People Reject (Superior) Algorithms Because They Compare Them to Counter-Normative Reference Points

Abstract

People often choose to use human forecasts instead of algorithmic forecasts that perform better on average; however, it is unclear what decision process leads people to rely on (inferior) human predictions instead of (superior) algorithmic predictions. In this paper, I propose that people choose between forecasting methods by (1) using their status quo forecasting method by default and (2) deciding whether or not to use the alternative forecasting method by comparing its performance to a counter-normative reference point that is often independent of the performance of the default. This process leads people to reject a superior algorithm when (1) the algorithm serves as their alternative forecasting method and (2) the algorithm performs better than their default forecasting method but fails to meet their reference point for forecasting performance. I present the results of five studies that are consistent with this decision process. In Studies 1 through 4, participants were less likely to use a superior algorithm to complete an incentivized forecasting task when they were assigned to a relatively higher performance goal. This behavior persisted when participants recognized that their performance goal did not provide information about the relative performance of their two forecasting options, and even persisted among those participants who believed that the algorithm was the best performing option. Study 5 shows that this pattern of behavior reverses when people are assigned to use an algorithm as their default forecasting method.

Keywords: Decision Making, Defaults, Decision Aids, Heuristics and Biases, Forecasting