AAAI Fall 1999 Symposium on

Psychological Models of Communication
in Collaborative Systems

November 5-7th, 1999,
Sea Crest Conference Center on Cape Cod
North Falmouth, Massachusetts, USA

Recently, the fields of artificial intelligence, human-computer interaction, computer-supported cooperative work, and computational linguistics have spilled much ink on the topic of collaborative systems. Collaborative systems act either as partners to human users or as mediators between communicating users. The inventors of collaborative systems, being human, all have intuitions about what collaboration is and how it should be supported, and so many collaborative systems embody their intuitions. However, there has been an increasing focus on data -- from looking at corpora of language use or behaviors during collaboration or people using systems. This link between data and systems is bi-directional: the data inform the design of systems, while the systems yield more data that highlight basic findings (about cognition, interaction, or the human use of technology) or applied findings (like whether the systems are easy to use or whether they do what they're supposed to do). The third part of the triangle is theory. In the field of psychology, researchers generate theories to explain interesting observations, and then they generate more observations to test these theories - another bi-directional relationship. But what is/has been/should be the relationship between psychological theory and collaborative systems? That was the inspiration for this symposium.

The symposium was attended by a diverse international group in which social scientists were well-represented; about 40% of the 50 registered participants were card-carrying psychologists (by way of their training or departmental membership). Overwhelmingly, it was an interdisciplinary group: When asked to characterize themselves as building things or studying things, many characterized themselves as doing both. A substantial number of the presentations focused on H. H. Clark's grounding theory and related ideas. These included not only attempts to formally model or implement the parameters of grounding, but also experiments looking at communication in pairs or larger groups, through different media, using different tasks, among participants in diverse roles, etc. Clark himself argued that communication with virtual partners is another form of disembodied language use, not unlike what happens whenever people read text or interpret language at a time when it is not being produced by an actual speaker. Finally, several systems were described and demos were presented of people interacting with and through collaborative systems, along with discussion of the psychological assumptions and implications of these systems.

The symposium focused mainly on a single, comprehensive theoretical framework, Clark's grounding model. Other interesting approaches were sampled briefly (e.g., gesture and language as an integrated system, speech acts, ACT-R as a modeling framework, the impact of cue phrases on comprehension, approaches to tutoring and pedagogy, and general contributions from the field of conversation analysis).