Autoethnographic Insights from Neurodivergent GAI "Power Users"

Research Background and Issues

• Issues and Challenges: The authors observe that while Generative AI (GAI) is rapidly becoming a significant tool in both daily and professional life, its potential in the field of accessibility remains underexplored, particularly for neurodivergent populations. They point out that existing assistive technologies often fail to meet the needs of neurodivergent individuals, especially in addressing challenges that arise during long-term use of new technologies, such as managing social pressures, emotional regulation, and information access.
• Significance of the Issue: For neurodivergent groups, societal expectations of normative behavior, and the resulting marginalization and discrimination, present significant daily challenges. A key question is whether GAI can help these populations achieve greater autonomy, alleviate social pressures, and enhance productivity.
• Research Motivation and Related Work:
  • The potential of GAI to improve accessibility for people with disabilities has been preliminarily explored. For example, the BeMyEyes application uses GAI to describe images for visually impaired users, while ADHD Companion and autism support generative models have seen significant use in mental health and emotional regulation applications.
  • However, existing research primarily focuses on specific use cases or short-term effects, lacking a deep understanding of the practices and complex dynamics of long-term GAI use by neurodivergent individuals. In particular, applications involving social and psychological pressures, as well as the impacts on intersections of multiple identities (e.g., race, gender, culture), require further study.

Solutions

• Methods and Solutions: This study employs a collaborative autoethnography approach to deeply explore the real-world experiences of advanced neurodivergent GAI users. All members of the research team identify as neurodivergent and come from diverse intersectional backgrounds (e.g., race, gender, language).
• Innovations:
  1. Utilization of GAI-assisted anonymous diary and interview tools (DiaryGPT and InterviewGPT) to protect privacy while capturing authentic and detailed user cases.
  2. Focus on the experiences of long-term "advanced users" rather than short-term or sporadic use.
  3. Examination of social adaptation behaviors (e.g., code-switching, camouflaging, masking) in GAI usage, as well as the cultural and identity-related limitations of GAI.
• Key Implementation Steps and Techniques:
  • Data Collection: Study participants recorded daily GAI usage cases using DiaryGPT and participated in structured interviews via InterviewGPT at the end of the study.
  • Data Analysis: Inductive coding was applied to participants' diary and interview data, followed by multiple rounds of discussion and thematic analysis to extract key themes, including time and cognitive savings, social communication and norm adaptation, emotional and informational support, among others.

Research Findings

• Specific Findings: The authors identified four main themes regarding GAI usage:
  1. Saving Time and Cognitive Resources: Users leveraged GAI to improve efficiency, particularly when handling academic or work tasks, or during periods of low energy due to cognitive challenges (e.g., brain fog, anxiety).
  2. Supporting Social Communication and Norm Adaptation: GAI was used to enhance language expression or adjust emotional tone to align with social norms (e.g., adjusting the tone of professional emails, expressing culturally appropriate emotions).
  3. Replacing Interpersonal Interaction to Reduce Social Costs: In situations of high social pressure or when seeking help from others was inconvenient, GAI served as a tool for anonymous emotional support, simulated conversations, or generating advice.
  4. Facilitating Understanding and Integration of Complex Information: GAI proved to be an effective tool for quickly accessing and interpreting complex technical, legal, or medical information, particularly for users with cognitive impairments.
• Advantages Over Existing Solutions:
  • Anonymity and Self-Paced Use: GAI's anonymity and 24/7 availability allow users to seek help without fear of social judgment.
  • Flexibility: GAI addresses a variety of needs that are not easily met by traditional technologies, such as instant translation of legal language or generating communication materials.
  • Autonomy: It enables neurodivergent users to more flexibly define and meet their own needs, rather than passively adapting to external expectations.
• Experimental or Evaluation Results:
  • Users generally reported that GAI positively contributed to enhancing their autonomy and reducing social adaptation pressures, with significant value observed in "streamlining workflows" and "psychological support."
  • The authors also identified risks associated with GAI, such as increased dependency and the potential for inaccurate outputs, especially when users experience cognitive fatigue or lack verification capabilities.
• Limitations and Future Directions:
  • Limitations:
    1. GAI still exhibits errors and biases (e.g., misrepresentation of racial and cultural language, inaccurate outputs).
    2. Users in high-dependency scenarios are more susceptible to the biases of GAI, particularly in sensitive areas such as health and law.
    3. Neglect of cultural and identity diversity: GAI often prioritizes mainstream cultures and languages, potentially marginalizing non-mainstream identities.
  • Future Directions:
    1. Improving GAI Accuracy: Develop built-in verification mechanisms to prompt users to carefully validate critical outputs.
    2. Enhancing Cultural and Diversity Sensitivity: Incorporate more diverse cultural representations in GAI training data.
    3. Exploring Mechanisms to Further Support User Autonomy in Neurodivergent Use Cases: Focus on empowering users rather than reinforcing adherence to social norms.

In summary, this study, through the first large-scale collaborative autoethnography, demonstrates how GAI can serve as a valuable tool for neurodivergent populations while also revealing the challenges and immense potential of its use.