An Overview of the ICE Lab/Studio

Our research and development practices centralize the following three concerns:

1) Computational Expression and Narrative
We believe that there is no holy grail of computational narrative, there will be a wide range of new narrative forms enabled by the affordances of computational media. Given this belief, we take an experimental approach, beginning with looking at how narrative imagining occurs at the most fundamental cognitive level. This approach provides semantic building blocks such as formal representations of concepts and algorithms describing how they can be composed. Such building blocks can be used to dynamically annotate, reconfigure, and generate media such as computer graphics in meaningful response to user input. We construct ontologies to support generation of concepts and metaphor via algorithms for blending, metaphor, and analogy. We define algorithms to compose narrative structures to arrive at new forms. We augment procedural computer graphics with semantic meta-data tags and compose media on the basis of composing these tags. The core idea is that the computational representations are inseparable from the expressive content. Traditional computer science has not considered the affective results or subjective interpretation of humanistic production as central questions, yet in order to make progress in developing truly expressive new narrative forms we must see the technical and expressive questions as intertwined. Such technology could be applied for many uses ranging from allowing people to narratively organize their digital photo albums, to narratively structuring mathematics learning curricula in educational software.

2) Computation and Social Identity
Current user representations in digital media are inadequate for capturing complex phenomena involving subjective experience of social identity. Current character creation tools allow for user representations to be customized on the basis of attributes associated with models of race, class, profession, and similar classifications, along with physical choice and construction of character models, skin tone, gender characteristics, and the like. Many popular current computer games duplicate and amplify many disempowering existing social structures. Such games hardcode stereotypes into their infrastructures. They reduce social constructs such as race to sets of numerical variables and cosmetic changes to avatar appearance. Tools for representing identity through more distributed modes such as text and image based profiles, participation in message boards, and knowledge sharing are rarely, if ever, used within gameplay. We seek to create tools that allow for some aspects of such identity constructions, but associated with these choices will be rule sets and processes that formally describe identity concepts. Socially realized identity effects will dynamically change based upon user interaction and discourse. Categories and classifications are constructed and reconfigured on the fly, based upon user interaction. Users selection of physical and artifactual representations will contain semantic information to be exploited within particular systems and will be co-created by participation in application specific activities such as dialogue with other characters or annotating profiles of self and others.

3) Computation and Culture
Deep understanding of the wide range of human cultural stories is necessary for true communication. We believe that diverse cultural practices can present new underpinnings for computational practices. As just one example, interactive narrative performance in a computer game can be informed by the construction of performance space and collective improvisation in African Disaporic narrative, just as artificial intelligence has been shown to be deeply informed by the philosophy (even theology) of thinkers like Descartes. The key is to make our cultural foundations explicit, and to use them as fertile resources for technical innovation. (Harrell, 2007) Developing formalizations for results in cognitive semantics research provides an empirical approach for examining and explicating such cultural forms in a manner amenable to computational implementation. Cognitive linguistics theories are concerned with the operational uniformity of cognitive processes across cultures and at different grain sizes of cognitive process (e.g. both every day commonsense reasoning and higher level literary narrative construction). Such theories emphasize identifying shared underlying cognitive principles that ground our approach to narrative world construction.

D. Fox Harrell, Director

Associate Professor of Digital Media

CMS | CSAIL | MIT