“Let Me Take That Back”: Designing and Testing Deletion Interfaces for Human-AI Interaction

Abstract

When interacting with large language models (LLMs), users have limited visibility and control over how their personal information is retained, particularly after incorporation into model training. Although some deployed LLMs provide options to opt-out of training, these mechanisms do not offer guarantees about the deletion of previously shared information. Hence, currently deployed LLMs rarely offer verifiable mechanisms to delete personal information from trained models or to demonstrate that deletion occurred. We conducted a within-subjects online experiment (N = 64) comparing three proposed interfaces, that enable users to request and receive feedback on the selective deletion of personal information in an LLM-based system: (1) a basic progress indicator, (2) a console-style technical display, and (3) an interactive verification chat. We measured user trust, perceived deletion, and usability. Participants appreciated transparent technical feedback, while interactive verification yielded the highest perceived deletion and trust without compromising usability. From our findings, we derive design recommendations for responsible, privacy-respecting, and user-controlled LLM systems.