AAAI 2004 Fall Symposium Series



Held on October 21-24, 2004, in Washington, D.C




Original Call for Papers

Download Papers

Lecture Slides – Coming Soon

View Pictures


Summary Report


This highly multidisciplinary symposium brought together researchers working with computational models for ‘style’ in a great variety of domains, including the visual arts, music, natural language, theater and cinema, game playing, and architecture. During very stimulating meetings we explored many different perspectives on style and its role in various facets of human behavior, attempting to create an integrated research community from our heretofore-disparate research enterprises. 


There was broad agreement that style is generally expressed by a confluence of a great many small ‘surface’ details in any particular work (painting, concerto, essay), which collectively point towards a particular coherent interpretation of a feeling or identity.  This ‘distributed’ character of stylistic expression was commented upon by several speakers, and was contrasted with the ‘compositionality’ of classical denotational accounts of meaning.  But style is far more than just a collection of features, and a deeper understanding must also take into account the history and social context of a work.  That is, production constraints, whether externally or internally imposed, are key to understanding how features may cohere into a recognizable style.  Also relevant was the observation of how stylistic patterns may be fundamentally altered by a perceiver’s changing point of view, underscoring the essentially context-sensitive nature of stylistic apperception.


At a high level, three main perspectives as to style’s place in cognitive processing were discussed.  One view of style had it as emerging from an interaction between production constraints and a coherent (though possibly initially vague) ‘artistic vision’.  An opposed view of style sees it as a realization of explicit affective and interpersonal goals, via many small features distributed in a work.  The third notion of style was as a consistent strategic stance enabling decision-making when various alternatives seem equally ‘good’ from a utility-theoretic perspective.  More specific approaches to analyzing stylistic features of works in various media were discussed, including the role of metaphor in stylistic expression, of anticipation and familiarity in emotional response, of ‘rules of style’ in construction of a ‘stylistic ideal, and of functional and rhetorical organization in constructing a style for a community of discourse.


It was clear that whatever it really is, style is an essential part of human intelligence, and that a computational understanding of style becomes more important as our lives become intertwined with ubiquitous information load and ever more capable information technologies.  In terms of applications, we all agreed that style research will be increasingly central to ‘building bridges of understanding’ between people and intelligent systems in areas such as machine translation, user-adaptive interfaces, or intelligent agents, to mention a few.  Beyond applications, though, an understanding of style, from the mundane (peoples walking styles, e.g.) to the sublime (Monet’s lilies, or Beethoven’s symphonies), is fundamental to understanding what it means to be human. A new community is coalescing to study this age-old problem, informed by the diverse traditions of artificial intelligence, cognitive science, and psychology, together with humanities scholarship, artistic expression, and philosophical inquiry.



Shlomo Argamon, Illinois Institute of Technology

Shlomo Dubnov, University of California at San Diego