk about other issues and options…
The bidding goes on for a while and the NSS shows a
visualization of the bids in the outcome space based on the
preferences of Piet and Mary and the estimated preferences of
the agent. After a while the NSS detects that the bidding is not
reaching a win-win situation. After finding new variables to
include in the negotiation to reach an agreement that suits both
parties they finally close a deal.
Face-to-Face Negotiation, Open Use. We decided to include
another scenario that has a face-to-face setting, but showing an
open usage of the NSS. This scenario is about a couple buying a
car. Our belief is that the car dealer’s setting enables people to
use the NSS more openly. When buying a car it is usually not
necessary to stick to one specific car dealer. No long-term
relationships need to be considered. Therefore, the couple in the
scenario openly states that it will be using the device and
explains what they can do with it. The focus of the scenario lies
in the advice of time-outs at strategic points during the
negotiation. During the process of looking at cars and refining
their preferences for the new car, they enter information about
the state of the negotiation into the NSS. They receive strategic
advice on how to proceed and when to take the time to
recapitulate.
He [the car dealer] shows them a range of more sporty looking
family cars and the couple chooses their favorite. They enter
that into the NSS. The NSS advices them to take a time-out and
check whether they have considered all their preferences and
whether all the information they need has been disclosed.
After they have found an interesting car the bidding starts in the
car salesman’s office. The NSS assists the couple by comparing
prices with similar cars online. They disclose to the salesman
that the market price is lower than his offer. The salesman drops
his price. They negotiate about a few extras and finally leave
with a new car and a deal they are satisfied with.
2.2 Storyboards and Videos
Due to their illustrative strength scenarios are a good means to
communicate design ideas within the project team as well as to
users or experts in the field. In order to exploit that strength
even more we decided to visualize the scenarios. First we
created a storyboard for each of the scenarios For the
collaborative preparation scenario see Figure 1. These
storyboards then served as a basis for the shooting and editing
of short (about two to three minutes) videos (for an example
video see http://mmi.tudelft.nl/video/scenario2/). Using videos
we were able to present the use contexts of our NSS very well.
Much of the functionality of the NSS was kept open for
interpretation to avoid limiting the discussion about the
functionality. The videos were used in the design process as
described in the focus group section. In the future they will also
be used alongside a questionnaire on users’ acceptability of an
NSS in different use contexts.
2.3 Claims Analysis
Due to the scenarios’ narrative nature many things are left
implicit. Often causal facts and relations underlying the actions
described are not revealed. Therefore it is useful to enumerate
such causal relations separately. This can be done through
claims analysis [5]. Each claim underlying a certain action or
design feature in the scenario is listed together with its tradeoffs. We used the claims slightly different, as proposed by
Neerincx [12], namely in order to test our hypothesis about
functionality and use contexts in the focus groups discussions
with the experts. We wrote down four to six claims per scenario
based on our hypothesis. Due to space limitations we cannot list
all the claims here, but only give examples. The first claim was
written for the face-to-face scenario and the second for the
negotiation on the phone scenario:
Advice claim: the NSS gives generic advice for different
negotiation phases in a text-based form (e.g. ask for reason of
concern, be sympathetic, and maintain the relationship).
+ Even though the user might know of such things due to a
good preparation, the NSS’s advice serves as a reminder during
the negotiation process.
- The user might not be able to put the advice to practice or the
way he tries to do so is not effective.
Graphical representation claim: the NSS shows the current
status of the negotiation graphically including all variables etc.
+ The variables and their influences on the negotiation process
are shown, so that the user can understand the process better.
+ The user can recapitulate and learn for future negotiations by
looking at the current status and the influences of the variables.
- The number of variables and influences is high and the user
finds it hard to learn from the graphical representation.
- The graphical representation is not understood by every type
of user.
2.4 Focus Groups
Focus groups [16] have been widely used in marketing to
exploit the dynamics of group discussions in order to receive
attitudes towards ideas or products. Bruseberg and McDonaghPhilp [3] have shown that focus groups are also useful during
the design process of new technologies. They help the
participants to articulate their ideas and provide the researcher
with inspiration for the design process. Lately, HCI researchers
have adopted the method and refined the techniques used to
stimulate the discussion. As for instance, Goodman and
colleagues [7] found out, it is profitable to use visual help such
Fig. 1. Collaborative preparation scenario
as pictures and also scenarios in focus groups. Furthermore,
tasks can start up a discussion. Based on these findings we used
the previously described scenarios in form of videos in the
focus groups.
