ALSA 2025 meeting

Generative AI is increasingly used in the legal field, offering low-cost public legal consultations and improving access to justice. However, issues such as unauthorized practice, unclear responsibility, and the public’s difficulty in judging AI accuracy remain. This raises the need to understand how people evaluate AI-generated legal advice and what factors shape those evaluations.
This study empirically investigates how people perceive and evaluate AI-generated legal advice by presenting identical consultation content attributed either to an AI lawyer or to a human lawyer. Drawing on Construal Level Theory , the study aims to explore how individuals’ information processing and judgment patterns differ when evaluating legal advice concerning a case involving a close acquaintance compared to a stranger’s case. To this end, a vignette experiment will be conducted with members of the general public, manipulating two variables: the source of advice (AI vs. human) and psychological distance (close vs. distant). Participants will then evaluate the advice in terms of its perceived quality, reasonable price, and willingness to follow the recommendation.
First, in psychologically close situations, participants are expected to assign a lower price to advice labeled as AI-generated and show lower intentions to accept it compared to advice labeled as coming from a human lawyer. Second, in psychologically distant situations, no meaningful differences are expected in price judgments or acceptance intentions regardless of the source of advice. Third, in close situations, the negative evaluation of AI-labeled advice is expected to be explained by lower ratings of its contextualization and customization.
By revealing the psychological mechanisms underlying people’s evaluation of AI as a legal advisor, this study provides empirical evidence relevant to discussions on the adoption, regulation, and potential overreliance on AI-based legal advisory systems.