abstract
-
This paper describes a human-computer interface which was designed to study the effectiveness of voice and text output modes, along with information abstraction, in decision making. The interface was tested in an exploratory study by population samples of both domain experts and non-experts in an apartment choice problem. We observed that there was no difference in user preferences for both user communities between text and voice or between text and text plus voice. However, text plus voice was preferred over voice alone. The speed of interaction with a pure voice interface was slower than either of the others. The choice of output mode, although it affected time efficiency, had no effect on the decisions made. We also observed that domain experts used higher level information abstraction more than detailed information, in contrast with non-experts, and heuristic experts were faster than analytic experts in completing decision tasks. An entropy calculation showed that domain experts were also significantly less random in their choice of information attributes. Our findings indicate that text or a combination of voice plus text can be effective output modes, and that information abstracted at several levels can be useful for certain types of users.