Training and AI Aversion

Artificial Intelligence (AI) outperforms humans in many tasks. However, accumulating evidence suggests AI aversion – a tendency to distrust and thus ignore AI advice, especially among experts. In a two-stage experiment, we show that compared to a control group with equivalent experience in a prediction task, trained experts are slightly more acceptive of AI advice in forming predictions. Meanwhile, trained experts are more sensitive to the AI’s incidental performance. Experts reduced their weight on AI advice more when an incidence where AI underperformed was observed, while they became more reliant when they were outperformed by AI. Follow-up studies show that the punishment is more significant and the effect extends to irrelevant tasks. This result challenges the notion of AI aversion, suggesting that experts are willing to adopt AI advice while being more responsive to its performance.