En creusant le sujet, je me suis rendu compte que de nombreuses recherches avaient lieu dans le domaine des émotions et notamment au niveau Européen. Le projet le plus révélateur est me semble t’il celui qui porte le nom de « HUMAINE »- HUman MAchine Interaction Network [9].
HUMAINE a pour but de poser les fondations du développement de systèmes pouvant enregistrer, modéliser et/ou influencer les processus et états émotionnels humains, mettre en œuvre des systèmes orientés émotion. De tels systèmes peuvent être centraux pour de futures interfaces, mais leurs fondations conceptuelles ne sont pas encore suffisamment avancées pour imaginer leur réel potentiel ou la meilleure façon de les développer. Une des raisons est que la connaissance pertinente est dispersée parmi de nombreuses disciplines. HUMAINE rassemble les experts éminents des disciplines clé dans un programme conçu pour réaliser cette intégration intellectuelle. Il identifie 6 domaines thématiques qui traversent les groupements traditionnels et il offre une structure pour une organisation approprier du travail :
-
Théorie des émotions.
-
Interfaces de signal/signe.
-
La structure des interactions émotionnellement colorées
-
Les émotions dans la cognition et l’action.
-
Les émotions dans la communication et la persuasion.
-
Application des systèmes orientés émotions.
Ces recherches n’intéressent pas que l’Europe. D’autres universités de part le monde ont emboîté le pas. On peut citer notamment les université de Tel Aviv et Haifa[10]
Parce qu’une présentation live vaut mieux qu’un long discours, ci-joint une vidéo de M Schroeder, un ponte du domaine présentant HUMAINE, il faut s’intéresser en particulier au passage démarrant à partir de 17 :00.
http://emotion-research.net/projects/humaine/aboutHUMAINE/MLMI-05-Schroeder.avi
Je ne sais pas vous mais lorsque je lis tout cela, je ne me sens pas particulièrement rassuré et je suis plutôt inquiet. Cette affaire n’est pas très médiatisée et nous ne percevons peut-être que les premières applications à l’insu de notre plein gré ! Qu’est ce que peut bien être un système orienté émotions ? Une intelligence artificielle pour le contrôle social ? Voilà bien une invention humaine qui pourrait ne pas vraiment l’être !
Poursuivons dans les trouvailles, connaissez-vous NECA[11] (Net Environment for Embodied Emotional Conversational Agents) ? NECA est un projet financé par la commission Européenne. NECA développe le concept de communication multimodale avec des personnages virtuels qui présentent des traits de personnalité et des comportements affectifs crédibles. C’est la combinaison de différentes branches de recherche dans le domaine de la parole et du langage naturel, de la sémiotique des expressions non verbales dans la communication et de la modélisation des émotions et de la personnalité. NECA est en fait une plateforme générique qui fournira les bases d’applications spécifiques qui pourront être développées.
Un exemple : Il concerne Internet et a pour objectif de mettre en relation les personnes. Les utilisateurs choisissent un avatar dynamique, formatent cet avatar à leurs goûts, préférences, personnalité et l’envoie sur des forums, mondes virtuels où il rencontrera d’autres agents, rentrera en contact avec ceux-ci et décidera de développer ou non une relation sociale virtuelle ! Le but est de rentre son avatar le plus populaire possible ! Bref la starisation virtuelle. On peut rêver mieux comme système pour créer des liens…Cela ne vous rappelle pas étrangement un programme qui s’appelle Second Life ? Il ne s’agit pas encore d’agents entièrement autonomes, mais cela ne devrait pas tarder.
On peut aussi se poser la question concernant les interventions d’un certain Seth Brundle Flight qui finalement pourrait être un de ces agents, lui bien réel dans ce monde virtuel !
Enfin, pour faire un tour d’horizon des « programmes » développés autour de NECA, vous trouverez en annexe quelques applications [12]. Tous nos sens comme vecteurs d’émotions sont concernés, de la musique aux jeux vidéo et la liste est loin d’être exhaustive [13].
Par exemple, serions-nous donc déjà dans la création de réflexes conditionnés à la Pavlov, en passant en boucle des évènements tragiques et leurs causes désignées (pour illustrer le propos, les attentats du 11/09 et l’association avec le terrorisme islamique) une façon de préparer le raisonnement motivé et les nouveaux conflits calculés ?
