Emotional and Intelligent:
The Tangled Knot of Cognition

http://www.iiia.csic.es/~lola/ei-fs98.html

American Association for Artificial Intelligence (AAAI)

1998 Fall Symposium Series

Omni Rosen Hotel, Orlando, Florida, October 23-25, 1998

The question is not whether intelligent machines can have emotions, but whether machines can be intelligent without any emotions.
Minsky, The Society of Mind

Visit the Emotion Forum page for information about research on emotion and related events.

Research in neurobiology has provided evidence that emotions pervade human intelligence at many levels, being inseparable from cognition. Perception, attention, memory, learning, decision making, social interaction or communication are some of the aspects influenced by emotions. Their role in adaptation has likewise been evidenced by these studies. In the AI community, the need to overcome the traditional view that opposes rational cognition to absurd emotion has also been acknowledged. Emotion is not regarded anymore as an undesirable consequence of our embodiment that must be neglected, but as a necessary component of intelligent behavior that offers a rich potential for the design of artificial systems, and for enhancing our interactions with them.

This symposium investigates the role of emotions in grounding intelligent behavior, both at the individual and social levels. The main focus is on artificial agents in all sorts of embodiments, and on the possibilities for cross-fertilization between research in artificial emotions and studies of emotions in animals and humans.

Submissions are welcome which discuss theories, architectures, implementations, or problems regarding the following issues (this is a non-exclusive list), in artificial or in biological systems:

Models, architectures, taxonomies
Origins and status of emotional phenomena
Emotions versus other affective phenomena (motivation, moods, temperaments, etc.)
Emotions in biological and artificial systems:
- the biology of emotions
- biologically-inspired models of artificial emotions
- artificial emotions generated through artificial evolution
- what insight can we gain from research in animal and human emotion? What can our models and implementations contribute to them?
- should we follow nature for emotion synthesis, or should we let our agents evolve their own emotions?
Embodiment aspects of emotion:
- how do emotional phenomena arise from/relate to the body?
- does any kind of embodiment seem to have a privileged status for the implementation of emotions in artificial agents? For what aspects?
Emotion expression, synthesis and recognition:
- mechanisms for emotional expression, perception, activation, etc.
- how can emotional states be "measured"?
Emotions and behavior:
- adaptive value of emotional states
- emotions as a mechanism for behavior control
- roles of emotions in social interaction
Emotions in cognition and learning:
- influence of emotions in other aspects of (lower- and higher-level) cognition
- emotional learning
- learning of emotions
- (how) does the lack of emotions impair the cognitive abilities of artificial autonomous agents?
Emotions in social interaction and communication:
- what do emotions communicate to others?
- modulation of communication by emotions
- expression and description of emotions in natural and artificial languages
- emotional expression and art
Design and implementation issues:
- what architectures are better suited for emotion synthesis?
- what kind of computational and representational mechanisms are needed to capture the complexity of emotional phenomena?
Philosophy:
- emotions in the philosophical tradition
- emotions and self-image
- epistemological issues
- ethical aspects
Applications: art, autonomous and believable agents, education, entertainment, interfaces, medicine, multi-agent systems, pets and personal robots, wearable computing, etc.

Contributions from fields others than AI, ALife, and robotics (e.g., arts, biology, humanities, social sciences), are also strongly encouraged.

Interaction among participants will be fostered. Discussion groups will be formed before and during the symposium. Presentations will be short and organized around panels. Poster sessions will allow for more detailed and technical discussions. All contributors are invited to bring a poster presenting their work.

