Michael Matessa and John Anderson
This paper describes a preliminary attempt to incorporate theories of communication (Clark and Schaefer, 1989; Poesio and Traum, 1998) into a general cognitive architecture (Anderson and Lebiere, 1998) as a first step in creating an interactive problem solving system. A restricted interface for a communicative task is introduced that facilitates system interaction with subjects, and an empirical study of paired subjects is presented that compares the restricted interface to an unrestricted interface. No difference is found between the interfaces by measures of time to complete problems, turns to complete problems, and final scores. However, some subjects using the unrestricted interface present information to their partner but do not wait for evidence of understanding before performing important actions. Implications of these findings and future modeling directions are discussed.