In total we had a number of 12 experts divided into three focus
groups. We divided the experts into different focus groups
according to their expertise. As explained by a number of
researchers, e.g. [16], the homogeneity of the group plays an
important role. The more similar the group members are the
more likely they are to voice their opinions. Therefore, we
formed one group with general negotiation experts, such as
negotiation trainers, lawyers, a judge etc., and two with job
negotiation experts, such as human resource employees and
labour union representatives. In the beginning participants were
introduced to each other and the project was described. Every
participant received a questionnaire that contained two claims
from the claims analysis (see previous section) per video. The
claims, however, were reformulated into statements that allowed
the experts to specify their level of agreement with. The two
claims named in the previous section were presented as the
following statements:
Statement: General tips and strategic advice [e.g. try small
talk, show sympathy for your opponents concerns] is more
useful for the user than specific behavior- and decision-advice.
Statement: The NSS should focus on helping the user to
understand the bidding process [e.g. graphical representation
of the bidding including history of bidding] rather than
proposing the next bid.
After watching each video the participants individually
specified their level of agreement on a 7-point Likert scale, and
provided comments. We chose this method to give everyone a
chance to think about their own attitudes and opinions in
silence. As pointed out by e.g. Carey [4] less confident
members may be encouraged to disclose more when having
written down their views in advance. Once every member
finished writing the moderator started a group discussion, by
asking the participants in turn to react to the claims and discuss
their ideas with the others. The moderator stimulated the
discussion without enforcing any existing views from the
project team. The discussion was taped for later analysis. In
addition, two researchers in every group took notes. Taking
notes is important since simple audio-recording cannot always
capture what is happening between the members of the group.
3. DATA ANALYSIS
Our approach results in two types of data, i.e., data from the
notes and data from the questionnaires. To analyze the
questionnaire data (values on a Likert scale) we used a standard
mean value calculation. Figure 2 presents the average level of
agreement of the experts with the claims that were presented in
the questionnaire. Considering the 95% confidence interval and
the value four as the middle of the scale the results suggest that
the majority of the experts leaned towards agreeing with the
claims: 1) open use of the device when buying a car benefits the
outcome; 2) device should help the user to understand the
bidding rather than giving the next bid; 3) general tips are more
useful than specific advice; 5) in preference elicitation ask for
core concerns (instead of specific values); 6) short training and
simulation enhances negotiation skills; and 7) short preparation
contribution positively to negotiation outcomes. The qualitative
data explains the rationale behind these positions and provides
additional ideas.
Focus groups provide large amounts of qualitative data, due to
the dynamic nature of the group and the contextual setting. As
discussed in detail in [4, 16] the data analysis of focus group
data is delicate. Researchers have to be aware that focus groups
are not meant to find consensus within the group and that
empirical generalization from the data is not possible. However,
according to Sim [16], the data from focus group can provide
theoretical insights with sufficient level of universality to be
projected to comparable contexts.
For the analysis of our data we used a method similar to
interpretative phenomenological analysis [17], which is a
bottom-up method often used in psychological qualitative
research. The idea is to go through the data from one focus
group to gather emerging themes from the text. Themes can be
recurring ideas, thoughts or feelings from the participants.
These themes are then clustered together and superordinate
concepts might emerge. This process is repeated for the other
focus groups and finally, the superordinate themes are
compared and converged to final themes or theories, i.e. in our
case transformed into design guidelines.
We analyzed the sessions separately on the basis of the notes by
at least two researchers. The recordings from the sessions were
only used in case the notes were not clear enough or
incomplete. Every idea or attitude was written on a post-it note.
Repeated ideas were not written down again, as we were not
trying to get empirical generality and furthermore, in groups
people tend to agree with or repeat thoughts and ideas.
