Unlock Life with a Chat(GPT): Integrating Conversational AI with Large Language Models into Everyday Lives of Autistic Individuals

Title of the Paper

Unlock Life with a Chat(GPT): Integrating Conversational AI with Large Language Models into Everyday Lives of Autistic Individuals

Paper Information

• Research Areas: Human-Computer Interaction, Artificial Intelligence, Neurodiversity, and Educational Technology
• Keywords: Conversational Agents, Large Language Models, Autism, Participatory Design, AI Ethics, User Experience

Research Background and Issues

• Problems or Challenges:

  • Autistic individuals face challenges in daily life and social interactions, including information management, problem-solving, and understanding social norms.
  • Traditional support networks may not meet their needs in a timely manner, and online support communities often lack tailored information for autistic individuals.
  • Despite rapid advancements in AI and natural language processing (NLP) technologies, research on their effectiveness in addressing the needs of autistic individuals remains limited.

• Significance:

  • Conversational agents (CAs) powered by large language models (LLMs), such as ChatGPT, offer capabilities for instant and structured information retrieval, potentially helping autistic individuals solve problems independently and receive emotional support.
  • The diverse and complex needs of autistic individuals highlight the importance of designing inclusive technologies to support their social participation and self-understanding.

• Research Motivation and Related Work:

  • Previous studies have focused on online social interactions and technology support for autistic individuals but rarely delve into open-domain conversational technologies like LLMs.
  • The academic community is increasingly shifting towards leveraging technology to better support the autonomy and self-expression of autistic individuals.

Solution

• Methods or Solutions:

  • This study engaged with 14 autistic individuals through focus group interviews and workshops to explore their expectations and challenges regarding LLM-driven conversational agents.
  • The research aims to center autistic individuals in the use of technology, empowering them to define the significance and role of these technologies.

• Innovative Aspects:

  • Emphasizing the proactive role of autistic users in adapting technology, rather than relying solely on the technology to adjust itself.
  • Providing foundational data for inclusive improvements to LLM-driven CAs through detailed user needs collection and multi-topic validation design.

• Implementation Steps:

  1. Group participants for focus interviews to gather their initial perceptions of AI and LLM-driven CAs.
  2. Encourage participants to explore strategies for solving daily problems using ChatGPT during workshops.
  3. Apply qualitative data analysis methods (e.g., thematic categorization, coding) to conduct multi-level analysis based on interview and interaction results.

Research Outcomes

• Main Findings:

  • Autistic participants expressed a desire for ChatGPT to support them in areas such as daily planning, navigating social situations, acquiring knowledge, emotional support, and self-reflection.
  • The study innovatively extracted 99 user-generated conversation topics, spanning diverse domains from personal health to time management.
  • ChatGPT demonstrated adaptability and provided online resources that helped autistic individuals articulate their unique needs.

• Advantages and Experimental Results:

  • ChatGPT showed significant advantages in providing instant information, adjusting social interaction strategies, and supporting autistic individuals in building self-confidence.
  • Experiments revealed that autistic participants highly appreciated ChatGPT's logical and detailed feedback.

• Limitations and Future Directions:

  • Limitations:
    • All participants were Korean users, which may limit the generalizability of the findings across different cultural contexts.
    • Data collection was conducted in controlled experimental environments, with limited reflection of real-world daily interactions.
    • Non-verbal autistic individuals were not included in the study.
  • Future Directions:
    • Conduct long-term follow-up studies to validate the usage patterns and effectiveness of conversational agents in daily life.
    • Design multimodal technology support tools for non-verbal autistic individuals.
    • Enhance the representation of neurodiversity in AI training datasets to improve adaptability for autistic individuals.

This study provides a theoretical and practical foundation for AI technology support tailored to autistic individuals, potentially driving the development of more inclusive artificial intelligence.