[9] HUMAINE.
http://researchprojects.kth.se/index.php/kb_1/io_9033/io.html
http://gaips.inesc-id.pt/gaips/proj.php
HUMAINE aims to lay the foundations for European development of systems that can register, model and/or influence human emotional and emotion-related states and processes - 'emotion-oriented systems'
HUMAINE aims to lay the foundations for European development of systems that can register, model and/or influence human emotional and emotion-related states and processes - 'emotion-oriented systems'. Such systems may be central to future interfaces, but their conceptual underpinnings are not sufficiently advanced to be sure of their real potential or the best way to develop them. One of the reasons is that relevant knowledge is dispersed across many disciplines. HUMAINE brings together leading experts from the key disciplines in a programme designed to achieve intellectual integration. It identifies six thematic areas that cut across traditional groupings and offer a framework for an appropriate division of labour - theory of emotion; signal/sign interfaces; the structure of emotionally coloured interactions; emotion in cognition and action; emotion in communication and persuasion; and usability of emotion-oriented systems. Teams linked to each area will run a workshop in it and carry out joint research to define an exemplar embodying guiding principles for future work in their area. Cutting across these are plenary sessions where teams from all areas report; activities to create necessary infrastructure (databases recognising cultural and gender diversity, an ethical framework, an electronic portal); and output to the wider community in the form of a handbook and recommendations of good practice (as precursors to formal standards).
http://emotion-research.net/projects/humaine/aboutHUMAINE/technical_annex_public.pdf
Beyond Robotics: in particular the development of cognitive (and emotion-oriented) robots that serve humans as assistants or companions, being able to learn in an active open-ended way and to grow in constant interaction and co-operation with humans.
The Disappearing Computer: emotion-oriented systems can play a vital role in the development
of future ambient systems – IT systems intimately integrated with everyday environments and supporting people in their activities – in particular when combined with wireless and mobile systems.
Life-like Perception and Cognition Systems: subsystems and complete autonomous artefacts that are inspired by architectures and mechanisms found in living systems. In addition to the issues highlighted under ‘Cognitive systems,’ emotions and their expressive and behavioural manifestations are one of the key elements to achieve ‘life-like’ qualities and the impression of life.
“Solving ‘trust and confidence’ problems so as to improve dependability of technologies, infrastructures and applications.” HUMAINE addresses ‘trust and confidence’ problems at two main levels. On the one hand, it aims to provide the technological community with a thorough ethical analysis, which gives users in general rational grounds to trust the systems (see Objective 4, section 2). On the other hand, it aims to provide a realistic assessment of the human dimension (the benefits and costs of emotion-oriented systems, cultural and gender diversity, etc.) which encourages a subjective sense of trust in the systems.
“Strengthening social cohesion by providing efficient, intelligent and easy to use systems for health, transport, inclusion, risk management, environment, learning and cultural heritage.” The relevance of HUMAINE to various strategic objectives of IST in FP6, such as eHealth, eInclusion, eSafety for road and air, Technology-enhanced learning, etc. has been extensively developed under the heading ‘scientific objectives’ in this section.
(a) Signal/sign interfaces (from raw signals to emotionally significant signs and vice versa) Emotion-related interactions are mediated by physical signals in various modalities – i.e. distributions of energy that fluctuate in time and/or space. The function of these signals is to carry signs – information structures that human or artificial systems use to signify particular types of object or event.
The signal/sign interface in emotion poses major technical challenges in both directions (signal to sign and sign to signal) and in several modalities. These include finding emotionally significant features in an image of a face, synthesising sound patterns that convey emotionally significant voice qualities, detecting patterns in physiological data that are systematically associated with emotion, and much more. The challenge in the signal/sign interface thematic area is to ensure that work on these low-level but difficult issues receives the attention it needs, and is both understood and informed by other parts of the community.
(b) Signs and emotional content in interactions (from signs to emotion descriptors and vice versa) Emotion-sensitive systems need to embody knowledge about the relationships between emotional states and the signs, in many modalities, by which they are conveyed. Standard analyses of those relationships concentrate on the simplest and clearest types of emotionality. The challenge in this thematic area is to develop analyses that extend at least to simple interactive situations. The task has multiple levels. It requires good ways of describing the types of sign that appear in those situations, and the variations in them that are important; it involves finding suitable ways of describing the emotional and emotion-related states that occur in
interactions; and it involves finding ways to relate the two. Good solutions to these problems would serve both synthesis (allowing appropriate combinations and sequences of signs to be generated) and analysis (allowing appropriate interpretation of combinations and sequences of signs).
