"Computer Says No": Disabled Welfare Experiences and Envisioned Futures Under AI Governance

Paper Title

"'Computer Says No': Disabled Welfare Experiences and Envisioned Futures Under AI Governance"

Publication Info

• Topic area: The intersection of AI governance, welfare systems, and accessibility for disabled individuals.
• Keywords: AI governance, welfare automation, disability rights, aphasia, participatory design, accessibility, digital exclusion, hostile design, inclusive AI, public services.

Background and Problem

• Problem / challenge: The digitisation and automation of welfare systems risk exacerbating inequalities for disabled individuals, particularly those with communication disabilities like aphasia. Current systems are often inaccessible, dehumanising, and fail to account for the lived experiences of disabled communities.
• Significance: Welfare services are critical for the safety, health, and well-being of disabled individuals. Ensuring accessibility and fairness in these systems is essential to prevent exclusion and harm.
• Motivation and related work: Prior research has highlighted the barriers disabled individuals face in digital and automated systems but has largely overlooked the specific experiences of people with aphasia. Existing studies have not sufficiently engaged disabled communities in shaping inclusive AI-driven welfare systems.

Solution

• Proposed approach: A four-stage participatory design study involving 42 co-designers, including people with aphasia, caregivers, and speech and language therapists (SLTs), to explore their experiences with digitised welfare systems and envision inclusive AI-automated futures.
• Novelty:
  1. Identification of five core challenges in current welfare systems: the cost of performing disability, geographies of inequity, navigating digital bureaucracy, the accessibility paradox, and hostile design.
  2. Development of tangible tools (e.g., card decks, journey-mapping materials) to facilitate co-design with people with aphasia.
  3. Creation of speculative design fiction videos to explore and critique future AI-automated welfare systems.
• Procedure and key techniques:
  • Stage 1: Advisory board formation and unstructured interviews to ground the research in lived experiences.
  • Stage 2: Workshops using tangible tools to map welfare access and emotional experiences.
  • Stage 3: Journey mapping to trace barriers and facilitators in accessing key services.
  • Stage 4: Speculative design fiction workshops to envision and critique AI-automated welfare futures.
  • Reflexive thematic analysis to identify key themes and insights.

Results

• Concrete findings:
  • Five challenges identified: performing disability, regional and digital inequities, inaccessible digital bureaucracy, paradoxical accessibility claims, and hostile system design.
  • Co-designers envisioned AI systems that are patient, multimodal, empathetic, transparent, and retain human oversight.
• Advantage over baselines:
  • The participatory approach directly incorporated the lived experiences of people with aphasia, offering unique insights into systemic barriers and inclusive design possibilities.
  • Proposed AI systems address specific accessibility needs, such as multimodal communication and proactive support.
• Experiments / evaluation:
  • Eight workshops with 42 participants, including 29 people with aphasia, using tangible tools and speculative videos to elicit insights.
  • Analysis of 2141 discussion instances to generate themes and recommendations.
• Limitations and future work:
  • Findings are context-specific to the UK and may not generalise globally.
  • Speculative design prompts were not based on deployed systems, limiting real-world applicability.
  • Future work should test proposed AI systems in real-world settings and expand participatory methods to other disability contexts.

Summary

This study investigates the experiences of people with aphasia in navigating digitised welfare systems and explores their visions for inclusive AI-automated futures. Through a participatory design process, the research identifies systemic barriers and proposes AI systems that are patient, multimodal, empathetic, and transparent while retaining human oversight. The findings highlight the need for accessible, bottom-up digital governance and compassionate AI systems that empower rather than exclude disabled individuals. Future research should focus on real-world testing and broader community engagement to ensure inclusive welfare automation.