# DialogGuard: Multi-Agent Psychosocial Safety Evaluation of Sensitive LLM Responses

arXiv: https://arxiv.org/abs/2512.02282 · HTML: https://mlsanity.com/paper/2512.02282/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Medical and scientific domain agents, one of 16 directions in LLM agents and tool use.
- Builds on 34 of its 87 references that are in the map.
- First posted on arXiv 1 Dec 2025.

## Placed in

- LLM agents and tool use / [Medical and scientific domain agents](https://mlsanity.com/agents/medical-and-scientific-domain-agents/)

## Builds on (most cited first, 30 of 34)

- [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://mlsanity.com/paper/1910.13461/) (2019)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset](https://mlsanity.com/paper/2307.04657/) (2023)
- [Automated Hate Speech Detection and the Problem of Offensive Language](https://mlsanity.com/paper/1703.04009/) (2017)
- [BOLD: Dataset and Metrics for Measuring Biases in Open-Ended Language Generation](https://mlsanity.com/paper/2101.11718/) (2021)
- [Justice or Prejudice? Quantifying Biases in LLM-as-a-Judge](https://mlsanity.com/paper/2410.02736/) (2024)
- [Hallucination is Inevitable: An Innate Limitation of Large Language Models](https://mlsanity.com/paper/2401.11817/) (2024)
- [PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preference](https://mlsanity.com/paper/2406.15513/) (2024)
- [HateCheck: Functional Tests for Hate Speech Detection Models](https://mlsanity.com/paper/2012.15606/) (2020)
- [Ex Machina: Personal Attacks Seen at Scale](https://mlsanity.com/paper/1610.08914/) (2016)
- [Understanding the Benefits and Challenges of Using Large Language Model-based Conversational Agents for Mental Well-being Support](https://mlsanity.com/paper/2307.15810/) (2023)
- [OpenAgentSafety: A Comprehensive Framework for Evaluating Real-World AI Agent Safety](https://mlsanity.com/paper/2507.06134/) (2025)
- [Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers](https://mlsanity.com/paper/2504.18412/) (2025)
- [The Art of Defending: A Systematic Evaluation and Analysis of LLM Defense Strategies on Safety and Over-Defensiveness](https://mlsanity.com/paper/2401.00287/) (2023)
- [Detecting High-Stakes Interactions with Activation Probes](https://mlsanity.com/paper/2506.10805/) (2025)
- [Evaluating the Consistency of LLM Evaluators](https://mlsanity.com/paper/2412.00543/) (2024)
- [Modularized Networks for Few-shot Hateful Meme Detection](https://mlsanity.com/paper/2402.11845/) (2024)
- [Understanding Silent Data Corruption in LLM Training](https://mlsanity.com/paper/2502.12340/) (2025)
- [AI-based Clinical Decision Support for Primary Care: A Real-World Study](https://mlsanity.com/paper/2507.16947/) (2025)
- [SafetyAnalyst: Interpretable, Transparent, and Steerable Safety Moderation for AI Behavior](https://mlsanity.com/paper/2410.16665/) (2024)
- [Towards Understanding the Robustness of LLM-based Evaluations under Perturbations](https://mlsanity.com/paper/2412.09269/) (2024)
- [AgenticEval: Toward Agentic and Self-Evolving Safety Evaluation of Large Language Models](https://mlsanity.com/paper/2509.26100/) (2025)
- [Beyond Pipelines: A Survey of the Paradigm Shift toward Model-Native Agentic AI](https://mlsanity.com/paper/2510.16720/) (2025)
- [Before It's Too Late: A State Space Model for the Early Prediction of Misinformation and Disinformation Engagement](https://mlsanity.com/paper/2502.04655/) (2025)
- [A Robot-Led Intervention for Emotion Regulation: From Expression to Reappraisal](https://mlsanity.com/paper/2503.18243/) (2025)
- [What People Share With a Robot When Feeling Lonely and Stressed and How It Helps Over Time](https://mlsanity.com/paper/2504.02991/) (2025)
- [Robot-Led Vision Language Model Wellbeing Assessment of Children](https://mlsanity.com/paper/2504.02765/) (2025)
- [A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents](https://mlsanity.com/paper/2505.22073/) (2025)
- [RADAR: A Risk-Aware Dynamic Multi-Agent Framework for LLM Safety Evaluation via Role-Specialized Collaboration](https://mlsanity.com/paper/2509.25271/) (2025)
- [LEXI: Large Language Models Experimentation Interface](https://mlsanity.com/paper/2407.01488/) (2024)

## Built on by (most cited first, 3 of 3)

- [Agentic Abstention: Do Agents Know When to Stop Instead of Act?](https://mlsanity.com/paper/2606.28733/) (2026)
- [SPASM: Stable Persona-driven Agent Simulation for Multi-turn Dialogue Generation](https://mlsanity.com/paper/2604.09212/) (2026)
- [Assessing Risks of Large Language Models in Mental Health Support: A Framework for Automated Clinical AI Red Teaming](https://mlsanity.com/paper/2602.19948/) (2026)

## Abstract (arXiv)

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 framework for assessing psychosocial risks in LLM-generated responses along five high-severity dimensions: privacy violations, discriminatory behaviour, mental manipulation, psychological harm, and insulting behaviour. DialogGuard can be applied to diverse generative 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 rationales. A formative study with 12 practitioners illustrates how it supports prompt design, auditing, and supervision of web-facing applications for vulnerable users.