Towards accessible and ethical AI tools for disability services: a stakeholder-informed study

Abstract

Integration of artificial intelligence (AI) into disability social services introduces both significant opportunities and complex challenges, exacerbated by social, institutional, and technical barriers. In this paper, we analyse the data from semistructured interviews with 30 actors from disability social services, drawing on literature in sociotechnical systems, disability studies, and value-sensitive design. Accessibility, algorithmic fairness, and socio-ethical considerations emerge as key barriers to effective AI adoption in the disability services. We identify that, in addition to technical specifications, a combination of social and institutional factors among disability service actors and technology providers may either hinder or accelerate disability-inclusive AI development and deployment. Based on the findings, we emphasise the importance of direct and systemic engagement of persons with disabilities in AI systems design through participatory approaches. To illustrate how disability-inclusive participatory design can be implemented in practice, we provide examples of good practice and lessons learnt informed by the perspectives of disability services actors.

Ähnliche Arbeiten