Reputational Algorithm Aversion
Gregory Weitzner
TL;DR
This paper addresses why humans resist incorporating algorithm forecasts despite potential benefits by embedding a Bayesian reputational cheap talk game with a binary state $omega$ and worker types $\theta_L,\theta_H$. It shows that the first-best use of information cannot be sustained in equilibrium: the high-skill worker reports his own signal while the low-skill mixes between signals, creating algorithm aversion, and the mixing probability $\gamma$ rises with algorithm uncertainty $\alpha$. The model yields a unique informative equilibrium where the low-skill sometimes overrides the algorithm even when it reduces forecast accuracy, aligning with empirical observations of overriden algorithm decisions. These results imply wage and adoption effects driven by reputational concerns, suggesting that improving collaboration between humans and AI may require coordinating decision authority with algorithm development or redesigning observability of the algorithm signal to mitigate signaling costs.
Abstract
People are often reluctant to incorporate information produced by algorithms into their decisions, a phenomenon called ``algorithm aversion''. This paper shows how algorithm aversion arises when the choice to follow an algorithm conveys information about a human's ability. I develop a model in which workers make forecasts of an uncertain outcome based on their own private information and an algorithm's signal. Low-skill workers receive worse information than the algorithm and hence should always follow the algorithm's signal, while high-skill workers receive better information than the algorithm and should sometimes override it. However, due to reputational concerns, low-skill workers inefficiently override the algorithm to increase the likelihood they are perceived as high-skill. The model provides a fully rational microfoundation for algorithm aversion that aligns with the broad concern that AI systems will displace many types of workers.