Carl Westin

80 Source bias effects Results mirror those of Wærn and Ramberg’s study. 164 In their study, partici- pants received advice to a pattern recognition problem from an automated aid on a computer screen or from a human source over the phone. Trust ratings collected af- ter each source condition showed that participants’ level of trust was higher for the human source (although the modality effect presents a possible confound). In con- trast, trust and self-confidence collected in association to each problem experienced did not yield any significant differences in trust or self-confidence between the two sources. In conclusion, different perceptions of trust that people have in automated and human sources may not carry over to the situation specific measures (e.g., trust, acceptance of agreement). Rather, a rating obtained during a task may reflect a more subliminal derivative measure tied to the current performance and perception of the aid. As an expla- nation, van Dongen and colleagues 50 reasoned that comparing reliability between automated source and oneself is a conscious and rationally driven process of (at- tributed to System 2 ) that typically is not triggered during the task and interaction with the automated aid. Instead, the automatic and effortless processing ( System 1 ), driven by the availability heuristic and anchoring, takes presence during task han- dling. Therefore, although one source can be perceived as more reliable in a post- simulation questionnaire, it was not relied upon more often during trials. System 1 and 2 refers to different thought systems based on seminal work by Kahneman and Tversky on decision-making processes. 41 System 1 thoughts are instinctive, stereo- typical, and emotional, while System 2 thoughts are slow, rational, and effortful. A similar perspective is offered by Dijkstra 173 who argued that people easily can agree with advisers who are perceived credible, even if their advice is incor- rect. According to the Elaboration Likelihood Model (ELM) people often use the peripheral decision route to make decisions based on external cues that trigger pat- tern recognition. In contrast, participants who disagreed with the automated source used the central route, which requires cognitive processing to analyze information and arguments more thoroughly before making a decision. Note that an opposite relationship also can occur. For example, Lyons and Stokes found that participants’ reliance on the human source decreased with in- creasing risk for successfully accomplishing mission objectives in the simulation. 185 Questionnaires measures, however, did not indicate any differences in reliance on the human and automated source. 4-5-2 Trust and user expertise The general preference for the human source may reflect a difference between novice and expert users. While research has shown that trust and acceptance can vary with the perceived pedigree of the decision aid, few studies have considered