Embodied Conversational Agents (ECAs)

Die Analyse der ersten zehn Minuten einer audio aufgezeichneten psychoanalytischen Sitzung fuhrt zu der Entdeckung und Beschreibung einer psychoanalytischen Prozedur, die wir als »Tanz der Einsicht« bezeichnen. Die Sitzung ist nach den Standards der Konversationsanalyse transkribiert und wird auf drei Ebenen analysiert. Zunachst kommt eine klinische Analyse zu dem Ergebnis, dass die Analytikerin bei zwei Gelegenheiten ihre Gesprachsposition verandert; sie spricht, als ware sie die Patientin. Die klinische Perspektive bringt diese Beobachtung mit dem Konzept der Modellszene und dem methodischen Prinzip der »Komplettierung der Szene« in Beziehung und einigen weiteren Befunden der Sauglingsforschung. Zum zweiten lasst sich mit den Mitteln der Konversationsanalyse erkennen, wie die Therapeutin dies Manover ausfuhrt und wie die Patientin darauf reagiert. Diese Analyse beschreibt detailliert die »slots«, an denen ein solches Manover auf naturliche und ungekunstelte Weise ausgefuhrt wird. ...

A voice assistant (VA), a type of voice-enabled artificial intelligence, is no longer just a character in science fiction movies. Currently, voice is embedded in a variety of products such as smartphones (mobile applications) and smart speakers in consumers’ homes. Furthermore, voice assistants are becoming integral to our daily lives. While human personalities shape the way we interact with the world, voice assistant personalities can also impact everyday interactions with our environment. This study identifies seven voice assistant personality traits (VAP) of three commonly used mobile applications: Microsoft’s Cortana, Google’s Assistant, and Amazon’s Alexa. To examine the effect of VAP on consumer experience, this study applies and extends flow theory to uncover why VAP has the effects it has and what facets of VAP drive the voice interaction flow experience that can influence consumers’ attitudes and behavioral intentions. Our study shows that voice interaction with a VA that incorporates functional intelligence, sincerity, and creativity empowers consumers to take control of their voice interactions with the VA, focus on their voice interaction, and engage in exploratory behavior. Consumers’ exploratory behavior leads to consumer satisfaction and consumers’ willingness to continue using voice assistant.

Mental illness is common in the United States. About one in four adults suffers from some form of mental illness in a given year (Holmes, 1/14/2015).

This level of occurrence is even higher for college students—perhaps as high as one in two according to the article, Delivering Cognitive Behavior Therapy to Young Adults with Symptoms of Depression and Anxiety Using a Fully Automated Conversational Agent (Woebot): A Randomized Controlled Trial (Fitzpatrick, et al., April-June, 2017.) In a related article, Depression and College Students, Michael Kerr points out that financial worries due to high debt and poor employment prospects coupled with failed relationships, lack of sleep, poor eating habits, and not enough exercise frequently result in the development of depression (Kerr, 3/29/2012). There are also many life transitions and challenges to a student’s identity.

In their effort to overcome this fear of stigma, Stanford researchers developed a virtual therapist, or conversational agent (often called a chatbot). The chatbot employs artificial intelligence and natural language processing to deliver cognitive behavior therapy (CBT) to college students self-identified as suffering from significant depression and/or anxiety.

Stanford's virtual therapist is named Woebot. Like many chatbots, Woebot uses Natural Language Programming to process student responses to questions posed by the virtual therapist, then guides the conversation to an appropriate node of a decision tree to provide suggested actions.

Perception and attention mechanisms are of great importance for entities situated within complex dynamic environments. With roles extending greatly beyond passive information services about the external environment, such mechanisms actively prioritise, augment and expedite information to ensure that the potentially relevant is made available so appropriate action can take place. Here, we describe the rationale behind endowing artificial entities, or virtual agents, with real-time perception and attention systems. We cover the fundamentals of designing and building such systems. Once equipped, the resulting agents can achieve a more substantial connection with their environment for the purposes of reacting, planning, decision making and, ultimately, behaving.

