A staged, human–computer interaction-informed model for designing art therapy digital interventions

Background Artificial intelligence (AI) is being used in mental health field, yet its integration into art therapy is developing. While digital tools are already used, introducing AI into therapeutic practice raises concerns. Therefore, the development of methods needs to focus on process execution and practitioner engagement instead of using technology as a replacement.Aims This study aimed to develop a staged, human–computer interaction-centred model guiding the design and evaluation of AI-supported art therapy interventions but maintains therapist agency and core therapeutic principles.Methods Three sequential empirical studies were conducted by the first author over three years, following the Double Diamond design framework. Study 1 employed semi-structured interviews investigating how art therapists currently use technology. Study 2 conducted co-design sessions with mid-fidelity prototyping to determine suitable AI functions in art therapy practice, followed by semi-structured interviews. Study 3 evaluated the feasibility of an AI-assisted intervention through survey-based feedback. Findings were synthesised across studies to create a generalised design framework.Results Synthesising findings across the three studies, a four-stage, three-dimensional (4 × 3) model is proposed, demonstrating how human–computer interaction methods and therapeutic focus interact across stages to support structured flexibility and user’s autonomy.Conclusions The proposed model shows how staged, human-centred design processes can support the responsible integration of AI into art therapy, extending existing human–computer interaction design frameworks to therapeutic contexts.Implications for future research This model provides process-level guidance for researchers developing AI-supported tools in art therapy, emphasising the importance of staged, user-centred human–computer interaction approaches in future digital mental health research.