Algorithm Aversion: People Erroneously Avoid Algorithms After Seeing Them Err
Berkeley J. Dietvorst, Joseph P. Simmons, Cade Massey
Journal of Experimental Psychology: General · 2015
Abstract
This paper documents that people are significantly more likely to lose confidence in algorithms than in humans after seeing them make mistakes. Even when the algorithm's errors are smaller than the human's, seeing the algorithm err makes observers revert to human predictors, which perform worse on average.
Methodology
Across five studies, participants made forecasts about real-world outcomes (e.g., student performance) and received recommendations from either a human expert or an algorithm. Some participants saw the predictors err before choosing which advisor to rely on.
Findings
Seeing the algorithm make a mistake caused a steep decline in trust. Observers punished algorithmic errors much more severely than human errors, even when they were shown evidence that the algorithm outperformed the human.
Applied nudge
To combat algorithm aversion in choice architecture, do not present algorithms as infallible. Instead, frame them as collaborative assistants and manage error expectations transparently.
Citation
Dietvorst, B. J., Simmons, J. P., & Massey, C. (2015). Algorithm aversion: People erroneously avoid algorithms after seeing them err. Journal of Experimental Psychology: General, 144(1), 114-126.
↗ Download PDF