Carl Westin

216 Literature review of human-automation trust T ABLE D-1: Antecedents of human-automation trust Antecedent of trust Finding Trust before use People often indicate higher liking in human sources than in automated sources. 49,177 People often indicate higher trust, higher reliability, greater expertise, higher authority, and higher performance expecta- tions of automated sources as opposed to human sources (e.g., automation bias and perfect automation schema). 63,165,166 Much trust in external advice may be attributed to a heuristic of relying in order to reduce mental effort. 63,173 Trust in humans and automated sources varies with pedigree: Operators may trust novice automation more than novice human, while trust in expert humans is greater than trust in expert automation. 167 Trust during use Determinants for intentions to use often do not translate well to determinants of actual use. 167,174 Trust in automated aids is primarily influenced by its performance. 171 Trust in human aids continues to be influenced by the dispositional attitudes such as knowledge, effort and expertise. 171 People’s confidence in a source does not determine agreement with advisories from that source. 171,185 Automated sources are judged more negatively if they fail than are human sources, 49,50,167,174 especially if the task appears easy to the operator. 294 People are more forgiving of errors made by a human source than errors made by an automated source. 49,165 Reliance on an automated aid is not affected by advisory timing (i.e., before or after the operator has made a decision). 50 Experts are more likely to critically consider arguments, while novices are more likely to rely on external clues when given advice. 173 Influence of perceptions The expertise of an automated aid is mainly associated with knowledge, while expertise of a human aid more broadly is associated with knowledge, experience, and education. 171 Advice from an automated source perceived more rational and objective than the same advice from a human source. 172 Trust in other humans may be biased by a worry of participants to admit distrust. This does not apply to automaton. 185 Human sources are generally perceived as imperfect and have less credibility, which negatively affects trust. 55,161 Motivation to perform well is critical to appropriate trust. 173 Framing effects Anthropomorphic features (e.g., automation’s appearance and behavior) can increase trust and reliance. 48,114,118,176 Reliance is driven more by how credible a source is framed, than by consideration of the actual problem at hand. 173,175 Attitudes of trust, confidence, and liking are sensitive to framing. 48,110,171