# AI testing, evaluation, verification and validation for accessibility: a comprehensive framework

## TL;DR

This paper introduces a framework to test and evaluate AI systems for accessibility, ensuring they work well for all users, including those with disabilities.

## Contribution

A novel framework for AI testing and evaluation focused on accessibility, incorporating red teaming and usability testing.

## Key findings

- Systematic evaluation of AI systems for accessibility improves inclusivity and effectiveness.
- Case studies show the framework helps identify accessibility barriers and biases.
- The approach supports the development of more equitable and universally usable AI systems.

## Abstract

As artificial intelligence (AI) systems continue to remain prevalent in society, ensuring their accessibility for all users, including those with disabilities, is of great importance. This paper presents a comprehensive framework for AI Testing, Evaluation, Verification and Validation (TEVV) focused on accessibility. The proposed methodology incorporates methods for red teaming, model testing, and field testing with a particular emphasis on usability testing for accessibility. The results demonstrate, through detailed case studies, that systematically evaluating AI systems for accessibility barriers and biases improves the inclusivity and effectiveness of AI technologies for diverse user populations. The findings suggest that this accessibility-focused TEVV framework provides a structured approach for developing more equitable and universally usable AI systems that benefit all members of society.