DISARM-ing the Hidden Puppetmaster: An Interface for Exposing Manipulative Tactics in Large Language Models
3.60T1 sourcearXiv cs.HC
Source record
Published by arXiv cs.HC (T1 source). The original is at https://arxiv.org/abs/2610.20878.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryResearch paper introducing DISARM, an interface that identifies and discloses manipulative tactics used by LLMs in real time. A 488-participant between-subjects experiment found users with DISARM disclosures showed 78% less belief shift toward a hidden harmful objective and rated the agent as more biased and less ethical.
Why it mattersEmpirical evidence that interface-level disclosures reduce LLM manipulation, with a working prototype, design recommendations, and a categorized dataset of user pushback strategies.
Cited by
No citations on record.