ATTENPlay: A Game-Based Attention Network Test for Autistic Children
Yuying Wan, Xin Tong, Kaishun Wu · 2026 · Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI '26) · doi:10.1145/3772318.3790265
Summary
This CHI 2026 paper addresses a long-standing methodological problem in autism research: the Attention Network Test (ANT) — the standard computerized tool for measuring alerting, orienting, and executive-control attention networks — is abstract, repetitive, and built for neurotypical adults, producing high non-completion rates and noisy data when applied to young autistic children. The authors redesign the ANT as ATTENPlay, a tablet-based single-tap game set in a forest where children catch fireflies under traffic-light Go/No-Go rules, and follow a three-phase inclusive design process: (1) eight co-design sessions with four domain experts (two pediatric neurologists, a special-education teacher, an HCI researcher) to surface motor, rule-comprehension, pacing, and sensory challenges; (2) a survey of 32 parents (17 autistic, 15 neurotypical children) evaluating device familiarity, rule comprehension of traffic-light vs. directional-arrow mappings, and story-based vs. abstract framings; (3) a comparative pilot with six children pitting an early ATTENPlay prototype against ANT-I-Child to surface remaining usability issues. The final system maps each attention network to a specific child-facing mechanic: a centrally presented white circle with chime for alerting, a curved firefly flight path for orienting, and Go/No-Go on green vs. red light for executive control. ATTENPlay was evaluated in a counterbalanced study at a special-education school with 52 children aged 3-10 (28 autistic, 24 neurotypical), with quantitative performance metrics, parent interviews, researcher field notes, and child preference probes.
Key findings
ATTENPlay closed the data-collection gap that has plagued ANT research with autistic children. Task completion for the autistic group rose from 71.4% on ANT-I-Child to 92.9% on ATTENPlay (chi-square=4.17, p<0.05), while neurotypical completion was 100% on both. Learning time was dramatically shorter for autistic children on ATTENPlay (2.85 vs. 5.95 minutes; t=6.60, p<0.001, d=2.10), and session duration was roughly halved (16.95 vs. 28.50 minutes, p<0.001, d=4.94). Accuracy rose from 41% on ANT-I-Child to 84.83% on ATTENPlay for autistic children, with the primary difference driven by a lower false-alarm rate — suggesting motor rather than attentional confounds in the traditional paradigm. On the cleaner ATTENPlay data, the alerting network did not differ significantly between autistic and neurotypical children (d=-0.57, p=0.67), but the orienting network showed reduced anticipatory benefits in autistic children (p<0.01), while reorienting costs were preserved — a selective predictive-attention difference rather than general inflexibility. The executive control network (measured via d') was the most affected (mean 2.62 vs. 3.76 for neurotypical, p<0.001), and executive control correlated with age in neurotypical but not autistic children, suggesting asynchronous developmental trajectories. Parent and child preference strongly favored ATTENPlay (95%+ in 'Again-Again' and 'Fun Sorter' probes), and parents highlighted gentle feedback, visible score display, and diegetic audiovisuals as key usability improvements.
Relevance
For accessibility practitioners working on cognitive assessment, pediatric digital health, or autism-inclusive research methodology, this paper is a strong concrete example of how 'true accessibility' for a neurodivergent population requires redistributing cognitive and interactional load across the whole task rather than layering gamification on an unchanged paradigm. The authors explicitly frame the work through the double-empathy problem (interactional failures belong to the tool, not the child) and articulate a transferable five-level design framework — interface patterns, mechanics, principles, conceptual models, and inclusive co-design methods — that can inform other cognitive tests being adapted for disabled populations. Practical contributions include a traffic-light Go/No-Go metaphor that lowers rule-comprehension burden, single-tap interactions that remove fine-motor confounds, diegetic audiovisual feedback that prevents sensory overload, and tolerance-aware gesture detection. Limitations are upfront: reliance on the traffic-light metaphor limits use with toddlers or children with significant cognitive delays, the study did not control for ADHD comorbidity, it has not yet undergone formal psychometric validation, and sample size limited deeper analyses. Especially useful for anyone designing assessment, intervention, or research-participation tools for autistic children.
Tags: autism · autism spectrum disorder · attention network test · cognitive assessment · game-based assessment · serious games · tablet · co-design · inclusive design · child development · executive function · double empathy problem · pediatric · mobile accessibility
Standards referenced: Declaration of Helsinki (2024)