The effects of disclosing an algorithm’s inner workings and analytic thinking on algorithm reliance

Abstract This study draws on the idea that disclosing information about how algorithmic decision aids process data (‘inner workings’) may increase algorithm reliance. In this context, the decision-maker’s thinking style is likely to play a crucial role, as analytic thinking may help to process algorithm information and therefore to foster algorithm reliance. In two between-subject experiments, I examine whether (the extent of) information provided about an algorithm and thinking style influence individuals’ reliance on the algorithm. The findings suggest that disclosing an algorithm’s inner workings significantly increases algorithm reliance up to a certain threshold, implying a concave relationship between information load and algorithm reliance. While analyses do not indicate an interaction effect between information disclosure and analytic thinking, a joint effect on algorithm reliance is shown. Furthermore, supplemental analyses of the study uncover the underlying mechanisms of algorithm reliance: Analytic thinking and other individual characteristics, such as a general faith in technology, are positively related to perceived advice source credibility that, in turn, is positively correlated with algorithm reliance. In this regard, the effect of advice source credibility appears to be stronger for analytic and weaker for intuitive thinkers. Additionally, I find that numerical differences such as the span between advices or advice direction correspond to varying levels of algorithm reliance. Thus, the study presents valuable insights for designing decision-making environments that may reinforce the utilization of data-driven approaches.