Process Measures of Dyadic Collaborative Interaction for Social Skills Intervention in Individuals with Autism Spectrum Disorders
Joshua Wade, Arpan Sarkar, Amy Swanson, Amy Weitlauf, Zachary Warren, Nilanjan Sarkar · 2017 · ACM Transactions on Accessible Computing · doi:10.1145/3107925
Summary
This paper presents DOSE (Dyad-Operated Social Encouragement), an open-source collaborative game and data acquisition platform designed for social skills intervention in individuals with Autism Spectrum Disorder (ASD). The system addresses a gap in existing technology-based interventions: while many systems show promise for improving social skills, most rely on survey data or post-interaction assessments rather than capturing quantitative process measures during collaborative activities. DOSE is built on a Pong-style game modified for collaboration rather than competition. Four gameplay modes are available: Practice (single-player against AI), Dyad vs AI (two players share control of one avatar to defeat an AI opponent—movement only occurs when both players press the same direction), Competition (traditional two-player adversarial play), and Rally (collaborative goal of keeping the ball in play as long as possible, where moving one's avatar quickly increases the partner's avatar size). Players can be collocated (sharing a keyboard) or remote (using Skype for communication). The system records three classes of real-time data at approximately 60 Hz: time-series data (avatar positions, velocities, ball state, player inputs), event data (match starts/ends, scoring, timeouts), and session metadata. Audio is recorded for offline transcription of verbal communication. From this data, multiple metrics can be computed including Individual Contribution, Contribution Disparity, Dyad Points per Minute, and Words per Minute—providing fine-grained quantitative measures of collaborative behavior during intervention.
Key findings
A pilot study with 24 participants (6 with ASD, 18 typically developing) aged 8-17 evaluated DOSE's feasibility and initial effectiveness. Participants were paired into ASD-TD dyads (n=6) and TD-TD dyads (n=6), matched for gender and age. The study used a pretest-training-posttest design with 5-minute collocated Rally games as pre/post assessments and 15 minutes of mixed remote competitive and collaborative play as training. Statistically significant improvements were observed in verbal communication metrics. Individual Words per Minute increased from a median of 18.21 to 27.68 (p < .05, r = .37), and Dyad Words per Minute increased from 31.54 to 60.15 (p < .05, r = .46). This suggests DOSE facilitates increased verbal communication between dyad members. Individual Contribution showed a nominal but non-significant increase (p = .052), indicating participants were trending toward more active engagement. Survey results indicated high acceptability across both ASD and TD participants. Both groups reported enjoying the games (median 5/5), feeling comfortable with their partners, and perceiving performance improvements. Notably, participants with ASD rated the games as less difficult than TD participants (median 5 vs 4, p < .05), though both found difficulty appropriate. There were no significant differences between groups on enjoyment, comfort, or perceived improvement.
Relevance
DOSE represents an important advancement in technology-based social skills intervention by capturing objective, quantitative process data during collaborative activities rather than relying solely on subjective post-hoc assessments. This approach enables researchers and clinicians to identify specific behavioral patterns and measure fine-grained changes in collaborative performance. For practitioners, DOSE is freely available as open-source software, allowing adoption and extension without licensing costs. The system runs on all major platforms (built with Unity3D) and supports both collocated and remote play, making it adaptable to various clinical and educational settings. The enforced collaboration mechanic in Dyad vs AI mode—where the shared avatar only moves when both players input the same direction—creates natural opportunities for communication and coordination practice. The study's limitations include a small ASD sample (n=6), manual speech transcription (future versions could use automated speech-to-text), and possible ceiling effects on task difficulty. However, the preliminary evidence that DOSE increases verbal communication during collaborative play is encouraging. The platform's ability to collect process data opens possibilities for more targeted interventions based on real-time behavioral metrics rather than solely outcome measures.
Tags: autism spectrum disorder · social skills · collaborative games · dyadic interaction · process measures · intervention · open source · serious games