Integrating Conversational Artificial Intelligence into Mental Healthcare: Stakeholders’ Perspectives on Ethical Challenges

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Background: Therapeutic Conversational Artificial Intelligence (CAI) is increasingly viewed as a promising response to the widespread prevalence of poor mental health. While commercial therapeutic CAI tools are gaining popularity, they raise significant ethical and practical concerns. This study explores ethical challenges of integrating CAI into mental healthcare by combining the Ethics by Design for AI framework of Brey and Dainow with stakeholder perspectives. Methods: Seven focus groups (n=32) were conducted with three stakeholder groups: user representatives with lived experience (n=10, 2 focus groups), psychologists and psychiatrists (n=12, 3 focus groups), and experts in AI ethics or technology in mental healthcare (n=10, 2 focus groups). Data were analysed using reflexive thematic analysis.Results: The analysis identified 23 subthemes and four potential levels at which CAI could be integrated into mental healthcare. Across all stakeholder groups, participants recognised CAI’s potential to complement existing services by improving access and supporting clients, particularly when developed and deployed under professional supervision. However, human contact was considered essential for effective treatment. Key concerns included unclear accountability, insufficient training for providers, limited AI literacy, and the absence of clear standards regarding user responsibility, moral values, and content quality. Participants emphasised that safe implementation requires co-design approaches, phased integration, and continuous evaluation within public or non-profit domains. Fully autonomous use of CAI was deemed premature due to low trust, regulatory gaps, and uncertain risk thresholds. Conclusion: This study helps bridge the gap between ethical theory and practical implementation by demonstrating how stakeholder-informed design can ensure that CAI supports, rather than replaces, human-centred mental healthcare.

Article activity feed