Algorithm aversion or appreciation: what research says
Should we fear that teams reject AI, or that they accept it unexamined? The literature documents both tendencies, under different conditions.
Documented aversion
Berkeley Dietvorst, Joseph Simmons and Cade Massey (2015, Journal of Experimental Psychology: General) show participants abandon an algorithm faster than a human after seeing it err, even when the algorithm remains more accurate. The effect is established in the lab, on forecasting tasks.
Equally documented appreciation
Jennifer Logg, Julia Minson and Don Moore (2019, Organizational Behavior and Human Decision Processes) find laypeople follow advice more when it is labelled algorithmic. The effect weakens among experts, who weight their own judgement more.
What the divergence teaches
The findings are compatible: trust depends on user expertise, error visibility and domain. These are controlled experiments; transfer to real teams, with status and accountability at stake, remains a hypothesis. (our executive and employee training programmes)
The limit of common practice
Training managers to 'trust AI' or 'distrust AI' is too crude. The useful question is calibration: where has the tool proved reliable, and how do we know?
A practical check in Lausanne
For a month, log each recommendation followed or overridden and its outcome. Then compare error rates across both groups: the only way to know whether the team suffers from aversion or overconfidence. To go further: explore the Managing in the Age of Artificial Intelligence training in Lausanne, or browse our executive and employee training programmes in Switzerland.
In pictures: Managing in the Age of Artificial Intelligence in Lausanne
- managing in the age of artificial intelligence
- Lausanne
- research
- training lausanne