The effects of human-computer communication mode, task complexity, and desire for control on performance and discourse organization in an adaptive task

The present study examined how different communication patterns affected task performance with an adaptive interface. A Wizard-of-Oz simulation (Gould, Conti, & Hovanyecz, 1983) was used to create the impression of a talking and listening computer that acted as a teammate to help participants interact with a computer application.; Four levels of communication mode were used which differed in the level of restriction placed on human-computer communication. In addition, participants completed two sets of tasks (simple and complex). Further, a personality trait, Desire for Control (DC), was measured and participants were split into high and low groups for analysis. Dependent measures included number of tasks completed in a given time period as well as subjective ratings of the interaction. In addition, participants' utterances were assessed for verbosity, disfluencies, and indices of common ground.; The largest performance differences were found between the groups that could communicate freely and those where communication was restricted or denied. As the level of restriction increased, performance decreased. Further, as communication restriction increased, the computer assumed greater control and levels of verbosity decreased. Performance on the simple tasks declined as communication restriction increased, but no differences were observed among communication modes for complex tasks. There were no performance effects due to DC, however high-DC participants rated their ability to communicate as easier than low-DC participants. The results of the present study are discussed with respect to differences between human-human and human-computer communication as well as research on adaptive environments.