With the rise in popularity of chatbot and virtual-agent platforms, from Apple, Amazon, Google, Microsoft, Facebook, IBM and more, a new design discipline is emerging: Conversational UX Design. While it is easy to create natural language interfaces with these platforms, creating an effective and engaging user experience is still a major challenge. Natural language processing (NLP) techniques have given us powerful tools for analyzing bits of language, but they do not tell us how to string those bits together to make a natural conversation. Natural conversation has a sequential organization that is independent of the organization of language itself. At IBM Research-Almaden, we are addressing this user experience (UX) design challenge by applying formal, qualitative models from the field of Conversation Analysis to the design of conversational agents. Our Natural Conversation Framework (NCF) is a design framework for conversational user experience. It provides a library of generic conversational UX patterns that are inspired by natural human conversation patterns and that are agnostic to platform and input method (text or voice). This chapter will cover the four components of our Natural Conversation Framework: (1) an interaction model, (2) common activity modules, (3) a navigation method and (4) a set of sequence metrics. In addition, it will briefly outline a general process for designing conversational UX: from mock-up to working prototype.

En este documento se presentan los diferentes tipos de agentes conversacionales como lo son: los Agentes Conversacionales Animados, Humanos Virtuales Creíbles, Asistentes Virtuales,
así como sus características, áreas de aplicación, y roles. Existen diversas empresas que se dedican a crear dichos agentes, los cuales deben contar con una base de conocimientos en un tema en específico. Para lograr que un agente conteste coherentemente se puede llevar a cabo un análisis de la comunicación, el cual es importante en la interacción cara-a-cara, con el fin de analizar la conversación y el discurso; los que tienen como base los actos del habla, para este caso en particular: los actos locutivos. Esto con el objetivo de llegar a crear un Sistema de Diálogo que tenga el rol de un Asistente Virtual.

"Este trabajo de investigación se origina a partir de un proyecto realizado en la institución relacionado con un agente conversacional corpóreo, ya que actualmente el uso de los Agentes Conversacionales Animados ha ido creciendo. Estos suelen ser una herramienta utilizada en ciertas funciones en las cuales pueden servir como guías, maestros, ayudantes, etc.

La problemática que abordamos en éste documento, es la implementación de un agente conversacional capaz de expresar diálogos creíbles en lenguaje natural, expresando en sus diálogos rasgos de personalidad, emociones y su intensidad. Para la generación de diálogos creíbles, proponemos utilizar la extensión que se le hizo al lenguaje AIML (Artificial Intelligence Markup Language), utilizando la nueva estructura para los archivos de la base de conocimientos basados en AIML, mediante diálogos creíbles, es capaz de crear escenarios más reales para que el aspirante disperse sus dudas con relación a la
Maestría en Ciencias en Ciencias de la Computación, manteniendo una conversación simulada por el agente conversacional. Para esto se presentó una metodología para organizar y diseñar la estructura de los diálogos de un Agente Conversacional Animado capaz de expresar emociones en su diálogo. Resultando de éste trabajo un Asistente Virtual con Diálogo Emocional que proporciona información sobre la Maestría en Ciencias en Ciencias de la Computación del ITCM."

Conversational agents are becoming an essential part of a growing number of personal and commercial encounters, bringing the issue of Conversational Marketing to a broader audience. A conversational agent is a developing technology that will be used in various fields throughout life, including e-commerce. The common characteristics of any conversational agent in whatever area are their capacity to engage in one-to-one personalised real-time dialogue with a human user and their availability 24 hours a day. Scale items for conversational agent phenomena have not been created scientifically or managerially in a business environment. The primary goal of this study was to develop and validate a new scale for conversational agents that could be used to quantify individual interactions in conversational marketing. As a result, the creation of a new scale for conversational agents with the objective of measuring individual customer interactions in conversational marketing was separated into two phases: Scale Development and Scale Validation. The Conversational Agent Usage Scale was developed and validated as a consequence of pilot studies. Additionally, this article discusses the practical consequences of conversational marketing, which can now be accomplished through the use of the Conversational Agent Usage Scale, which may be used by Customer Service & Support, Marketing, and Sales departments.

