CONCEPT
Algorithm Aversion
The empirically documented tendency to trust human judgment over algorithmic judgment <em>even when the algorithm demonstrably outperforms the human</em> — a bias whose welfare cost scales with the stakes of the domain.
Algorithm aversion designates the asymmetric reaction to errors made by algorithms versus identical errors made by humans. The asymmetry is irrational in the technical sense — the source of the error should be irrelevant to assessing its severity — but it is psychologically real and politically consequential. Research across domains from medical diagnosis to financial forecasting to parole decisions documents that people become less willing to use algorithmic tools after observing algorithmic errors than they do after observing equivalent human errors, even when aggregate algorithmic performance exceeds human performance by substantial margins. The aversion means deployment of AI decision support faces social resistance disproportionate to actual risk, and the resistance is not responsive to evidence about comparative accuracy. Understanding and designing around algorithm aversion is essential to realizing the welfare gains that Choice Engines make possible.
In The You On AI Field Guide
The welfare cost of algorithm aversion is measured in outcomes foregone. In bail decisions, algorithms outperform human judges by margins that, if