A Checklist for Trustworthy, Safe, and User-Friendly Mental Health Chatbots

TL;DR

This paper addresses the lack of standardized guidance for mental health chatbots amid rising demand and safety concerns. It employs a PRISMA-inspired literature review and thematic analysis to distill core design principles into an auditable checklist, then demonstrates its practicality by applying it to Woebot. The checklist covers transparency, boundaries, accuracy, user-centric design, meaningful conversations, safeguards, inclusivity, and trust-building, and it is framed as both a design guide and an auditing tool. The findings suggest the checklist can support safer, more trustworthy chatbot design and regulation, while also highlighting the challenges of evaluating subjective dimensions like transparency and empathy. The work advances responsible design practices and provides a concrete artifact for developers, end-users, and regulators to assess digital mental health tools.

Abstract

Mental health concerns are rising globally, prompting increased reliance on technology to address the demand-supply gap in mental health services. In particular, mental health chatbots are emerging as a promising solution, but these remain largely untested, raising concerns about safety and potential harms. In this paper, we dive into the literature to identify critical gaps in the design and implementation of mental health chatbots. We contribute an operational checklist to help guide the development and design of more trustworthy, safe, and user-friendly chatbots. The checklist serves as both a developmental framework and an auditing tool to ensure ethical and effective chatbot design. We discuss how this checklist is a step towards supporting more responsible design practices and supporting new standards for sociotechnically sound digital mental health tools.