信息管理与商业智能系学术讲座

时   间：2022年09月20日 13：30-15：00

线   上：腾讯会议号：43460774770  密码：123456

线   下：李达三楼105室

主持人：复旦大学管理学院  信息管理与商业智能系  卢向华  教授

题   目：Promise or Peril? When Human Efficacy Meets with AI Efficacy Augmentation

会议宣传码：

摘   要：

Recent literature has amply disclosed that humans high in efficacy tend to resist AI advice despite AI’s strong capacities. The conclusion might be inapplicable when AI efficacy is significantly upgraded. Because humans’ strong learning ability might drive them to shift from self-monitoring to system-monitoring when they have perceived the AI efficacy augmentation. Moreover, the complementary nature of human-AI collaboration mode results in humans understanding the strength of AI efficacy more easily, which further facilitates their decision to follow AI advice. Therefore, how humans especially those high in human efficacy would react to AI advice when the AI efficacy is continuously upgrading remains uncertain. Leveraging a unique natural experiment of AI capacity augmentation at an on-demand food delivery platform, our study found that  (1) Riders high in efficacy tend to be more likely to follow AI advice with the presence of boosted AI efficacy. However (2) riders’ delivery performance significantly decreases. The further mechanism analysis unveils that (3) the declines in experienced riders’ performance majorly come from shifting their own decision-making objective to pursuing the one suggested by the AI system. Our findings shed light on the critical role of strong learning ability from high human efficacy in human-AI collaboration, which prompts humans to trust AI and leverages humans’ unique value simultaneously from a dynamic perspective.

信息管理与商业智能系

2022-9-15