Social robots for children with autism are often evaluated through engagement and interaction quality, assuming the robot acts as a social scaffold. We report a mixed-methods "withdrawal" study that tests a harder question: what changes when the robot is removed. In an 8-week home-based randomized controlled trial (N=40), children either retained a consumer social robot (Qrobot) or had it withdrawn after initial use. Quantitatively, continued access reduced anxiety (SCARED/RCADS), yet was associated with lower parent-reported social motivation and weaker gains in emotion recognition (SMS/RMET) compared to withdrawal. Interviews with guardians contextualized this divergence: removal sometimes prompted children to seek human interaction, while continued use could keep social behavior siloed within the child-robot dyad, despite exceptionally high usability (SUS). We synthesize a UXR point of view: for vulnerable users, "engagement" can mask ecological downsides. Success should be judged not by retention, but by designed separation that bridges back to human relationships.

翻译：自闭症儿童社交机器人常通过参与度和互动质量进行评估，其假设是机器人充当社交支架。我们报告了一项混合方法“撤回”研究，旨在检验一个更棘手的问题：移除机器人后会发生什么变化。在一项为期8周的家庭随机对照试验（N=40）中，儿童要么保留使用消费级社交机器人（Qrobot），要么在初始使用后将其撤回。定量分析显示，与撤回组相比，持续使用可降低焦虑水平（SCARED/RCADS量表），但家长报告的社会动机较低，且情绪识别能力（SMS/RMET量表）的提升较弱。对监护人的访谈揭示了这一差异的成因：撤回有时会促使儿童主动寻求人际互动，而尽管机器人具有极高的可用性（SUS量表），持续使用却可能使社交行为局限于儿童-机器人二元组内。我们综合得出用户体验研究视角：对于弱势用户，“参与度”可能掩盖生态层面的负面影响。成功与否不应以用户留存为评判标准，而应通过有设计地分离来重新搭建与人类关系的桥梁。