Does training induce algorithm aversion?
15:30-17:00, Friday, April 29, 2022
Tencent Meeting (Meeting ID: 543 946 735)
The last decade has witnessed rapid advances in artificial intelligence (AI) based on new machine learning techniques. Despite the hypes and prospects of AI-enabled decision-making support applications, 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 not less willing to incorporate AI advice in making predictions. Meanwhile, they are more reactive to AI's incidental performance. In follow-up experiments, we show that there is asymmetry in expert's reactions to AI's incidental performance. Compared to the control group participants, trained experts reduced their weight on AI advice more when an incidence where AI underperformed was observed, while they did not become more reliant when they were outperformed by AI. The training effect carries over to other tasks. We discuss the theoretical mechanism for the observed training effect from the perspective of construal-level theory and mental heuristics.
For more information of the seminar, scan the following QR code(s) to join Tencent QQ group (904 544 292) or WeChat group named "IAER Seminar (3)", please.
(QR code is valid until 02/5/2022)