CounselReflect: A Toolkit for Auditing Mental-Health Dialogues

TL;DR

CounselReflect is a comprehensive toolkit designed to enable transparent, multi-dimensional auditing of mental-health dialogues generated by conversational AI systems, supporting users and professionals.

Contribution

It introduces a novel, multi-faceted evaluation system combining model-based and rubric-based metrics, with flexible deployment options and demonstrated human and expert usability.

Findings

Supports transparent inspection with session summaries and turn-level scores.

Integrates 69 literature-derived and custom metrics via LLM judges.

Human and expert evaluations indicate it is understandable, usable, and trustworthy.

Abstract

Mental-health support is increasingly mediated by conversational systems (e.g., LLM-based tools), but users often lack structured ways to audit the quality and potential risks of the support they receive. We introduce CounselReflect, an end-to-end toolkit for auditing mental-health support dialogues. Rather than producing a single opaque quality score, CounselReflect provides structured, multi-dimensional reports with session-level summaries, turn-level scores, and evidence-linked excerpts to support transparent inspection. The system integrates two families of evaluation signals: (i) 12 model-based metrics produced by task-specific predictors, and (ii) rubric-based metrics that extend coverage via a literature-derived library (69 metrics) and user-defined custom metrics, operationalized with configurable LLM judges. CounselReflect is available as a web application, browser extension, and command-line interface (CLI), enabling use in real-time settings as well as at scale. Human evaluation includes a user study with 20 participants and an expert review with 6 mental-health professionals, suggesting that CounselReflect supports understandable, usable, and trustworthy auditing. A demo video and full source code are also provided.