Effects of information source, pedigree and reliability on operators' utilization of decision support systems

Two experiments are described that examine operators' trust in and utilization of decision support systems. In Study 1 we examined operators' preconceived trust in and perceived reliability of different sources of decision support (human adviser vs. automated aid) that varied in pedigree (expert vs. novice), prior to operators' interacting with these advisers. Results revealed that automated advisers were perceived as more reliable than human advisers across pedigrees. However, subjective trust in advisers varied as a function of pedigree. Automated 'novices' were trusted more than human 'novices'; but human 'experts' were trusted more than automated 'experts'. Participants used different criteria to judge human and automated advisers. Humans were judged by dispositional features such as "years of experience"; automated aids were assessed by situational factors such as their "current use".; In Study 2 we examined participants' utilization of and post hoc trust in the above sources of advice and pedigree levels after their objective experience of advisers' diagnostic accuracy (high vs. low) in an airline luggage-screening task. Biases toward automation (observed in Study 1) led to greater dependence on 'novice' automated aids than on 'novice' human advisers. However, biases in favor of automation broke down when automated aids believed to be 'experts' generated errors, leading to a significant drop in compliance/reliance on 'expert' automation relative to 'expert' humans. Contrary to the results of Study 1, perceived reliability of advice was significantly influenced by the interaction of source and pedigree, while subjective trust was always higher in automaton than in humans. The results have important implications for the development of theoretical models of operator interaction with decision aids, while providing a reference point for deriving optimal indices of user dependence on decision support systems.