“You see, there’s a primal joy in hitting a thing in motion. It’s one of the oldest pleasures there is. Something moves, boo, you wing it. Beast, bird or human, the thing to do is to knock it down. It's primal, Davy. It's basic to the origin of the species.”
-Don DeLillo, Americana

We do not consider it presumptuous to write about the ineffable nature of God nor the strategy for engaging in the War in Heaven.  In point of fact the “War” can be discussed in scientific terms with an admixture of correlated so-called “mystical” terms.  During the “End Times”—sometimes called the “Paradigm Shift”—this Cosmic battle is being projected into Earth’s fractal-microcosmic Media-sphere where the signs of conflict between forces of “harmony” and of “disharmony” are abundantly evident. These signs can be interpreted in the same way that Jungian dream symbols are interpreted in order to foster coherent states of Psyche for the purpose of healing (making whole or coherent) the humanitarian persona. This healing or “making whole” must be understood in terms of Heart-Mind coherence which is now measurable in terms of mediated images in the dreamscape of the Media-sphere. (Schafer, Stephen Brock, all; McCraty, Rollin/HeartMath, all) During this time of paradigm shift in scientific parameters in which a “fractally structured Trinitarian” Cosmos exhibits an emergent holographic ontology, a new scientific language for “Divinity” must be established in order to optimize benign humanitarian strategies for opposing the existential forces of chaos.  From ancient times, prophets and heretics throughout the world have spoken about the scientific nature of the War in Heaven.

This chapter deals with the communication of persuasion. Only a small percentage of communication involves words: as the old saying goes, “it’s not what you say, it’s how you say it”. While this likely underestimates the importance of good verbal persuasion techniques, it is accurate in underlining the critical role of non-verbal behaviour during face-to-face communication. In this chapter we restrict the discussion to body language. We also consider embodied virtual agents. As is the case with humans, there are a number of fundamental factors to be considered when constructing persuasive agents. In particular, one who wishes to persuade must appear credible, trustworthy, confident and non-threatening. Knowing how not to behave is also a vital basis for effective persuasion. This includes resolving task constraints or other factors with the social perception considerations. These social virtual agents face many of the same problems as humans have in controlling and expressing themselves in an appropriate manner so as to establish and maintain persuasive interaction. All along the chapter, much of our discussion will handle concepts applicable both to agent and human behaviour.

"Es conocido que la comunicación no verbal involucra el uso de expresiones faciales, gestuales y posturas para comunicar intenciones, objetivos, sentimientos e ideas; lo que permite reforzar un mensaje dentro de una dinámica de diálogo. Las habilidades de comunicación no verbal son propias de los seres humanos y por más de una década los investigadores han intentado transferir dicha riqueza a un personaje virtual, y a través de esto facilitar la comunicación hombre-máquina.
En la actualidad la comunicación no verbal juega un papel importante en el desarrollo de personajes virtuales, debido a que es un medio que permite mostrar rasgos de personalidad, manifestar un estado de ánimo y reforzar la expresión verbal, buscando así un vínculo de empatía entre el usuario y el personaje.
En esta tesis se desarrolló una metodología para identificar a través de un diseño experimental y una serie de pruebas estadísticas, el conjunto mínimo de atributos que permita caracterizar las expresiones gestuales semióticas (brazos y manos) de personajes virtuales, considerando la influencia de la personalidad, el estado emocional y actos del habla. Los resultados de esta investigación muestran que el conjunto mínimo con el que se caracterizó a la expresión gestual permite discriminar entre gestos deícticos e icónicos.
La integración de los rasgos de personalidad y estado emocional durante el proceso de caracterización tiene como propósito emular la riqueza de la comunicación humana, buscando tornar creíble al personaje virtual y por consiguiente incrementar la inmersión en el mundo virtual en el que se desenvuelve el personaje."

