CAIA in Practice: Field Evaluation of an AI-Assisted Support System for Text-Based Online Counselling

AI tools must preserve professional judgment: The study found that counsellors sustained use of CAIA only when the system supported their autonomy rather than replacing their decision-making. Interpretive functions that "stimulate professional reflection" were particularly valued, suggesting AI works best as a thinking partner rather than an automated decision-maker in sensitive contexts like mental health care.

Accuracy is non-negotiable for trust: Information accuracy emerged as a "decisive" factor for whether counsellors would continue using the AI system. This finding underscores that in high-stakes domains like counselling, even occasional errors can undermine trust and adoption, requiring rigorous quality assurance before deployment.

Co-design with end-users improves responsible outcomes: The tool was "co-designed" with counsellors rather than imposed on them, which likely contributed to the strong adoption rates (34 professional counsellors actively using it across 321 messages). This suggests involving practitioners in AI development from the start helps ensure systems meet real-world needs and ethical standards.

AI can expand access to mental health support responsibly: By supporting counsellors in asynchronous email counselling—described as a "crucial low-threshold channel"—CAIA demonstrates how AI might help address "rising global demand for mental health support" without replacing human expertise, potentially making care more accessible.

Transparency about AI capabilities matters: The study's focus on which specific functions counsellors valued (interpretive tools over others) suggests that clear communication about what AI can and cannot do helps professionals use it appropriately and maintain accountability to their clients.