To define the general themes that can be transformed into
design guidelines four researchers independently clustered the
post-it notes. We intentionally included one researcher
unrelated to the project. Therefore, we could compare unbiased
data with the data from the project researchers. Themes thus
identified were then compared across all focus groups. This
revealed that researchers used two arguments to categorize the
themes. Either they considered the system’s functionality or
they looked at the phases in negotiations process. The system’s
functionality perspective led to four categories, namely
negotiation tactics, usage of an NSS, information the NSS should
provide, adaptivity of the NSS to the user, and the interaction
with the interface. The negotiation perspective extracted
categories for all negotiation phases, such as training, extracting
preferences, context analysis, interaction with the opponent, and
analysis of the bidding process. In particular, the participants
emphasized that the device should motivate the user to prepare,
as even a short preparation will be beneficial for the negotiation
outcome. Furthermore, they expressed that the device should
help people understand the bidding process instead of just
proposing next bids. Note, that although the discussions were
triggered by the statements and the filmed scenarios, they also
gave insights that cannot be directly linked to the statements.
New themes arose, e.g., the importance of context and the
adaptivity of the system to the user. All themes fall into the
categories resulting from the clustering.
95% agreement score
Fig. 2. Mean values of agreement with claims (1=strongly disagree, 7=strongly agree) including 95% confidence interval.
4. FROM THEMES TO GUIDELINES
In the following we elaborate on the themes and construct eight
design guidelines from the themes. Themes are presented in
bold and guidelines in italics.
An NSS device adds higher value in the preparation and
training phase than during a negotiation. Training needs to
be interactive and the NSS needs to react intelligently.
All experts across the groups agreed on the fact that any
preparation for a negotiation is useful. However, some experts
mentioned that a technical device should add more value to the
preparation than just reading a book on negotiation. They
emphasized the importance of training and simulation and
pointed out that the system needs to be able to respond to the
user in an intelligent way. In detail, one idea that was
mentioned was that the system needs to make people aware of
what they can negotiate about. In addition, the system needs to
ask questions to the user similar to the ones asked in job
negotiations. In one group it was mentioned that multiple short
sessions of preparation might be better than one long one.
1) An NSS should support interactive preparation sessions of
different lengths.
2) The preparation module should have a simulation mode in
which the user interacts with an intelligent negotiation agent.
In a face-to-face situation it is hard for the user to focus both
on the device and the opponent.
Most experts were of the opinion that an NSS should not be used
in face-to-face negotiations. Especially the job negotiation
experts mentioned that the way the applicant or employee
presents him/herself is important as well as focusing on the
negotiation partner. While using a device the interaction with
the opponent becomes awkward and might be embarrassing.
Furthermore, the experts were concerned that understanding and
processing the device’s information and advice takes too much
time and is too much cognitive load for the user in a face-toface situation.
3) The cognitive load of the information representation
provided by the NSS during a face-to-face negotiation should be
minimized.
The
context
including
atmosphere,
non-verbal
communication and emotions plays a major role for the
negotiation process.
In two focus groups it was emphasized that especially in job
negotiations the non-verbal communication and the atmosphere
in the room play an important role. Furthermore, emotions
influence the decision-making process and the course of
negotiation. This means that the system needs to be able to
obtain this context information and take it into account when
reasoning about next steps. People are generally better at
interpreting emotions, non-verbal communication and
atmosphere than computers. One way of enabling the system to
understand the context is to build a context model within the
system and let the user enter information about the context
during the negotiation. To reduce the data that the user needs to
feed into the system other techniques like emotion recognition
or using (e.g. sound) sensors might be a solution.
11) An NSS should suggest time-outs at appropriate stages in
the negotiation process.
4) In the training module the user should be trained on being
aware of the context.
Preferences of collaborating partner’s should be put in
separately.
5) Advice from an NSS should consider information about the
context of the negotiation.
Across the focus groups there was a consensus that in the
process of generating a preference profile for collaborating
partner’s, e.g. couples, they should put in the their preferences
separately. That avoids that one partner is more dominant than
another. In our scenario we proposed that the system then
merges the preferences and shows the clashes to the users. The
experts did not agree on doing it this way. They pointed out that
showing those clashes triggers arguments between the partners
instead of a discussion about underlying values. It is more
important that the partners talk about such values and come to a
conclusion. The system could also directly suggest solutions. It
was also proposed that a user indicates the importance of every
preference.
The NSS is strong in the rational part of a negotiation, by
offering new options and for storing and managing data. It
should provide domain knowledge in terms of facts that the
user can use to persuade.
