Abstract or Description
When people communicate they try to establish mutual knowledge. Garrod and Anderson (1987) proposed that a way to minimize effort during this process would be to follow a “output/input coordination” principle, where output to a partner is formulated according to the same principles of interpretation as those needed to interpret input from a partner. A computational model of establishing mutual knowledge efficiently can be given in the ACT-R architecture (Anderson & Lebiere, 1998) where goals that are completed successfully can be retrieved and used later. Applied to communication, goals of presenting and accepting information include semantic and syntactic representations of that information, and these goals can later be retrieved to provide templates for the creation of new utterances. Results from an ACT-R model communicating with human subjects show similar performance to that of human subjects communicating together.