טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
Ph.D Thesis
Ph.D StudentGavish Nirit
SubjectThe Influence of Descriptive Information and Experience
on Trust in Decision Support Systems
DepartmentDepartment of Industrial Engineering and Management
Supervisors Professor Emeritus Daniel Gopher
Mr. David Sinreich (Deceased)
Full Thesis text - in Hebrew Full thesis text - Hebrew Version


Abstract

Unlike real time, dynamic decision support systems such as path finding and navigation aides in aviation systems, users of decision support systems for the long term planning and policy making receive only very delayed feedback on the value and success of system's recommendations. This lag creates a basic acceptance and "trust" problem in the use of this type of decision support systems. In order to build users’ trust, designers are required to provide them with information on the anticipated performance of the system. Most commonly, such information is given in a descriptive format using graphs, tables, statistical parameters, etc. Research demonstrated that users have low trust in systems recommendations, which underestimates the actual predictive power of the developed aid. The present research investigated the value of simulated interactive experience on the development of better matched trust levels. During a simulated training session, subjects were given the opportunity to observe and experience the effect of their given trust levels on simulated system performance.  The effects of simulated experience were tested when accompanied with general descriptive information or without it. When provided, descriptive information could precede or follow experience. Decision making literature has shown that description based decisions tend to overestimate extreme probabilities, while experience based decisions tend to underestimate them. Assuming that acceptance and trust levels of decision support systems are lower than optimal because of overweighting of lower probability unfavorable consequences, experience may elevate trust levels (change them in the right direction).  

The study examined this hypothesis using two simulated decision support systems, one high and the other low positive expected values. In 4 experiments subjects experienced only one or both systems, with or without overall descriptive information.  Results showed that when trust was solely based on viewing descriptive information, it was relatively low. With experience, trust levels increased significantly, and were much closer to the optimal levels. Heterogeneous experience helped to maintain trust over changes in the absolute levels of positive expected values. Presentation of preliminary descriptive information had no effect on the trust levels obtained after experience. However, past experience did influence trust levels based on the following presented descriptive information.

The results of the present study support the introduction of simulated interactive experience to improve acceptance and trust in delayed feedback decision support systems. In addition, it shows the value of heterogonous experience in developing a better distinction between trust policy and overall expected values levels.