Investigation of voice and text output modes with abstraction in a computer interface

A human-computer interface is described, which was designed to study user preferences and the effectiveness of output modes and levels of information abstraction in a decision making environment. The interface was tested in an exploratory study of an apartment selection problem. It was observed that text plus voice was preferred over voice alone, but there was no significant difference in preferences between text and voice or between text and text plus voice. This indicates that adding text to voice output improves the perceived acceptability of voice, but adding voice to text does not alter the perceived acceptability of text. The text mode was most efficient in performing information search, followed by voice mode and text plus voice mode in that order. We observed inconsistencies between the users' perceived importance of information attributes and the actual usage of these attributes, and inconsistencies between the perceived importance of and the usage of abstraction levels. We did not observe significant differences between users with task domain experience and those which did not have domain experience, but cognitive style did affect task performance. Our findings suggest that a user interface should either provide flexible access at different abstraction levels, or should organize information based on its perceived importance to the user rather than its level of abstraction.