Creating Disability Story Videos with Generative AI: Motivation, Expression, and Sharing

Paper Title

Creating Disability Story Videos with Generative AI: Motivation, Expression, and Sharing

Publication Info

• Topic area: Generative AI for disability storytelling and advocacy
• Keywords: Generative AI, disability storytelling, accessibility, digital storytelling, identity representation, emotional expression, video creation, social media, assistive technology, advocacy

Background and Problem

• Problem / challenge: Creating disability-related videos is hindered by accessibility barriers, limited media production skills, and ableist biases in existing tools. Generative AI (GenAI) offers potential but often misrepresents disabilities and perpetuates stereotypes.
• Significance: Storytelling videos are critical for raising awareness, advocating for disability rights, and fostering community. GenAI could lower barriers to video creation and enable broader participation in disability advocacy.
• Motivation and related work: Prior research has explored storytelling as a reflective and expressive practice for people with disabilities (PwDs) and the potential of GenAI in creative work. However, gaps remain in understanding how GenAI supports video-based storytelling, especially for PwDs with limited media experience.

Solution

• Proposed approach: The study introduces the Momentous Depiction framework, which identifies four key affordances of GenAI for disability storytelling: non-capturable depiction, identity representation and non-disclosure, context realism and consistency, and emotion & social experience articulation.
• Novelty:
  1. Framework for understanding GenAI’s affordances and limitations in disability storytelling.
  2. Exploration of how GenAI supports PwDs in creating and sharing videos about lived or imagined disability experiences.
  3. Identification of challenges in GenAI outputs, including biases, inconsistencies, and misrepresentations.
  4. Design implications for integrating GenAI into storytelling tools for PwDs.
• Procedure and key techniques:
  • Participants (9 PwDs) used ChatGPT for scripting, DALL-E for image generation, and ElevenLabs for voiceovers to create six-scene, 1-minute videos.
  • The study involved pre-study training, GenAI-assisted video creation, and post-study interviews.
  • Data analysis included thematic coding of prompts and interview responses to address motivations, co-creation processes, and sharing intentions.

Results

• Concrete findings:
  • GenAI enabled participants to depict real-world challenges (e.g., inaccessibility, societal barriers) and express emotions and imagined futures.
  • Participants used GenAI to create alternative identities, ensuring privacy while maintaining authenticity in storytelling.
  • GenAI outputs often included inaccuracies, such as inconsistent visuals, misrepresented assistive technologies, and stereotypical portrayals.
• Advantage over baselines:
  • GenAI reduced the effort required for video creation and allowed participants to visualize moments that are difficult to capture in real life.
  • Enabled storytelling without requiring participants to appear on camera, addressing privacy and accessibility concerns.
• Experiments / evaluation:
  • Participants created videos using standardized prompts and tools, followed by semi-structured interviews.
  • Themes were identified through qualitative analysis of prompts and interview data.
• Limitations and future work:
  • Limited participant diversity; future research should explore broader disability representation.
  • Focused on novice creators; experienced creators may use GenAI differently.
  • Examined only specific GenAI tools (ChatGPT, DALL-E, ElevenLabs); newer tools may offer additional capabilities.

Summary

This study investigates how Generative AI (GenAI) can support people with disabilities (PwDs) in creating storytelling videos, focusing on motivations, expression, and sharing intentions. The proposed Momentous Depiction framework identifies four affordances of GenAI: non-capturable depiction, identity representation and non-disclosure, context realism and consistency, and emotion & social experience articulation. While GenAI enabled participants to create impactful videos and lowered barriers to media production, challenges such as biases, inconsistencies, and misrepresentations were identified. The findings highlight the potential of GenAI to enhance disability advocacy and storytelling, with design implications for improving GenAI tools to better support PwDs.