(c) Emotion in cognition and action
Modern research in various disciplines has emphasised that emotion has pervasive effects on the
mind: it entails distinctive ways of perceiving and assessing situations, processing information,
and prioritising and modulating actions. In effect (though not necessarily explicitly), an emotion-sensitive system needs to embed models of these phenomena. Otherwise, for instance, it cannot “understand” a user’s emotional state well enough to anticipate how s/he is likely to act, with or without particular interventions by the system; nor can it synthesise actions which are consistent with the emotion it sets out to portray. The challenge in this thematic area is to develop ways of embedding appropriate models
of emotionality in systems that use emotion-sensitive interfaces.
(d) Functions of emotion-related elements in communication and persuasion
There are many potential applications of emotion-oriented systems where the goal is not to manipulate or match emotion as such, but to use emotion-related means as a way of achieving pragmatic ends (Stock, 2002) – for instance, ensuring that a person is receptive to information, or motivated to carry out a task. There is a long-established literature on these issues in social psychology (e.g., Lazarus, 1991), and recently the technological community has become increasingly interested in the ways that these effects may be used. For instance, several researches have suggested that the user’s productivity and performance is enhanced by the use of emotionally coloured embodied agents. Users spend more time interacting with an agent with a stern face than one with a neutral expression (Walker, Sproull, & Subramani 1994).
The challenge in this thematic area is to understand the way emotion-related means can be used to enhance systems’ ability to communicate and persuade.
(e) Theories of emotion
There is a well-established psychobiological tradition that aims to discover the essential nature of emotion in humans and (particularly since Darwin) in animals (e.g., Plutchik 1980, Scherer & Ekman, 1984). It is backed by a large body of empirical research. It is clearly important that the insights developed by that tradition should be available to technologies concerned with emotion. The converse may also be true – it is quite likely that attempts to develop emotion-oriented systems may cast useful light on pure theory.
The challenge in this thematic area is to develop a body of understanding that effectively links theoretical work on emotion to technological issues entailed in developing emotionoriented systems.
(f) Usability of emotion-related systems
Any attempt to develop novel systems for human use needs to be supported by work on humans’ reactions to the systems (Höök et al., 2003). Usability traditionally focuses on goals such as effectiveness, efficiency, safety, utility, learnability, and memorability. These objective usability goals contrast with user experience goals, which cover subjective qualities such as being fun, rewarding, motivating, satisfying, enjoyable, and helpful. Usability goals and user experience goals often stand in complex relationships, involving tradeoffs such as safety vs. fun or efficiency vs. enjoyability (Preece et al., 2002). Introducing emotion thus raises many new dimensions for research on usability to address. The challenge in this area is to ensure that research on emotion-sensitive systems can be effectively informed by understanding of users’ responses to them. This does not mean simply passive evaluation: it has profound implications for the design process.
(g) Functions of emotion-related elements in communication and persuasion
The thematic area is related to a tradition of literature on ‘social influence’, which looks (among other things) at ways of using or adapting to emotion. That kind of issue underlies many of the obvious applications – teaching, informing, selling, supporting, entertaining, pre-empting problems. Systems in these areas need to know not only how to express and recognise various emotions and emotion-related states (such as boredom or irritation or humour), but also how to use them or avoid them so that desired goals can be achieved (Pelachaud, Carofiglio, De Carolis, & de Rosis, 2002). From an applied point of view we can say that future intelligent interfaces will have contextual goals to pursue. They may aim at inducing the user – or in general the audience – to have some beliefs and perform some actions in the real world. They will have to take into account the “social environment”, exploit the situational context, and give due weight to emotional aspects in communication.
Some foreseeable scenarios of this kind are: dynamic advertisement, preventive health care, social action, and edutainment. In all these scenarios rational exchange of information is not enough. For people to adopt the desired stance, often what really matters is not only the content but also the overall impact of the communication. Hence there is a strong motivation to provide interfaces with the capability of reasoning on the effectiveness of the message, and not only on the high-level goals and content. Persuasion is a computationally interesting example. It is clearly concerned with elements that transcend argumentation. Inducing emotions is a means for obtaining a given result, either directly or indirectly by devices like the use of specific languages, such threatening or promising, which have emotional overtones. They all can be regarded as resources for inducing the receiver to act in a desired way. Elements that are relevant in the computational study are:
-
The cognitive state of the participants (beliefs and goals of both the user and the interface)
-
Their social relations (social power, shared goals etc.)
-
Their emotional state (both the emotional state of the user and the one expressed by the system)
-
The context in which the interaction takes place.
Persuasive technology is a theme becoming very popular in the Computer Human Interaction milieu.
