GenRole: Personalizing Role Play for Educators Supporting Autistic Students’ Social Interaction Learning

Paper Title

GenRole: Personalizing Role Play for Educators Supporting Autistic Students’ Social Interaction Learning

Publication Info

• Topic area: Generative AI for personalized role play in autism education
• Keywords: Autism, social interaction skills, role play, generative AI, personalization, educators, progressive learning, visual components, cognitive development, usability

Background and Problem

• Problem / challenge: Existing role-play methods for autistic children rely on fixed content, requiring significant manual preparation by educators, which limits personalization and scalability.
• Significance: Personalized role-play activities can better align with autistic children’s diverse needs, fostering social interaction skills and improving their quality of life.
• Motivation and related work: Prior research highlights the effectiveness of role play and generative AI in education but lacks tools that integrate personalization for autistic learners. Existing systems often rely on predefined content, limiting adaptability to individual needs.

Solution

• Proposed approach: GenRole, a generative AI-powered system that enables educators to design personalized role-play activities tailored to the diverse needs of autistic children.
• Novelty:
  1. Integration of generative AI to create personalized role-play scripts, visuals, and reinforcers.
  2. Progressive role-play modes (Echo, Dialogue, Character, Exploration) to scaffold learning from concrete to abstract tasks.
  3. Empirical validation of GenRole’s usability and effectiveness through pilot and user studies.
• Procedure and key techniques:
  1. Teachers input social skills and preferences into the system.
  2. GenRole generates personalized scripts, visuals (characters, scenes, reinforcers), and progressive learning modes.
  3. Teachers and children engage in role play, progressing through structured modes to practice and generalize social skills.

Results

• Concrete findings:
  • Post-test scores (M = 3.03, SD = 0.81) were significantly higher than pre-test scores (M = 2.18, SD = 0.73; z = 4.35, p < 0.01).
  • System Usability Scale (SUS) score averaged 76.88, indicating high usability.
  • Teachers rated AI-generated stories positively (Mean = 4.16, SD = 0.69).
• Advantage over baselines: GenRole reduced educators’ workload by automating content generation and provided more engaging and personalized learning experiences compared to traditional role-play methods.
• Experiments / evaluation:
  • Pilot study with 16 educators to refine the system.
  • Two-week user study with 11 teacher-student pairs to evaluate usability and learning outcomes.
  • Data collected through pre/post-tests, SUS, and semi-structured interviews.
• Limitations and future work:
  • Short study duration limited observation of long-term effects.
  • All participants were boys, and the study was conducted in China, limiting generalizability.
  • Future work should expand personalization parameters, include diverse cultural contexts, and explore real-world scenario integration.

Summary

GenRole is a generative AI-powered system designed to help educators create personalized role-play activities for autistic children, enhancing their social interaction skills. The system features progressive learning modes and customizable components such as scripts and visuals, tailored to individual needs. Empirical studies demonstrated significant improvements in children’s social skills and high usability for educators. Future research will focus on expanding personalization, addressing cultural diversity, and incorporating interactive elements to further enhance engagement and learning outcomes.