Microeconomics SeminarDivisive by Design: Shaping Values in Optimal MechanismsBen Chen (University of Hong Kong), VMP 5 (Room 0079)

8 January 2026

Abstract:  While generative artificial intelligence (GenAI) promises to enhance productivity, its practical exploitation remains inconsistent. In legal applications where accuracy is paramount, bad initial experiences may induce professional hesitancy, thereby stifling further exploration of the technology’s capabilities. This study investigates whether task-specific training can break this dynamic by driving human adoption of GenAI for assignments requiring specialized knowledge or judgment. An experiment conducted on 164 law students finds that an intervention comprising a training video and a short quiz significantly increased use of a Large Language Model (LLM) on an issue-spotting examination. Moreover, those who received the intervention achieved significantly better scores than those who were not exposed to treatment but were also allowed to use the LLM. These results suggest that task-specific training can be critical in ensuring that the legal profession fully captures the benefits of emergent technology.