Organizing Committee

Dolores Cañamero (Chair), IIIA-CSIC, Spain
Kerstin Dautenhahn, University of Reading, UK
Hiroaki Kitano, Sony CSL, Japan
Ramón López de Mántaras, IIIA-CSIC, Spain
Chisato Numaoka, Sony CSL, France
Marvin Minsky, MIT, USA
Andrew Ortony, Northwestern University, USA
Rosalind Picard, MIT, USA
Takanori Shibata, MEL, Japan
Aaron Sloman, University of Birmingham, UK

Contact person:

Dolores Cañamero
Spanish Scientific Research Council (IIIA)
Artificial Intelligence Institute (CSIC)
Campus de la U.A.B.
E-08193 Bellaterra, Barcelona, Spain
Phone: +34- 93-580-9570
Fax: +34-93-580-9661
E-mail: lola@iiia.csic.es

The Call for Papers is available here.

Symposium Schedule:

Friday, October 23:

09:00 - Welcome
09:05 - Invited talk : Affective Computing: Research Progress and Grand Challenges, Rosalind W. Picard
10:00 - Session 1: Emotion, Cognition, and Learning I (Chair: Paolo Petta)

Wisely Non-Rational---A Categorical View of Emotional Cognitive Artificial Perceptions (10:00 - 10:15), Zippora Arzi-Gonczarowski
To Believe and To Feel: The Case of Needs (10:15 - 10:30), Cristiano Castelfranchi
10:30 - 11:00 Coffee break
11:00 - Session 1: Emotion, Cognition, and Learning II

Exploring the Role of Emotions in Autonomous Robot Learning (11:00 - 11:15), Sandra Gadanho and John Hallam
Learning and Emotional Intelligence in Agents (11:15 - 11:30), Magy Seif El-Nasr, Thomas Ioerger and John Yen
Emotions as Weak Biases in Metaphor Formation (11:30 - 11:45), William Stubblefield
Emotion and Learning: Solving Delayed Reinforcement Learning Problem Using Emotionally Reinforced Connectionist Network (11:45 - 11:55), Stevo Bozinovski and Liljana Bozinovska
An Algorithm Based on an Appraisal Theory to Control a Khepera Robot (11:55 - 12:05), Gérald Foliot
Discussion
12:30 - 2:00 Lunch
2:00 - Session 2: Architectures (Chair: Dolores Cañamero)

Modeling Emotion within Symbolic Cognitive Architectures (2:00 - 2:15), Eva Hudlicka
Emotion Synthesis: Some Research Directions (2:15 - 2:30), Christine Lisetti
An Architecture for Emotion (2:30 - 2:45), Lee McCauley and Stan Franklin
Modeling Emotion-Based Decision-Making (2:45 - 3:00), Juan Velásquez
Emotions---The Missing Link? (3:00 - 3:15), Rodrigo Ventura, Luís Custodio, and Carlos Pinto-Ferreira
Discussion
3:30 - 4:00 Coffee break
4:00 - Session 3: Emotion in Social Interaction (Chair: Andrew Ortony)

A Commitment Theory of Emotions (4:00 - 4:15), Michel Aubé
Early Experiments Using Motivations to Regulate Human-Robot Interaction (4:15 - 4:30), Cynthia Breazeal (Ferrell)
Emotionally Intelligent Agents: The Outline of a Resource-Oriented Approach (4:30 - 4:45), Alastair Burt
Personality Development through Interactions in Virtual Worlds (4:45 - 5:00), Chisato Numaoka
A Computational Model of Trust and Rumor (5:00 - 5:05), Michael Prietula and Kathleen Carley
Modeling Strategic Emotions (5:05 - 5:10), William Turkel
Discussion
5:30 - End of Sessions and Evening Opening Reception

Saturday, October 24:

09:00 - Session 4: Emotion Expression and Recognition I (Chair: Chisato Numaoka)

Bayesian Networks for Modeling Emotional State and Personality: Progress Report (9:00 - 9:15), Jack Breese and Gene Ball
From Motion to Emotion: Synthesis of Interactivity with Gestural Primitives (9:15 - 9:30), Insook Choi
Dominance and Valence: A Two-Factor Model for Emotion in HCI (9:30 - 9:45), Christopher Dryer
Expressing Emotions: Using Symbolic and Parametric Gestures in Interactive Systems (9:45 - 10:00), Paul Modler
How Well Can People and Computers Recognize Emotions in Speech? (10:00 - 10:15), Valery Petrushin
Affective Pattern Classification (10:15 - 10:30), Elias Vyzas and Rosalind Picard
10:30 - 11:00 Coffee break
11:00 - Session 4: Emotion Expression and Recognition II

