MODELING SELF-CONFIDENCE IN USERS OF A COMPUTER-BASED SYSTEM SHOWING UNREPRESENTATIVE DESIGN

While a great deal of research has demonstrated that users' self-effic acy beliefs have a major impact upon both their attitudes to technolog y and their performance, a related construct, self-confidence, has bee n largely ignored within the domain of human-computer interaction. Thi s is surprising given the vast literature on the calibration of confid ence which can be found within the judgment and decision literature. I n this study, 60 participants were asked to complete a novel computer- based task, and to provide measures of self-confidence in terms of the ir anticipated performance at each stage of the task. Users' confidenc e judgements showed sensitivity to their rate of improvement on the ta sk, but were poorly calibrated with actual performance at each stage. Furthermore, confidence judgements were insensitive to the complexity of the individual task components, even though these different compone nts led to very different levels of performance. The style of computer interface was also found to affect anticipated performance independen tly of actual performance. The ability of existing models of confidenc e judgement to deal with these data is discussed. (C) 1998 Academic Pr ess.