Preventing knowledge transfer errors: Probabilistic decision support systems through the users' eyes

Development and use of probabilistic decision support systems benefit by good communication between the developer on the one hand, and the user and the domain expert on the other hand. Communication is difficult because large differences in training and experience, and thus also expectations, exist between the two. This causes knowledge transfer errors. Preventing these necessitates user-centered design of the representations used in this communication, and attention to the translation of user terms to model terms and vice versa. A systematic approach to developing user-centered representations and preventing knowledge transfer errors is outlined in this paper. We demonstrate how five heuristic guidelines can be fruitfully applied in different developer-user interaction situations in different phases of decision-support system construction, and briefly outline an evaluation of a data-entry system developed using these guidelines.