Another theme is the general attitude of the participants to communication and specifically with aspects of pleasurable communication. Relevant studies include, among others, the case of humour. Specific workshops concerned with Computational Humour have taken place in recent years; the latest one at the big CHI convention in 2003.Recently, there has been growing acceptance that humour is inseparable from an emotional element (e.g., Ekman 1999). Ruch (1997) postulates a specific emotion underlying humour called exhilaration. Humour also has the typical emotional effect of facilitating change in a person’s global
appraisal of the priorities in a situation (Oatley 1992). The joint research in this area will consist of two phases. The first will aim to achieve as broad as possible an overview of the various issues that arise in this area. It will result in recommendations to the plenary meeting after 16 months, and will underpin the workshop at month 24. The second phase will consist of joint work towards specifying a system or systems that would constitute an exemplar for the area, showing how these issues can be addressed in a principled way that links to principled treatments of emotion per se. As suggested above, areas under consideration
include persuasion and humour.
CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE (CNRS)
The research group at CNRS bring to HUMAINE expertise in multimodal communication and in spoken language processing.. Research on multimodal communication includes typologies of cooperation between modalities, creation / annotation / analysis of corpora of multimodal humanhuman and human-computer interactions, design of bi-directional multimodal interfaces featuring cooperation between input modalities (e.g. speech and gestures) and output modalities (e.g. specification of 2D embodied conversational agents including emotional tags), and application to user interfaces for autistic people. Research on spoken language processing includes the creation and annotation of spoken dialogs and the design of spoken interactive systems, nd the question of perceptual detection and prosodic cues analysis of emotional behavior in a spontaneous speech corpus of real Human-Human dialogs. Future plans concern the relations between emotions and spoken /multimodal behaviors, including the following directions: survey of spoken and multimodal cues of
emotional state, creation of multimodal human-human and human-computer corpora including real human-computer interactions / emotion eliciting protocols / theatre / movie samples, specification of mark-up language for the annotation of emotions, use of cues in the multimodal behavior for easing the annotation of emotion in speech, produce knowledge out of multimodal corpora analysis for both the automatic analysis of emotions in user’s behavior, and the synthesis of multimodal emotional behavior in embodied agents. LIMSI-CNRS has been involved in several projects such as IST ISLE (International Standards for Language Engineering: annotation of multimodal behavior), IST NICE (Natural Interaction with Computers for Edutainment: multimodal embodied agents for a game application), IST Amitiés (Automated Multilingual Interaction with Information and Services: creation
of a corpus of real agent-client dialogs at a Stock Exchange Customer Service Center).
http://www.affective-sciences.org/areas-applications-overview
Understanding how individual and contextual factors shape and modulate emotion regulation goals and skills may help educators, managers, parents, and policy makers create more effective socialization practices, laws, and interventions, which in turn should help foster family cohesion, a favorable organizational climate, and community cooperation. Finally, affective phenomena play useful and interesting roles in areas of art, music, and theater and are used by advertisers to influence consumers and sell their products.
[10] En dehors de la communauté Européenne (exemples non exhaustifs)
UNIVERSITY OF HAIFA (HU)
The Center for Interdisciplinary Research of Emotions, founded in 1998, aims to further systematic, integrative, and in-depth understanding of feelings. Research will synthesize work in a wide array of disciplines: philosophical and literary analysis, anthropological observation, psychological and psychiatric research and experimentation, biological investigation and physiological assessment, historical inquiry, and artistic investigation. Behavioral and humanistic research will complement each other in the effort to illuminate the causes, determinants, and consequences of emotional expression and behavior in a variety of social and cultural contexts. The research group attached to HUMAINE focuses on two topics - an integrative framework for describing emotions; and emotions on the Internet. The group aims to investigate whether an emotion can be considered as a general mode (or style) of the mental system. The integrative framework will then be applied to describing emotions in cyberspace. The present project focuses upon a few central emotions that occur in cyberspace, and in particular romantic love and sexual desire. The nature of these emotions in cyberspace will be examined and compared to their counterparts in offline circumstances.
TEL AVIV UNIVERSITY (TAU)
The group is centered at the Department for Communication Disorders, Tel Aviv University. Members are engaged in a number of ongoing research subjects related to the expression of emotion in speech and its manifestation in physiological measurements. The group also have particular expertise in data collection. They have put together a number of corpora of emotional speech, including a corpus recorded during event recollection, a corpus of anger in naturally occurring Hebrew speech and a corpus of sentence utterances recorded during a computer game that includes various degrees of risk, with the final objective of gaining a sum of money. In the latter, speech was recorded simultaneously with a number of physiological parameters. Acoustic analysis is being carried out.
