Laypeople’s use of generative AI in legal information-processing tasks

Investigate how laypeople use generative AI chatbots for legal information-processing tasks such as legal information retrieval and translation by conducting naturalistic evaluations of everyday user interactions, in order to characterize typical usage patterns and associated risks of errors.

Background

The paper notes that everyday users increasingly turn to technology for legal advice and information-processing tasks (e.g., translation, legal information retrieval), but there is limited evidence about how chatbots affect these users. Errors in such contexts can be catastrophic, as in asylum applications where machine translation errors can lead to rejections.

The authors emphasize the need for naturalistic evaluations to understand real-world use, highlighting current gaps in knowledge about how laypeople interact with generative AI for legal tasks.