Social interaction in virtual universe opens new horizons in the field of information technologies, and particularly, in the use of video games for professional purposes (a.k.a. serious game). Introducing a certain level of realism in the social and emotional interactions of virtual characters makes it possible to consider many professional applications, which could have only been developed until now in the field of human-human interaction.
The main difficulty of this approach is to produce a human interaction dynamic that makes sense for the tasks and actions that must be achieved by the characters. In order to develop this kind of dynamics, we have referred to (i) the approaches of situated cognition (so as to take into account the role of the objects of the environment), and (ii) representational approaches for the production of speech acts.
We focus on the production of verbal, nonverbal and paraverbal interactions between a human and a virtual character in the context of virtual therapy. Initially, we were interested in the description of the dynamics of the visual, gesture and verbal interactions between a patient and a real therapist. This study enabled us to categorize the phases of dialogue and to elaborate a model of the dynamics of the interaction, taking into account the personality of the actors, their intentions, and their verbal and gesture behaviours.
From this model, we have elaborated a generic architecture, which is articulated around a behavioural model and a kinesics model. The behavioural model selects the speech acts and
the attitude of the character (empathic, neutral, etc). The kinesics model selects and integrates the facial, posture and gesture expressions related to the results of the behavioural model.
From a theoretical point of view, we were interested in the concept of presence in order to identify which was the nature of the relevant couplings making it possible to produce a feeling of social presence (i.e. a feeling of living a social interaction not mediated by the elements of the technical device for the human actors which animate their characters).
Finally, the device was used within the framework of a research program on the concept of mirror neurons. The assumption is that the presentation of a virtual image of its handicapped member to the patient would enable him to better recover. For this purpose,
IRMf tests associated with each experiment would allow to confirm or to invalidate the role of these mirror neurons in the process of recovery.

"Este trabajo de investigación se origina a partir de un proyecto de la Universidad de Barcelona para desarrollar habilidades en el diagnóstico del Trastorno de Ansiedad Generalizada a estudiantes de psicología y psiquiatría utilizando un chatbot, el cual es un programa que intenta simular una conversación humana con una persona [Laven, 1996].

La problemática que abordamos en éste documento, es convertir un chatbot en un agente conversacional capaz de generar diálogos creíbles y dinámicos en lenguaje natural, expresando en sus diálogos rasgos de personalidad, emociones y su intensidad. Para la generación de diálogos creíbles, proponemos realizar una extensión al lenguaje AIML (Artificial Intelligence Markup Language), proponiendo una nueva estructura para los
archivos de conocimiento basados en AIML, la cual será la estructura de la base de conocimientos de nuestro agente conversacional.

El agente conversacional, mediante diálogos creíbles, es capaz de crear escenarios más reales para que el estudiante diagnostique de forma contundente el padecimiento simulado por el agente conversacional. Para determinar si los usuarios perciben que a través del diálogo textual se expresa una emoción y para determinar si una imagen refuerza la emoción transmitida en los diálogos se aplicó una encuesta. Las emociones evaluadas fueron Miedo, Alegría, Tristeza, Enojo, Angustia, Placer, Sorpresa y Resignación. La encuesta se dividió en dos partes y muestra una pequeña conversación entre un usuario y el ECA, en la primera parte, se utilizaron sólo frases y en la segunda, éstas fueron reforzadas con imágenes del agente conversacional, acordes a la emoción a transmitir en el diálogo. Los resultados muestran que el porcentaje de identificación de la emoción y de su intensidad en las frases fueron del 57.50% y del 43.92% y que los porcentajes de identificación al reforzar las frases con imágenes fueron del 70.51% y del 55.35%, respectivamente. Estos resultados indican que utilizar imágenes reforzando a las frases contribuye en la identificación de las emociones y de sus intensidades."

