Accessibility Considerations in the Development of an AI Action Plan

2025-03-14

AI Summary

This paper addresses the widespread neglect of accessibility requirements for persons with disabilities in AI systems by proposing the first comprehensive, lifecycle-spanning accessibility framework for AI. Methodologically, it integrates eight interrelated dimensions: open-source model development and validation, accessible dataset curation, inclusive UI/UX design, specialized tooling integration, privacy-preserving mechanisms, bias detection and mitigation, and cross-cutting ethical and governance considerations. Its key innovation lies in enabling cross-layer, stage-aware modeling and empirical validation of accessibility requirements across AI research, deployment, and evaluation. The resulting actionable roadmap explicitly supports meaningful participation of disability experts and demonstrably reduces both direct and indirect discriminatory risks in AI-enabled accessibility contexts. (136 words)

Technology Category

• Humans and AI: AI for Accessibility
• Philosophy and Ethics of AI: Bias, Fairness & Equity
• Computer Vision: Bias, Fairness & Privacy

Application Category

• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization
• Responsible Web: Algorithmic accountability and transparency on the web
• Security and Privacy: Security and privacy of machine learning and AI applications

Abstract

We argue that there is a need for Accessibility to be represented in several important domains: - Capitalize on the new capabilities AI provides - Support for open source development of AI, which can allow disabled and disability focused professionals to contribute, including - Development of Accessibility Apps which help realise the promise of AI in accessibility domains - Open Source Model Development and Validation to ensure that accessibility concerns are addressed in these algorithms - Data Augmentation to include accessibility in data sets used to train models - Accessible Interfaces that allow disabled people to use any AI app, and to validate its outputs - Dedicated Functionality and Libraries that can make it easy to integrate AI support into a variety of settings and apps. - Data security and privacy and privacy risks including data collected by AI based accessibility technologies; and the possibility of disability disclosure. - Disability-specific AI risks and biases including both direct bias (during AI use by the disabled person) and indirect bias (when AI is used by someone else on data relating to a disabled person).

Problem

• Addressing accessibility in AI development for disabled users.
• Ensuring AI models include accessibility in data and interfaces.
• Mitigating disability-specific risks and biases in AI applications.

Innovation

• Open source AI development for accessibility contributions
• Data augmentation to include accessibility in training sets
• Accessible interfaces for disabled AI app users