Most experts agreed that the strength of a device would lie in
handling the rational part of a negotiation. It can store and
manage vast amounts of data, deal with the computational
complexity during the bidding and offer new options to the
user. Furthermore, domain knowledge should also include
mainly facts, such as prices or salaries, which the user can use
to persuade his/her opponent.
6) An NSS should support the user by calculating bids and
offering new options to negotiate on.
7) It should have a data storing and managing function that
gives the user easy access to the information needed at a
certain point in time.
Both generic and specific advice is useful but needs to
applied carefully.
One of our claims was that generic advice is more useful than
specific advice. The attitude towards this claim differed between
the experts. Many of them saw a danger in specific advice
because if the system cannot sense the context specific advice is
often inappropriate. Generally both generic and specific advice
could be useful but is dependent on the negotiation phase and
the capabilities of system and user.
8) An NSS should generally provide the user with more generic
advice that the user can apply to the situation he/she is in.
The NSS needs to adapt to the user’s behavior and his
knowledge or experience.
At several points in the discussion it was mentioned that the
system advice or reactions need to be adapted to the experience
of the user and his/her behavior. Regarding advice given by the
system it was mentioned that novice users who are not good
negotiators should get more specific advice whereas more
advanced users are able to apply more generic advice. During
the bidding the system should adapt its behavior to that of the
user and recalculate the next bids in case the user changed
his/her strategy.
9) An NSS should be able to adapt to the user‟s skill level and
experience and more in specific to the user‟s bidding behavior.
10) System advice should be based on the capabilities of the
user to apply them in practice.
Interruptions are seen controversial. Time-outs, however,
are good.
The majority of the experts thought that active interruptions by
the system through vibrating and beeping during a tense
situation are not useful. The users would either ignore the
system or become more upset. However, most experts agreed
that time-outs are very useful for reflection of the negotiation
process. As the user is not always aware of when to take a timeout the system should suggest it.
12) Partners should put in their preferences separately and
assign an (emotional) value to each preference.
5. CONCLUSION
Overall these guidelines boil down to the following overall
insight: the preparation phase of a negotiation and the actual
negotiation with an opponent require different interaction
styles. In the preparation phase NSS should provide a
negotiation training that is rich, content-full and contextual.
Preferably it should make use of an adaptive scenario including
socially intelligent opponents to provide a real setting. During
the negotiation with an opponent, on the contrary, the system
should provide concrete, personalized advice regarding offers
and generic advice regarding the negotiation process with easy
interpretable hints. The interaction style in this case should be
as little interrupting as possible.
The major implication of these guidelines is that NSS need to
have intelligence and reasoning capabilities in order to process
the information entered by the users and give personalized
output. Furthermore, the system needs to possess an accurate
user model that is updated during the interaction to be able to
adapt to the user. Furthermore, the interaction styles need to be
carefully selected for each phase of the negotiation.
With regard to our approach we learned that the addition of
video material as stimuli in focus groups facilitates idea
generation and discussion within the group. Participants were
able to directly reflect upon the potential usage of the NSS. The
discussion was vivid and constructive. During the focus groups
we got a detailed account of real life negotiations from the
viewpoint of negotiation experts such as negotiation trainers,
judges, labor union representatives and human resource
employees. This enabled us to understand the task negotiators
are facing and the mistakes people make. We learned what kind
of support an NSS should give to its users and in which form.
A major drawback of making concrete stimulus material is that
several experts also commented on and discussed particular
implementations, while these were only included in the videos
as examples and not as intended functionality. This happened
even though experts were explicitly instructed not to pay
attention to these details. We conclude that careful weighing is
necessary regarding the amount of detail put into concrete
stimulus material in order for a focus group to react upon the
right level of abstraction.
In the future we will test users’ acceptability of NSS in the
different use contexts and conduct field studies in order to get a
grasp of negotiation practice from a users’ point of view.
6. ACKNOWLEDGMENTS
We would like to thank the project team as well as the
negotiation experts for valuable input. This research is
supported by the Dutch Technology Foundation STW, applied
science division of NWO and the Technology Program of the
Ministry of Economic Affairs. It is part of the Pocket Negotiator
project with grant number VICI-project 08075.
