Artificial Intelligence (AI) has advanced rapidly in recent years, moving from Research & Development (R&D) labs and startups into broader use. Organizations are using AI to screen job candidates, streamline the application process, monitor employee actions, and provide employee training. When not designed and implemented to consider diverse users, AI technologies can increase the risk of workplace discrimination, including for people with disabilities.
Equitable AI refers to AI technologies that humans intentionally design, develop, and implement to result in more equitable outcomes for everyone, including people with disabilities. Organizations can adopt Equitable AI Principles to guide their design, development, and implementation of AI to mitigate discrimination and reduce bias in AI technologies. Employers who adopt Equitable AI Principles can help ensure a diverse, equitable, and inclusive workplace for all existing and future employees.
PEAT developed the AI & Disability Inclusion Toolkit to help your organization navigate the potential risks of implementing AI technologies (specifically for people with disabilities), to outline practices you can adopt to try making AI implementations more equitable, and to help you make a business case for Equitable AI to organizational leaders.
Because people with disabilities are members of all communities, disability is a dimension of diversity that crosses all others. This AI & Disability Inclusion Toolkit can support organizations’ Diversity, Equity, Inclusion, and Accessibility (DEIA) efforts, and the content within the Toolkit should be considered through an intersectional lens. The principles of Equitable AI can be a guidepost for your organization as you work toward creating an organizational culture of inclusion where everyone can advance and thrive, regardless of race, ethnicity, gender identity, sexual orientation, religion, or disability.
This Toolkit contains the following sections. To navigate, you may use either the buttons at the bottom of each page or the left sidebar.
Before you get started, there are a few things you should know about the concepts presented in this Toolkit and how we differentiate Equitable AI and Equitable AI Principles from other approaches impacting the development, design, and use of AI technologies in the workplace.
Our Approach to Content Selection
PEAT’s work in the AI space focuses specifically on supporting people with disabilities in the workplace. The content in this Toolkit is geared toward helping employers adopt and implement Equitable AI Principles that can ensure fair and unbiased treatment of job seekers and employees with disabilities. In addition, much of the information below can address AI discrimination and bias for all employees from underrepresented groups.
We selected content to help employers learn how AI tools function, how organizations can and do use AI tools in the workplace, and how to navigate the challenges in making AI implementations equitable for employees with disabilities. We also included practical guidance, tips, and resources to help employers integrate Equitable AI into their organizations.
In the Toolkit, we did not include guidance specifically for vendors who offer AI workplace tools. However, the content in the document can help vendors understand how to work with their customers (i.e., employers who purchase their products and services).
Notes on Terminology
AI experts may use terms other than Equitable AI when discussing efforts to mitigate AI discrimination and bias and concepts related to the development of fair and unbiased AI. These terms include, but are not limited to Responsible AI, AI Fairness, Ethical AI, and Accountable AI. PEAT considers the concepts behind all these terms essential to ensuring the design, development, and use of Equitable AI. This AI & Disability Inclusion Toolkit focuses on Equitable AI due to the strong intersections of this work with organizations’ broader DEIA efforts.
For more information about these terms and other terms used throughout this document, refer to the Toolkit Glossary.