Evaluating the quality of generated nonverbal behavior remains a major challenge in the development of generative models.While human evaluations are reliable, they are costly and impractical for large-scale or iterative optimization.In this work, we propose an objective evaluation framework based on aggregated ranks across multiple fidelity and diversity metrics, computed from both raw features and learned latent representations.Compared to existing works, our framework emphasizes consistency across multiple metrics, aiming to provide a more holistic assessment.

Research on non-verbal behavior generation for social interactive agents focuses mainly on the believability and synchronization of non-verbal cues with speech. However, existing models, predominantly based on deep learning architectures, often perpetuate biases inherent in the training data. This raises ethical concerns, depending on the intended application of these agents. This paper addresses these issues by first examining the influence of gender on facial non-verbal behaviors. We concentrate on gaze, head movements, and facial expressions. We introduce a classifier capable of discerning the gender of a speaker from their non-verbal cues. This classifier achieves high accuracy on both real behavior data, extracted using state-of-the-art tools, and synthetic data, generated from a model developed in previous work. Building upon this work, we present a new model, FairGenderGen, which integrates a gender discriminator and a gradient reversal layer into our previous behavior generation model. This new model generates facial non-verbal behaviors from speech features, mitigating gender sensitivity in the generated behaviors. Our experiments demonstrate that the classifier, developed in the initial phase, is no longer effective in distinguishing the gender of the speaker from the generated non-verbal behaviors.

This paper presents a study that assesses the effectiveness of Motivational Interviewing (MI) interventions on the engagement of adults over 50 years to do more physical activity. While MI's four step process may easily be adapted to online devices, modeling the MI approach based on empathy may be more challenging (as opposed to directive interviews in which the user is encouraged to adopt a favorable attitude from the outset). Three devices are compared: a non-directive MI-based website, a non-directive MI-based virtual agent and a directed MI-based one. Our results show that the non-directive virtual agent tends to be perceived as more empathetic and trustworthy than the directed one, and that it significantly raises the participants' self-efficacy to overcome barriers, and positively impacts intrinsic motivation. We discuss mentions several implications for the adaptation of MI to the context of ECA.

Training people for facing dangerous situations has always been expensive and difficult or dangerous to organize. New approaches use virtual environments and simulation in which the role of collaborative people is played by cognitive agents, which allows making training sessions when they were not previously possible, or else decreasing their cost. Two problems arise: firstly, it is necessary to simulate teams, composed of virtual agents and the learner, interacting in a collaborative context. Secondly, we need to support the definition of a variety of explainable and professional behaviours (i.e. representative of the expected behaviours of an expert in a given profession). It is extremely complex and time consuming. To facilitate the design of tools allowing to build adequate agents model, we propose a meta-model of human behavior simulation. The originality of this meta-model is to make the cognitive models, used for explainability, and the agent model completely independent, to avoid repeated implementations of various behaviours. We show how task selection works on a simple agent model, by means of a graph of influences and preferences. This means has the advantage of being intelligible by its visual nature and of facilitating the addition of new behaviours impacting on the selection. A preliminary evaluation has shown that human users perceive the variability of behaviour and that the provided explanations are relevant.

This paper proposes a logic-based model to study erroneous decision making by humans in accident reports. Our model is based on minimal belief revisions and forward chaining. It computes possible mental states that could explain the operator’s behavior. From this sequence, we extract logical patterns that correspond to possible cognitive biases responsible for the erroneous decision making. We apply this model on the crash of Air France’s 447 flight in 2009.

Embodied Conversational Agents (ECAs) are suffering from a lack of adoption from users. This paper emphasizes virtual intimacy as a lever to support social dimension of human-agent interaction and enhance user experience. Here, we present an autonomous virtual counselor, expert in tourism, which provides multimodal intimacy-related behaviors. Real tourists interacting in the wild with our intimate agent only perceived its honesty and genuineness, which was insufficient to enhance social presence and user experience. However, our results suggest that the perception of virtual intimacy could be a strong predictor of social presence and user experience. When perceived, virtual intimacy may be, then, an important mean to foster user experience and ECAs adoption.

Health care delivery in military conflict, in peacekeeping missions or in the aftermath of disaster, implies high stress environments with danger exposures, life-threatening events and high levels of work demand. Crisis and emergency risk communication remains a real challenge. The rapid response of emergency medical teams has an important role to play in preventing serious adverse events. During critical events, medical errors can be related to human or system factors, including ineffective team leadership, non-standardized team communication, a lack of global situation awareness, poor use of resources and inappropriate triage and prioritization. The VICTEAMS project aims at building a virtual environment for training rescue team leaders to non-technical skills. Depending on the leader abilities, the virtual characters playing the subordinates (e.g. nurses) reproduce a variety of behaviors like erroneous actions, followership attitudes or stress-based behaviors. A pedagogical director tailors the difficulty according to the dynamic profile of the learner.

