Understanding Algorithm Aversion: Forecasters Erroneously Avoid Algorithms After Seeing them Err

When faced with the choice between a human judge and a (superior) statistical algorithm, forecasters often choose the human judge, but the basis for this choice is unknown. We hypothesized that forecasters are less likely to choose an algorithm after seeing it err. In three experiments, participants either saw an algorithm’s prediction performance, a human’s prediction performance, both, or neither. Then they decided whether to tether their incentives to the future predictions of the algorithm or the human. Participants who saw the algorithm’s performance were less likely to choose it, even for those who saw the algorithm outperform the human.