How Did Clinicians Feel About Game-Based Interventions That Help Autistic Children with Motor Skills?—Field Insights for a Future Game Platform

Why Read This from an HCI Perspective

This article is not just about how to make a ‘good game.’ It shows why interactions used in real therapy settings often fail. In particular, the importance of designing in a way that preserves the therapist’s judgment rather than relying on automatic scoring, provides a structure in which children can make choices, and controls sensory stimulation directly connects to HCI/UX practice. When applying AI and games in health and care contexts, it is especially meaningful to prioritize the experience and intervention pathways over raw performance.

CIT's Commentary

What stands out is that this study redefines the game not as a ‘fun element,’ but as a work tool that the therapist actively adjusts. In particular, simple feedback structures like approve / partial / retry demonstrate that what matters more than the sophistication of automatic scoring is how a human interprets failure and can intervene. This approach carries over directly to products that incorporate LLMs or sensor-based AI: the key is not just replacing judgment with a model, but designing how and when users can roll back, and how transparently the system’s state can be viewed. The same holds in Korea’s service context—within the fast productization environment of Naver, Kakao, and startups, the bigger difference is often ‘the ability to tweak and endure in the field’ rather than whether the feature simply works. Also, proposing Wizard-of-Oz not as a temporary technique but as an ongoing interaction model expands the research questions by challenging the assumption that automation is always the correct answer.

Questions to Consider While Reading

• Q.In real practice, how does a design that leaves the therapist’s judgment intact—rather than relying on automatic scoring—reduce certain burdens, and what new burdens might it introduce?
• Q.To increase a child’s autonomy while still keeping therapy goals aligned, how far should we open up the range of choices?
• Q.If this Wizard-of-Oz-style structure is later transformed into an AI-supported system, what should be designed first: the state indicators or the intervention pathways?