Algorithm Aversion: People Erroneously Avoid Algorithms After Seeing Them Err

Abstract

A landmark study demonstrating that people are more likely to abandon algorithmic forecasters after witnessing them make errors, even when the algorithm consistently outperforms human forecasters. This asymmetry in error tolerance reveals a fundamental bias in how we evaluate human versus machine competence.

Key Findings

• People lose confidence in algorithms more quickly than in human forecasters after seeing errors
• This aversion persists even when the algorithm demonstrably outperforms humans
• Giving people the ability to slightly modify the algorithm reduces aversion significantly

Source