Symposium "Computational modeling of emotion: Issues and alternatives"

 
Discussant:
Peter Goldie (http://www.kcl.ac.uk/kis/schools/hums/philosophy/staff/p_goldie.html)

Presenters:
Lola Cañamero (http://homepages.feis.herts.ac.uk/~comqlc/)
Jonathan Gratch/Stacy Marsella (http://emotions.usc.edu)
Paolo Petta (http://www.oefai.at/~paolo/)
Klaus Scherer/Etienne Roesch (http://www.unige.ch/fapse/emotion/)
John Taylor (http://www.mth.kcl.ac.uk/~jgtaylor/)

The purpose of this symposium was to discuss the different approaches of the HUMAINE teams interested in elaborating computational models of emotion elicitation and differentiation.

The discussions were structured as follows: After a brief introduction to the issues and questions raised by computational modeling of emotions, the speakers were invited to comment on their own model in the context of one or several of the following questions. The explicit goal of this exercise was to allow an in-depth discussion regarding what constitutes central issues and choices in computational modeling of emotions, and increase the knowledge that can contribute to our understanding of emotions.

1. What is the input that your model uses?
How does your model deal with the input from the environment? How does the processing of the input elicit and differenciate the emotional states modeled? What is the basis for the differentiation and what are the functional considerations involved?

2. What are the internal dynamics in your model?
What are the mechanisms used for input transformation and integration? What functions are involved (e.g., nonlinear dynamics, hysteresis)? What is the architecture used to model dynamic and recursive processes (e.g., rule-bases, excitation spreading)?

3. What are the output variables provided by your model?
How can output be measured and compared to the results published by emotion researchers? How does it account for the current views of emotional processes (e.g., qualia, discrete emotion categories, dimensional perspectives)? What insight does it bring to the current understanding of action tendencies as an essential part of emotional processes?

4. What are the emergent properties of your model?
What are the potential behaviors that have not been explicitly programmed into algorithms? For instance, how does your model deal with unexpected changes in the environment? Does the model learn? How does it form schemata based on prior exposure? Does it associate earlier response patterns to such schemata?

5. What were the primary design goals that influenced your model's development?
Is your model intended to resolve theoretical distinctions between alternative theories of human emotion? Is it intended to faithfully model human mental processes? Or, is it intended to enhance the "naturalness" or "believability" of virtual characters, to model visible manifestations of emotions?

Speakers' descriptions of their own model with respect to these issues, were followed by a more general discussion among the symposium speakers and with the public at large. In order to avoid having to spend precious discussion time on the description of published models, this page provided the audience with pdf-copies of a single paper that best describes the models.

If you are having troubles downloading any of the following, please contact Etienne.Roesch@pse.unige.ch.

Speakers' articles

Avila-Garcia, O., & Cañamero, L. (2005). Hormonal modulation of perception in motivation-based action selection architectures. Paper presented at the AISB'05 Symposium "Agents that Want and Like: motivational and emotional roots of cognition and action", Univ. Hertfordshire, UK. [pdf]

Gratch, J., and Marsella, S. (2004). A Domain-independent framework for modeling emotion. Journal of Cognitive Systems Research, 5(4), 269-306. [pdf]

Rank, S., & Petta, P. (submitted). Appraisal for a character-based story-world. Paper submitted to the 5th Intl. working conference on Intelligent Virtual Agent 2005 (IVA'05), Kos, Greece. [pdf, slides]

Scherer, K. R. (2001). Appraisal considered as a process of multilevel sequential checking. In K. R. Scherer, A. Schorr & T. Johnstone (Eds.), Appraisal processes in emotion: Theory, methods, research. Oxford, New-York: Oxford University Press. [pdf]

Taylor, J. G., & Fragopanagos, N. (2005). Modelling the interaction of attention and emotion. Paper will be presented at the International Joint Conference on Neural Networks 2005 (IJCNN'05), Montreal. [pdf]

Speakers' presentations

Outline of the symposium: Goldie, P. (2005, May 24-27). Computational modeling of emotion: Issues and alternatives. Paper presented at the HUMAINE Plenary Session 2005, New Castle, Ireland. [pdf]

Cañamero, L. (2005, May 24-27). An embodied model of emotion. Paper presented at the HUMAINE Plenary Session 2005, New Castle, Ireland. [pdf][avi]

Fragopanagos, N. (2005, May 24-27). Computational modelling of attention and emotion. Paper presented at the HUMAINE Plenary Session 2005, New Castle, Ireland. [pdf]

Gratch, J., and Marsella, S. (2005, May 24-27). EMA: A Computational Model of the Appraisal Process. Paper presented at the HUMAINE Plenary Session 2005, New Castle, Ireland. [pdf][avi]

Petta, P. (2005, May 24-27). HUMAINE 2nd Plenary, WP3 Symposium - WG2.2: Tabasco. Paper presented at the HUMAINE Plenary Session 2005, New Castle, Ireland. [pdf][avi]

Scherer, K. R. (2005, May 24-27). Computational models of emotion. Paper presented at the HUMAINE Plenary Session 2005, New Castle, Ireland. [pdf]