Zusammenfassung Die Analyse der ersten zehn Minuten einer audio aufgezeichneten psychoanalyti-schen Sitzung führt zu der Entdeckung und Beschreibung einer psychoanalyti-schen Prozedur, die wir als »Tanz der Einsicht« bezeichnen. Die Sitzung ist nach den Standards der Konversationsanalyse transkribiert und wird auf drei Ebenen analysiert. Zunächst kommt eine klinische Analyse zu dem Ergebnis, dass die Analytikerin bei zwei Gelegenheiten ihre Gesprächsposition verändert; sie spricht, als wäre sie die Patientin. Die klinische Perspektive bringt diese Beobachtung mit dem Konzept der Modellszene und dem methodischen Prinzip der »Kom-plettierung der Szene« in Beziehung und einigen weiteren Befunden der Säug-lingsforschung. Zum zweiten lässt sich mit den Mitteln der Konversationsanalyse erkennen, wie die Therapeutin dies Manöver ausführt und wie die Patientin darauf reagiert. Diese Analyse beschreibt detailliert die »slots«, an denen ein solches Manöver auf natürliche und ungekünstelte Weise ausgeführt wird. Die Analyse der Gesprächsstruktur wird vervollständigt durch einige Skizzen zur prosodischen Form, deren akustische Realität in PRAAT-Graphiken illustriert wird. Zu Beginn spricht die Patientin vielfach ohne jede Modulation der Grundfrequenz – eine Intonation, die wir mit der klinischen Beschreibung der Affektisolation und der geringen Selbstbehauptung der Patientin verbinden. Die Therapeutin manövriert die Patientin aus dieser Position, indem sie deren Rolle einnimmt. Sie tut das nicht nur mit der Bedeutung der Wörter, sondern auch mit prosodischen Tech-niken: Sie artikuliert ihre Redezüge mit einem Plateau hoher Töne, das auch in anderen Zusammenhängen Assertionen vom Sprecher entkoppelt, was wir in diesem Falle tentativ als »Zuschreibung von Selbstbehauptung« deuten. Das Gesamt dieser Operation ist, was wir »Tanz der Einsicht« nennen, eine Bewegung des Gesprächs, die zu einer Veränderung metaphorischer Konzepte und der Ge-sprächspositionierung führt. Die Prosodie unterstützt dabei den symbolisch-se-mantischen Prozess. Summary In this paper we use the first ten minutes of an audiorecorded 28th psychoanalytic session in order to detect and describe a new therapeutic procedure which we call «Dancing Insight ". The session is transcribed according to the standards of conversation analysis and is analyzed at three levels: first, a clinical analysis which leads to the interpretation that the analyst at two opportunities actively turns the table-speaking as if she were the patient. The clinical view can bring this observation in relationship to the concept of «model scenes " and «completing the scene " and some further findings from infant research. Second, the conversation analytic approach observes how the therapist does this maneuver and how the patient responds to it. A further analysis is directed to the «slots " where such a maneuver can be conducted with a natural flair and without any irony. Third, we complement the picture by an acoustic analysis of some prosodic features using PRAAT-software which shows some interesting patterns. At the beginning the patient speaks with a flat intonation, a prosodic property which we relate to affect isolation and its consequence, the lack of assertive commitment. The therapist manages to drag the patient out of this stance by assuming her role and she does this not only by the meaning of words, but also with prosodic means: She pronounces her turns at a «high plateau " , the basic meaning of which we tentatively describe as «projection of the assertion to the partner ". This is the procedure we call «dancing insight " : The conversational move which leads to a change of metaphorical concepts and to a change of positioning is achieved by a prosodic technique.

One of the major problems of user’s interaction with Embodied Conversational Agents (ECAs) is to have the conversation last more than few second: after being amused and intrigued by the ECAs, users may find rapidly the restrictions and limitations of the dialog systems, they may perceive the repetition of the ECAs animation, they may find the behaviors of ECAs to be inconsistent and implausible, etc. We believe that some special links, or bonds, have to be established between users and ECAs during interaction. It is our view that showing and/or perceiving interest is the necessary premise to establish a relationship. In this paper we present a model of an ECA able to establish, maintain and end the conversation based on its perception of the level of interest of its interlocutor.

The first part of the paper develops the argument that geographers should learn to decompose human memory into its constituent parts because then and then alone will we become attuned to the full range of ways in which we incorporate places into our beings. The second part of the paper articulates Stephen Hill's comments on episodic memory with my recent work on wisdom.

