Cooperative Game-based Vocabulary Learning for Children on the Autism Spectrum

Abstract

Language acquisition challenges significantly impact the quality of life for children with autism spectrum disorder (ASD), yet access to evidence-based interventions remains limited by cost, geography, and resource availability. This thesis presents the design, development, and evaluation of CoVoL (Cooperative Vocabulary Learning), a web-based tool that translates Applied Behavior Analysis Verbal Behavior (ABA VB) tacting interventions into a digital format while introducing novel cooperative gameplay mechanics. Following a Design Science Research methodology, the project progressed through three iterations addressing distinct research questions: (1) alignment between literature-derived design principles and expert-elicited requirements, (2) technical feasibility of real-time speech recognition for vocabulary learning tasks, and (3) potential of shared-screen cooperative mechanics to facilitate language acquisition.The developed artifact utilizes Vosk speech recognition toolkit, achieving 247ms mean recognition latency and 88% functional accuracy through error recovery algorithms when tested on 11,005 adult speech samples. Expert evaluation (n=20 across iterations) revealed strong alignment between theoretical principles and practical requirements, while identifying domain-specific needs such as programmable reinforcement delays and customizable rewards. The cooperative two-player mode received conditional support from experts, who recognized its potential for future research on peer modeling and social engagement in ASD.Key contributions include demonstrating promising digital translation of ABA VB mechanics, introducing functional accuracy as an alternative speech recognition metric for behavioral contexts, and providing an open-source architecture for accessible autism interventions. While limitations include the absence of end-user testing with children and evaluation limited to adult speech samples, the modular design enables potential future expansion to other verbal behaviors and end-user tests. This work bridges the gap between evidence-based practices and accessibility, with potential for community-driven development of digital autism support tools.