arXiv:2512.02282v1 Announce Type: new
Abstract: Large language models (LLMs) now mediate many web-based mental- health, crisis, and other emotionally sensitive services, yet their psychosocial safety in these settings remains poorly understood and weakly evaluated. We present DialogGuard, a multi-agent frame- work for assessing psychosocial risks in LLM-generated responses along five high-severity dimensions: privacy violations, discrimi- natory behaviour, mental manipulation, psychological harm, and insulting behaviour. DialogGuard can be applied to diverse gen- erative models through four LLM-as-a-judge pipelines, including single-agent scoring, dual-agent correction, multi-agent debate, and stochastic majority voting, grounded in a shared three-level rubric usable by both human annotators and LLM judges. Using PKU-SafeRLHF with human safety annotations, we show that multi- agent mechanisms detect psychosocial risks more accurately than non-LLM baselines and single-agent judging; dual-agent correction and majority voting provide the best trade-off between accuracy, alignment with human ratings, and robustness, while debate attains higher recall but over-flags borderline cases. We release Dialog- Guard as open-source software with a web interface that provides per-dimension risk scores and explainable natural-language ratio- nales. A formative study with 12 practitioners illustrates how it supports prompt design, auditing, and supervision of web-facing applications for vulnerable users. Read More
