Design and Evaluation of a Collaborative Virtual Environment (CoMove) for Autism Spectrum Disorder Intervention
Lian Zhang, Qiang Fu, Amy Swanson, Amy Weitlauf, Zachary Warren, Nilanjan Sarkar · 2018 · ACM Transactions on Accessible Computing (TACCESS) · doi:10.1145/3209687
Summary
This paper presents CoMove, a collaborative virtual environment (CVE) designed to measure and potentially enhance collaborative interaction and verbal communication skills of children with autism spectrum disorder (ASD) when playing with typically developing (TD) peers from remote locations. Unlike prior approaches that use computer-controlled agents or co-located shared devices, CoMove is a distributed system where two players at separate computers interact in a shared virtual space through collaborative puzzle games and voice chat. The system has two distinguishing characteristics. First, the collaborative games are designed with embedded strategies that require three specific collaborative behaviours: sequential work (turn-taking), information sharing (one player has information the other needs), and simultaneous interaction (both players must act together to move a piece). A hybrid automaton governs puzzle piece movement so that collaboration is mechanically enforced rather than merely encouraged. Second, CoMove provides an objective measurement framework that logs game performance data, human actions, and audio to quantify collaboration and communication without relying solely on subjective observation. Seven tangram puzzle games and a castle-building game were designed with varying configurations of who can see, move, and rotate pieces, creating escalating collaboration demands. A feasibility study involved 28 children (ages 7-17): 7 ASD/TD pairs and 7 TD/TD pairs, each completing a 50-minute session with pre-test, game-playing, and post-test phases.
Key findings
All participants completed the experiments with zero dropout, validating the system's feasibility and engagement. Children with ASD showed statistically significant improvements in collaborative performance from pre-test to post-test: success frequency in tangram games doubled (7 to 14, p < 0.05) and collaborative movement ratio increased significantly (0.11 to 0.22, p < 0.05, rho = 0.30). Notably, children with ASD achieved success frequencies comparable to TD children in both castle and tangram games, suggesting that the structured collaboration strategies effectively scaffolded their participation. Communication measures showed trends toward improvement but did not reach statistical significance given the small sample — children with ASD increased their question-asking frequency in tangram games to levels comparable to TD peers by post-test. The enforced-collaboration games requiring simultaneous actions produced the most significant increases in collaborative movement ratio, suggesting that games demanding real-time coordination may be particularly effective for promoting collaboration in ASD. The TD children paired with ASD children (TD1 group) also showed significant improvements in several measures including word frequency and information-sharing responses, indicating that the system benefited both partners. The authors note important limitations: participants had average IQ and phrased speech, so results should not be generalized to all children with ASD.
Relevance
CoMove demonstrates a promising technology-mediated approach to social skills intervention for children with ASD that addresses several limitations of existing methods. Evidence-based ASD treatments are expensive and scarce, making technology-based alternatives particularly valuable for increasing access. The distributed CVE design solves the sensory overload problem of face-to-face co-located interaction — by removing visual co-presence and limiting communication to voice, the system reduces the multisensory integration demands that can overwhelm children with ASD while still enabling meaningful collaboration. For researchers and clinicians, the objective measurement framework offers a standardized way to quantify collaborative behaviours that is less subjective than observational coding alone. The enforced-collaboration game design — where the system mechanics require collaboration rather than merely encouraging it — provides a model for structuring technology-mediated social learning that could extend beyond ASD to other populations who benefit from scaffolded social interaction. For the broader accessibility field, the work exemplifies how virtual environments can create controlled, safe social practice spaces that bridge toward real-world interaction skills.
Tags: autism · collaborative virtual environment · social skills · peer interaction · puzzle games · verbal communication · objective measurement · intervention · children