Building a long-term relationship between human and virtual agent remains a challenge. This paper explores the influence of intimate behavioral cues and the impact of different interaction modalities (voice, text, gesture) on our perception of intimacy. We built a virtual Tourism Information (TI) counselor capable of intimate behaviors based on grounded theories of intimacy developed in psychology. We studied how external observers perceive the social behaviors of the agent during its interactions with a human tourist. Our originality is to evaluate the perception of social skills across the interaction. Our results show that nonverbal behaviors of the virtual agent reinforces the impact of verbal cues on human perception on virtual intimacy. Findings also suggest that textual medium of communication has a negative impact on the perception of virtual intimacy, whatever the level of animation and intimacy exhibited by the TI counselor.

International audience

In the context of the development of Serious Games, this position paper sheds light on their possible use to support and enhance social inclusion through the presentation of three research projects in this domain. We first describe the context of our works and research. We then report on several events joining the scientific community that focused on the research on Digital Games for Empowerment and Inclusion. Last, we propose some results of our experience in this domain. The presented work particularly highlights the importance of user-centered design in interdisciplinary cooperation with psychologists, sociologists, vulnerable end users and the practitioners. We stress the relevance of progress beyond the state-of-the-art in recently focused areas of AI, such as, in computational theory-of-mind, in affective multimodal computing, in learning of socio-emotional skills, and in recovery of human attention processes. Typical examples are given to illustrate the projects' research results.

Digital Games for Empowerment and Inclusion have the potential to improve our society by preparing particular groups of people to meet social challenges in their everyday lives, and to do so in an enjoyable way through games. These games are developing rapidly to exploit new algorithms for computational intelligence supported by increasing availability of computing power to help analyze players' behavior, monitor their motivation and interest, and to adapt the progress of the games accordingly. The workshop on Intelligent Digital Games for Empowerment and Inclusion (IDGEI) explores the use of machine intelligence in serious digital games. In this context, we summarize the third international workshop on IDGEI held at the International Conference on Intelligent User Interfaces (IUI) 2015.

This paper presents a method for evaluating the credibility of agents’ behaviors in immersive multi-agent simulations. It combines two approaches. The first one is based on a qualitative analysis of questionnaires filled by the users and annotations filled by others participants to draw categories of users (related to their behavior in the context of the simulation or in real life). The second one carries out a quantitative behavior data collection during simulations in order to automatically extract behavior clusters. We then study the similarities between user categories, participants’ annotations and behavior clusters. Afterward, relying on user categories and annotations, we compare human behaviors to agent ones in order to evaluate the agents’ credibility and make their behaviors explicit. We illustrate our method with an immersive driving simulator experiment.

In this paper, we consider the problem of modeling complex systems at several levels of abstraction. We design SIMLAB, a multi-level model for multi-agent based simulation. Our approach is based on the coexistence of different levels during simulation to enhance the model with complementary experts’ opinion. We present how a same concept can be defined independently of its granularity using the notion of modeling axis. We consider recursive agents with interactions and influences which captures the inter-level dynamics. We also propose observations to detect and to reify macroscopic entities.

International audience

One of the main challenges faced by the video game industry is to give life tobelievablenonplayer characters (NPCs). Research shows that emotions play a key role in determining the behavior of individuals. In order to improve the believability of NPCs' behavior, we propose in this paper a model of the dynamics of emotions taking into account the personality and the social relations of the character. First, we present work from the literature on emotions, personality, and social relations in computer science and in human and social sciences. We focus on the influence of personality on the triggering of emotions, and the influence of emotions on the dynamics of social relations. Based on this work, we propose a dynamic model of the socioemotional state and its implementation as part of a tool for game programmers. This tool aims at the simulation of the evolution of emotions and social relations of NPCs based on their personality and roles.

This paper presents a new semantic relatedness measure on semantic networks (SN) that uses both hierarchical and non-hierarchical relations. Our approach relies on two assumptions. Firstly, in a given SN, only a few numbers of paths can be considered as “semantically correct” and these paths obey to a given set of rules. Secondly, following a given edge in a path has a cost (which depends on its type, is-a, part-of, etc.) and its position in the path. We propose an evaluation of our measure on WordNet with two different benchmarks, using the part-of relation. We show that, in this context, our measure does better than the classical semantic measures.

A challenge for pervasive computing is the seamless integration of computer support with users' activities in a very dynamic setting, with deep human and resource mobility. Portable devices such as notebooks, PDAs and smartphones are becoming more and more popular, as their computational power increases and their prices fall. Moreover, the quick spread of wireless networks has permitted the exchange of data in places as diverse as university campus, airports, coffee houses and family homes. In this paper we introduce campus, a framework that supports the development of multi-agent, context aware, pervasive computing applications. Campus is designed to provide the necessary infrastructure for ambience intelligence applications.

The intelligent agents are very useful for World Wide Web applications. Intelligent agents are designed and implemented for a variety of tasks in diverse range of applications: managing e-mail, navigating and retrieving information from the Internet,