This is a CFP for my forthcoming book, "Embodied Research Methods," slated to be a Routledge graduate textbook. We are seeking original methods designed to address embodied data (collection and/or analysis) from all fields of study. Please see guidelines attached, and please forward this to any other interested parties.

The increase in the spread of conversational agents urgently requires to tackle the ethical issues linked to their design. In fact, developers frequently include in their products cues that trigger social biases in order to maximize the performance and the quality of human-machine interactions. The present paper discusses whether and to what extent it is ethically sound to intentionally trigger gender biases through the design of virtually embodied conversational agents. After outlining the complex dynamics involving social biases, social robots, and design, we evaluate the ethics of integrating gender cues in conversational agents, analysing four different approaches to the problem. Finally, we suggest which approach in our opinion might have the best chances to reduce the negative effects of biases and discriminatory visions of gender dynamics.

In order to engage with user on more personal level, natural human-machine interaction is starting
to virtual or even physical entities resembling human collocutors. Namely, through human-like entities the
designed multimodal interaction models, try to adapt to user’s context and to facilitate context of the
conversational situation. The designed multimodal processes tend to follow the ‘’rules’’ and ques as implanted
during face-face interaction among humans. In this way, the integration and exploitation of embodied
conversational agents (ECA) for human-like interaction seems only natural. The ECA’s artificial body and
articulation capabilities are already close to those found on real humans. From skin, face, hands, and body
posture, these virtual entities tend to look and behave as realistically as possible. Furthermore, ECAs tend to
imitate as many features of human face-face dialogs as possible, and integrate them into interaction as
synchronized as possible. One of the essential functions in face-to-face interaction is the ability to reproduce
synchronized verbal and co-verbal signals coupled into conversational behavior. Other signals, such as: social
cues, attitude (emotions), personality, eye-blinks, and spontaneous head movements are equally important, and
have to be blended into multimodal expressions. However, designing a realistic entity that acts realistically like
humans do, is a daunting task. Modern 3D environments and 3D modeling tools, such as: Maya, Daz3D,
Blender, Panda3D and Unity, have opened up a completely new possibilities to design virtual entities, which
appear almost (if not completely) like real-life persons. However, the modern 3D technology mostly covers the
design and deployment part of such realism, while the realism of the behavior itself and its diversity and
dynamic nature are generally not the focus of 3D modeling frameworks. Namely, most of the animation
prepared in 3D frameworks is planned, designed in advance, and over well-written and well-designed
situations. Therefore, it integrates limited diversity as well as limited capacity to adapt to a new set of
parameters. As a result, 3D frameworks have a limited capacity to handle highly dynamic and interchangeable
contexts present in human interaction. In this paper we, therefore, outline EVA U-Realizer, the second
generation of proprietary behavior realization module. The goal of the realizer is to integrate the diversity,
responsiveness, and adaptiveness that is required for the facilitation of conversational responses with animation
capacities and realism provided by 3D modeling frameworks. As a result, the presented novel realizer is built
over Unity 3D game engine’s core. The proposed realizer considerably improves the capabilities of the
animation engine itself, by providing interpreter and executor for incorporating dynamic and real-time
generated conversational artefacts, similar to those found in conversations among humans.

Recent research has concentrated on the question how virtual characters (‚agents’) and their set of reaction can be enhanced to better suit a certain purpose in a virtual reality environment (VRE). Another field of active research is augmented reality (AR), where computer generated graphics is superimposed over a real world image. Thus AR is able to add detail and enhance the underlying information.

Our idea is to create an augmented reality environment (ARE) with an agent for conversation centred areas such as presentations. Our goal is to find evidence in literature and research how agents are employed as effective means of communication. This includes a review of existing systems and a summary of user studies. We will a look at combining AR technology with agents, and what special issues arise. With this background we will develop design objectives and requirements of agents in AR. A prototype implementation of an agent architecture with a digital character as user front-end will be implemented. From our literature review and our practical implementation we will identify a list of fruitful areas of research and derive a list of challenging new questions.