7. REFERENCES
[1] Bellucci, E., Lodder, A. R., and Zeleznikow, J. 2004
Integrating artificial intelligence, argumentation and game
theory to develop an online dispute resolution
environment. Proceedings of the 16th IEEE International
Conference on Tools with Artificial Intelligence. pp. 749754.
[2] Bui, T. 1994 Evaluating Negotiation Support Systems: A
Conceptualization. In: 27th Annual Hawaii International
Conference on System Sciences, pp. 316—324. IEEE
Press, Hawaii.
[3] Bruseberg, A. and McDonagh-Philp, D. 2002 Focus
groups to support the industrial/product designer: a
review based on current literature and designers’
feedback. Applied Ergonomics 33 (1), 27—38
[4] Carey M.A. 1995 Comment: concerns in the analysis of
focus group data. Qualitative Health Research 5, 487–
495.
[5] Carroll, J. M. 2000 Making Use: Scenario-based Design
of Human-Computer Interactions. MIT Press, Cambridge.
[6] Carroll, J. M. 2000 Five reasons for scenario-based
design. Interacting with Computers 13, pp. 43—60.
[7] Goodman, J., Dickinson, A.,and
Syme, A. 2004
Gathering Requirements for Mobile Devices using Focus
Groups with Older People. To appear in Designing a
More Inclusive World, Proceedings of the 2nd
Cambridge Workshop on Universal Access and Assistive
Technology (CWUAAT), Springer
[8] Harvard Business Essentials: Negotiation. Harvard
Business School Publishing Corporation, Boston 2003
[9] Hindriks, K.V. and Jonker C.M. 2008 Creating HumanMachine Synergy in Negotiation Support Systems:
Towards the Pocket Negotiator. In: HuCom, pp. 47—55.
Delft.
[10] Kersten, G. E. and Lo, G. 2003 Aspire: an integrated
negotiation support system and software agents for e-
business negotiation, International Journal of Internet and
Enterprise Management 1 (3) pp. 293 – 315.
[11] Lee K.C., Kang I. and J.S. Kim, J.S. 2007 Exploring the
user interface of negotiation support systems from the
user acceptance perspective. Computers in Human
Behavior 23(1) pp.220—239.
[12] Neerincx, M.A.2003 Cognitive task load design: model,
methods and examples. In: E. Hollnagel (ed.), Handbook
of Cognitive Task Design. Chapter 13. Mahwah, NJ:
Lawrence Erlbaum Associates, pp. 283–305.
[13] Nunamaker, J. F., Briggs, R. O. , Mittleman, D.D. (1996).
Lessons from a Decade of Group Support Systems.
Proceedings of the 29th Hawaii International Conference
on System Sciences Volume 3: Collaboration Systems
and Technology, IEEE Computer Society.
[14] Rangaswamy, A., and Shell, G. R. 1997 Using Computers
to Realize Joint Gains in Negotiations: Toward an
Electronic Bargaining Table. Management Science 43
(8), pp. 1147—1163.
[15] Scaife, M., Rogers, Y., Aldrich, F., and Davies, M.1997
Designing for or designing with? Informant design for
interactive learning environments. In: Proceedings of
CHI'97: Human Factors in Computing Systems, pp. 343-350. ACM, New York
[16] Sim, J.2001 Collecting and analysing qualitative data:
issues raised by the focus group. Journal of Advanced
Nursing 28 (2), 345—352.
[17] Smith, J.A. and Osborn, M. (2003) Interpretative
phenomenological analysis. In J.A. Smith (Ed.),
Qualitative Psychology: A Practical Guide to Methods.
London: Sage.
[18] Stewart, D. and Shamdasani, P. 1990 Focus Groups:
Theory and Practice. Sage
[19] Swaab,R., Postmes, T., and Neijens P. 2004 Negotiation
Support Systems: Communication and Information as
Antecedents of Negotiation Settlement. International
Negotiation 9, pp. 59—78
[20] Thompson, L. L. 2005 The Heart and Mind of the
Negotiator. Pearson Prentice Hall, New Jersey
[21] Vetschera, R., Kersten, G. and Koeszegi, S. 2006 User
Assessment of Internet-Based Negotiation Support
Systems: An exploratory Study. Journal of Organizational
Computing and Electronic Commerce 16 (2), pp. 123—
148.