CollaDrum: Designing a Tangible Interactive Sonic System to Foster Collaboration in Children with Autism Spectrum Disorder
Yinhan Gao, Yalong Luo, Wenwei Jiang, Wanling Cai, Yucheng Jin · 2026 · Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems (CHI EA ’26) · doi:10.1145/3772363.3799088
Summary
CollaDrum is a co-designed tangible sonic system intended to scaffold task-based peer collaboration for autistic children in classroom settings, with explicit attention to not adding cognitive overhead from novel interaction paradigms. The authors interviewed five educators (1–10 years experience) at a private Chinese institution serving autistic children and used inductive thematic analysis to surface four design considerations: calibrated sensory cues for attention without distraction (DC1), interaction rules that force awareness of peer state rather than parallel play (DC2), graded difficulty for mixed-ability groups (DC3), and explicit non-punitive success/failure feedback as regulatory rather than evaluative signals (DC4). The hardware is a cloud-shaped, 3D-printed PLA enclosure built around an ESP32, with three silicone-covered colored pads (capacitive touch), an MPU6050 IMU for shake detection, a WS2812 RGB strip for diffuse light feedback, and a DFPlayer Mini for both percussion and non-instrumental (e.g., animal) sounds. The tap-and-shake interaction is deliberately drawn from the percussion instruments children already use in classroom music sessions. A coordination computer maintains shared state across four tangible devices over Wi-Fi, and a mobile educator interface lets a second educator switch game modes, adjust difficulty parameters such as timing tolerance, mute devices, and trigger synchronized light/sound reminders. Two collaborative game modes were co-designed: Turn-Taking, which requires reproducing a sound sequence with the active child indicated by a white light, and Simultaneous-Play, which requires designated children to act within a shared time window. A pilot study used two N=3 groups ("high-functioning" and "low-functioning" by the authors' terminology).
Key findings
The two groups responded very differently. The high-functioning group picked up device use quickly, attended to peers' interactions, verbalized collaborative intentions (described as spontaneous verbal coordination), and asked to attempt higher difficulty levels. The low-functioning group struggled with rule comprehension and showed limited responsiveness to the multimodal sensory feedback, suggesting that the abstraction from concrete drum to networked rule-based game broke down without further scaffolding. A specific design failure surfaced across both groups: red-light plus buzzer failure feedback induced anxiety and avoidance in some children, and educators recommended replacing it with gentler, more encouraging cues, undermining the original DC4 framing of failure signals as regulatory rather than punitive. Practical reflections include that drawing the tapping/shaking interaction from existing classroom percussion practice reduced learning overhead, and that distributing facilitation between an educator playing alongside the children and a second educator using the mobile interface kept guidance embedded in the activity rather than delivered as explicit instruction. The authors are clear that the pilot is exploratory and provides no quantitative evidence of impact on social initiative.
Relevance
The paper offers a concrete pattern for accessibility designers building collaborative classroom technology for autistic children: design the interaction modality from artifacts the children already know (here, drums and percussion), embed collaborative scaffolding inside individual devices rather than in shared tabletops, and split educator roles between in-game participant and out-of-game coordinator. The DC4 reflection on failure feedback is the most useful negative result; even thoughtfully designed non-punitive signals can be received as punitive by autistic children who already experience high social-evaluation anxiety, and the authors flag this as a real iteration target rather than papering over it. A point worth flagging for practitioners: the paper uses functioning-level language ("high-functioning" / "low-functioning") that the autism self-advocacy literature increasingly rejects in favor of support-needs framing, since the binary correlates poorly with IQ, communication ability, or daily-life support, and tends to over- and under-serve children at both ends. Limitations are substantial: total N=6, a single classroom context, only two game modes, no quantitative measures, and the system has not been evaluated against the educator-mediation alternative it claims to reduce. Long-term value will depend on the longitudinal case studies the authors plan.
Tags: autism · autistic children · tangible interfaces · classroom technology · collaboration · turn-taking · co-design · special education · sonification