In the last few years, the use of ontologies has spread thanks to the irruption of the Semantic Web. They have become a crucial tool in information systems as they explicitly state the meaning of information, making it possible to share it and to achieve higher levels of interoperability. However, being knowledge representation models as they are, other fields can take advantage of their characteristics to extend their capabilities. In particular, in the context of Embodied Conversational Agents, they can be used to provide them with semantic knowledge and, therefore, enhance their intellectual skills.
In this paper, we propose an approach to explore the synergies between these technologies. Thus, we have developed a multimodal ECA that exploits the knowledge provided by the Linked Data initiative to help users in their search information tasks. Based on a semantic-guided keyword search, our approach is flexible enough to: 1) deal with different Linked Data repositories and 2) handle different search/knowledge domains in a multilingual way. To illustrate the potential of our approach, we have focused on the case of DBpedia, as it mirrors the information stored in the Wikipedia, providing a semantic entry to it.

One smart speaker to rule them all, one sensor to find them, one device to bring them all and in the Internet of Things bind them. Conversational User Interfaces (CUI) represent an interactive phenomenon that is disrupting social interactions as we knew them. The spectrum of legal issues that arises from these user-device relationships has a wide range of impacts on society. This paper addresses seven legal issues arising from these interactions that affect Privacy and Law concerning consumer protection, profiling, liability, security, neutrality, influence and data protection. It aims to prompt an interdisciplinary discussion between Human-Computer Interaction, legal and Privacy researchers in order to address the lack of regulation around aspects of CUI and stimulate legal, policy and design confrontations. CCS CONCEPTS • Applied computing → Law; • Security and privacy → Human and societal aspects of security and privacy; • Human-centered computing → Natural language interfaces; Empirical studies in HCI .

When humans are addressing multiple robots with informative speech acts (Clark & Carlson 1982), their cognitive resources are shared between all the participating robot agents. For each moment, the user's behavior is not only determined by the actions of the robot that they are directly gazing at, but also shaped by the behaviors from all the other robots in the shared environment. We define cooperative behavior as the action performed by the robots that are not capturing the user's direct attention. In this paper, we are interested in how the human participants adjust and coordinate their own behavioral cues when the robot agents are performing different cooperative gaze behaviors. A novel gaze-contingent platform was designed and implemented. The robots' behaviors were triggered by the participant's attentional shifts in real time. Results showed that the human participants were highly sensitive when the robot agents were performing different cooperative gazing behaviors.

Conversational agents are attributed humanlike characteristics; in particular, they are often given a gender. There is evidence that gender sets up expectations that have an impact on user experiences with the agent. The objective of this paper was to explore gender affordances of conversational agents. Our examination takes a holistic approach to the analysis of the application of gender stereotypes to nine chatterbots: six embodied (three male and three female), two disembodied (male and female), and a robot embodiment. Building on social psychology research, our objective was to test the persistence of gender stereotypes in the selection of conversation topics and in the elicitation of disinhibition and verbal abuse. Our study is based on quantitative textual analysis of interaction logs. A dictionary of English sexual slang and derogatory terms was developed for this study. Results show that gender stereotypes tend to affect interaction more at the relational (style) level then at the referential (content) level of conversation. People attribute negative stereotypes to female-presenting chatterbots more often than they do to male-presenting chatterbots, and female-presenting chatterbots are more often the objects of implicit and explicit sexual attention and swear words. We conclude by calling for a more informed analysis of user interactions that considers the full range of user interactions.

The use of conversational agents within the aerospace industry offers quick and concise answers to complex situations. The aerospace domain is characterized by products and systems that are built over decades of engineering to reach high levels of performance within complex environments. Current development in conversational agents can leverage the latest retrieval and language model to refine the system's question-answering capabilities. However, evaluating the added-value of such a system in the context of industrial applications such as pilots in a cockpit is complex. This paper describes how a conversational agent is implemented and evaluated , with particular references to how state-of-the-art technologies can be adapted to the domain specificity. Preliminary findings of a controlled user experiment suggest that user perception of the usefulness of the system in completing the search task and the system's responses to the relevance of the topic are good predictors of user search performance. User satisfaction with the system's responses may not be a good predictor of user search performance.

