Huda Al Hussari, Christian Brück, Thorsten Knauer, Christian Kremer
Better Decisions with Good Advice from Humans or Algorithms?
Abstract
In this study we investigate the effects of algorithm aversion in a business advisory setting. Distinguishing between two central components of advice - the quality of advice and the advisor - we examine how superior advice provided by an algorithm affects decision-making quality. In a laboratory experiment, we vary the advisor (human vs. algorithm) and the quality of advice (superior vs. inferior). First, proving egocentric discounting in advice-taking, we show that in situations of inferior quality advice, decision-making quality is lower compared to when advice is absent, and in situations of superior quality advice, decision-making quality is higher compared to when advice is absent. Furthermore, algorithm aversion fosters a worse reputation for an algorithm advisor compared to a human advisor, which leads to lower decision-making quality in cases of an algorithm-based superior quality advice. However, the results suggest that decision-making quality is still better when the algorithm provides superior advice than when no advice is given. We show that decision-makers discount advice according to its quality, and discount algorithm advice more than human advice. Overall, we advise companies to account for the effects of algorithm aversion when establishing algorithm advice.