Discussion
11:20 - Session 5: Psychological Models of Emotion (Chair: Hartvig Dahl)

Cognition-Emotion in the Self-Organization of Personality and Behavior (11:20 - 11:35), Marc Lewis
Psychoanalytic Concepts for the Control of Emotion in Pet-Like Robots (11:35 - 11:50), Stephane Zrehen
Feedback Loops in Expression and Experience: Emotion as Cause and Effect (11:50 - 12:00), Marc Coulson and Simon Duff
Discussion
12:30 - 2:00 Lunch
2:00 - Session 6: Emotion in Arts and Entertainment (Chair: William Stubblefield)

A Talking Head Architecture for Entertainment and Experimentation (2:00 - 2:15), Kim Binsted
Towards New Lyrical Forms (2:15 - 2:30), Alain Bonardi and Francis Rousseaux
Modelling the Dynamics of Musical Engagement (2:30 - 2:45), Jonathan Impett
Affect-Driven Generation of Expressive Musical Performances (2:45 - 3:00), Josep-Lluís Arcos, Dolores Cañamero, and Ramon López de Mántaras
AI and Musical Emotions (3:00 - 3:05), Mladen Milicevic
Wolfgang: "Emotions" plus Goals Enable Learning (3:05 - 3:10), Doug Riecken
Discussion
3:30 - 4:00 Coffee break
4:00 - 4:45 Poster session and demos
4:45 - 5:30 Working groups
5:30 - End of Sessions and Evening Plenary Session

Sunday, October 25:

09:00 - Session 7: Theoretical Issues (Chair: Andrew Ortony)

A Formal Interpretation of the Concept of Emotion (9:00 - 9:15), Jonathan Liu
What AI Needs is a Theory of the Functions of Emotions (9:15 - 9:30), Hartvig Dahl and Virginia Teller
Emotional Thought or Thoughtful Emotions? (9:30 - 9:40), Valerie Hardcastle
Issues in the Design of Emotional Agents (9:40 - 9:50), Dolores Cañamero
Discussion
10:30 - 11:00 Coffee break
11:00 - Summary of groups and general discussion
12:30 - End of Symposium Series

Short Report for AI Magazine

(A longer report forthcoming soon)

This multidisciplinary symposium brought together researchers in artificial and human emotions, in an effort to investigate the roles that emotions play in grounding intelligent behavior, and the possibilities for cross-fertilization between theoretical and engineering approaches to the study of emotion.

Presentations were organized around seven main topics: emotion, cognition, and learning; architectures for emotion; emotion in social interaction; emotion expression and recognition; psychological models of emotion; emotion in arts and entertainment; and theoretical issues. Lively discussions took place at the end of each session, and were furthered in small working groups meetings.

The last session was devoted to identifying open problems and issues to be taken into account when designing artificial systems endowed with (some aspects of) emotions. Consensus was reached on several ideas, such as:

The degree of complexity of emotional phenomena that is meaningful to model within an AI system strongly depends on whether our concern is an engineering or a scientific one.

Designers of artificial systems cannot assume the adaptive character of emotions only on the grounds of biological evidence. The contributions of emotions must be shown by our results and carefully evaluated.

Design must be guided by the requirements on the system and on the interactions with its particular environment, and our architectures related with the desired functionalities.

The final ``moral'' could be summarized as: when engineering artifacts, we should not put in them more emotion than what is required by the complexity of the system-environment interaction.

Dolores Canamero
Spanish Scientific Research Council (IIIA-CSIC)

Last modified: December 11, 1998

Go to Dolores Cañamero's Home Page

Emotional and Intelligent: The Tangled Knot of Cognition