Engagement Is Not Transfer: A Withdrawal Study of a Consumer Social Robot with Autistic Children at Home

Why Read This from an HCI Perspective

This article encourages readers to view social robots not as ‘machines that respond well,’ but as ‘interaction environments within a child’s everyday life.’ In particular, it empirically shows that high usability and satisfaction do not necessarily translate into real improvements in social skills. For HCI/UX practitioners, it prompts a renewed question: what should be considered a success metric? For researchers, it offers key clues on how to measure ‘transfer’ and which variables to examine in real home contexts.

CIT's Commentary

The study’s greatest value lies in demonstrating, in the real-world context of the home, that ‘engagement is not the same as effectiveness.’ One interpretation is that the more comfortable and enjoyable the interaction with the robot is, the more it may produce a negative effect—reducing practice opportunities for interacting with people. This concern carries over directly to AI systems where safety is critical. Designs that only increase autonomy or convenience can narrow users’ pathways for involvement, so it becomes essential to consider how transparent the system’s state is and when humans can step back in to adjust it. Another interesting point is the suggestion to redesign the evaluation method itself. The idea that withdrawal—removing the robot for a short period—could serve as a ‘test sheet’ that reveals whether transfer occurs is also relevant to HCI research methodologies. In domestic service environments, this framing could be applied differently. Platforms like Naver, Kakao, and many startup products tend to be used by families, centered on mobile, and with shorter usage cycles, so rather than simply importing conclusions from global studies, we should examine more closely ‘which behaviors end up returning to people.’ The more interactions that the AI generation can accept comfortably from the start, the more we should continuously check whether comfort is replacing relationships rather than supporting them.

Questions to Consider While Reading

• Q.If this robot’s high usability did not lead to social transfer, what clues might emerge when we look more closely at the actual interaction flow in a home?
• Q.If withdrawal is used as a tool to verify transfer, how could we design a method that measures it reliably while placing less burden on the child and family?
• Q.In Korea’s mobile- and family-centered service environment, if we apply this ‘transfer-first’ approach, which interface elements should change first?