This paper presents a new adversarial learning method for generative conversational agents (GCA) besides a new model of GCA. Similar to previous works on adversarial learning for dialogue generation, our method assumes the GCA as a generator that aims at fooling a discriminator that labels dialogues as human-generated or machine-generated; however, in our approach, the discriminator performs token-level classification, i.e. it indicates whether the current token was generated by humans or machines. To do so, the discriminator also receives the context utterances (the dialogue history) and the incomplete answer up to the current token as input. This new approach makes possible the end-to-end training by backpropagation. A self-conversation process enables to produce a set of generated data with more diversity for the adversarial training. This approach improves the performance on questions not related to the training data. Experimental results with human and adversarial evaluations show that the adversarial method yields significant performance gains over the usual teacher forcing training.

One of the major problems of user's interaction with Embodied Conversational Agents (ECAs) is to have the conversation last more than few second: after being amused and intrigued by the ECAs, users may find rapidly the restrictions and limitations of the dialog systems, they may perceive the repetition of the ECAs animation, they may find the behaviors of ECAs to be inconsistent and implausible, etc. We believe that some special links, or bonds, have to be established between users and ECAs during interaction. It is ...

We evaluated user perception of certain attention behaviours, eye head and body orientation in particular, of humanoid agents in a virtual environment for the purposes of interaction initiation. Users were shown a number of scenarios involving agents making gaze, gesture and locomotion behaviours and were asked to report their impressions of how attentive the agents were and their perceived likely-hood to engage in interaction. These evaluation studies are critical for furthering our understanding of the early phases of human-computer interaction where distance may be involved and there is uncertainty as to the intention of the other to interact. Amongst other  results, we establish here that the human perception of interest, interaction seeking and openness from a humanoid agent is based, in part, on the time-varying behaviour of its eyes, head, body and locomotion directions. Application domains include social robotics and embodied conversational agents.

L'interaction sociale en univers virtuel ouvre de nouveaux horizons dans le domaine des technologies de l'information et en particulier de l'usage des jeux video a des fins professionnelles (" serious game "). En introduisant la possibilite de gerer avec un certain niveau de realisme des interactions sociales et emotionnelles entre les agents conversationnels animes (ACA) representant des acteurs humains, cette approche permet d'envisager de nombreuses applications professionnelles qui jusqu'alors ne pouvaient etre envisagees que dans le cadre domaine de l'interaction homme-homme. La difficulte principale de cette approche est de produire une dynamique des interactions humaines qui ait un sens dans le cadre des taches et des actions qui doivent etre accomplies par les acteurs. Afin d'ancrer theoriquement cette question, nous avons fait reference aux approches de la cognition situee (afin de prendre en compte le role mediateur des objets de l&#39...

(Unfinished draft) Our Digital Culture research group in the Department of Arts and Cultural Studies at the University of Copenhagen made a collective visit in Milan to see the very well received exhibition at the Milan Osservatorio titled "Training Humans." (Prada 2019) The very idea or concept behind the exhibition, coming from Kate Crawford as an artificial intelligence (AI) researcher and Trevor Paglen as an artist-researcher, spoke quite directly to a set of concerns in our research group, and in our department in general. We have been at work for quite some time on issues concerning digital images and film, networked visual cultures, media art, and algorithmic cultures. Several of us have written about or participated in projects inquiring into the considerable advances of AI during the last decade. (Ekman 2020 (forthcoming)) The exhibition promised to address almost all of these concerns in one or more ways. It appeared a very obvious choice to visit the first major photography exhibition focusing on showcasing training images, the base collections of photos used in research labs to train AI systems to recognize and categorize the world, including such things as human beings. The way in which the sense of the title of the exhibition oscillated was itself of interest to us, even before we went. Anthropocentrism was far from certain. Perhaps a posthuman message respecting power, agency and knowledge was indeed at stake. For the title might be taken to hint at a basic, mundane set of images of humans to use for training in order that some rather