Who's the real expert here? Pedigree's unique bias on trust between human and automated advisers

We assessed the effects of source type bias (human or automation) on adviser trust in a dual adviser decision-making task. Source type and reliability's effects on adviser trust have been studied in a dual-adviser context, but the influence of pedigree (perceived expertise) across source types lacked robust investigation. As situations with two decision-aids of uneven pedigree can easily arise, it is critical to understand how operators are biased towards a decision-aid of a certain source type and pedigree. A decision-making task similar to the paradigm of Convoy Leader (Lyons and Stokes, 2012) was given to participants, where a military convoy route had to be selected in the presence of IEDs and insurgent activity. We measured behavioral reliance and trust attitudes. Pedigree was manipulated via controlled adviser descriptions, in a manner consistent with past investigations (Madhavan and Wiegmann, 2007a). We found a trust bias towards the human adviser, reversed only when there is a far greater pedigree in the automated adviser. Trust attitudes were also strongly indicative of reliance behaviors. Pedigree is a strong influencer of trust in a decision-aid and biased towards human advisers. Trust is highly predictive of reliance decisions. System designers must take care with how “expert” automation is portrayed, particularly if it is used in conjunction with other human advisers (e.g.: conflicting advice from air-traffic control and an onboard system). •Pedigree is crucial in trust development in co-present human & automated advisers.•There is a trust bias for human decision-aids compared to automated decision-aids.•Human trust bias is reversed when automation pedigree far outweighs human pedigree.•Trust attitudes are predictive of behavioral reliance outcomes. Participants completed a decision-making task with the help of a human decision-aid and an automated decision-aid. The pedigree, perceived expertise, of each decision-aid was systematically manipulated. It was found that there is generally higher trust in human decision-aids, even when the automated aid is seen as similar to an expert.