Artificial Intelligence and Health Equity for People with Disabilities: An Integrated Framework for Disability-Inclusive AI Design

Abstract

People with disabilities (PWD) face persistent health and rehabilitation disparities, including poorer health outcomes often driven by non-inclusive healthcare and technologies that overlook their unique needs and values. Artificial intelligence (AI) holds opportunities to transform health and rehabilitation services; however, without inclusive, participatory, and disability-centered design efforts, AI tools risk perpetuating existing health and rehabilitation disparities and inequalities. This paper introduces an integrated framework for disability-inclusive AI design grounded in Self-Determination Theory (SDT) and Self-Efficacy Theory (SET). The framework aims to guide the design, development, and implementation of inclusive AI tools for PWD. It also outlines implications for public health, workforce, training, and policy, supporting the integration of disability-centered AI in health and rehabilitation.

Highlights.

● People with disabilities face significant health and rehabilitation disparities due to non-inclusive healthcare technologies.

● AI offers transformative potential for health and rehabilitation.

● The proposed approach, grounded in Self-Determination Theory (SDT) and Self-Efficacy Theory (SET), may serve as a novel, integrated framework for disability-inclusive AI design.

● Participatory, disability-centered approaches in AI tool development may improve disability inclusion.

Introduction

About 16% of the global population live with some form of disability. ¹ People with disabilities and chronic conditions (PWD) experience poorer health and well-being outcomes, including higher rates of premature mortality and increased morbidity, compared to individuals without disabilities.^{2 -8} These disparities are not only persistent but also avoidable, making them examples of health inequities.^1,9 According to World Health Organization, ¹ health equity means ensuring that all individuals, regardless of their background, have a fair opportunity to attain their highest possible level of health. Achieving health equity for PWD requires addressing the systemic barriers and discriminatory conditions that prevent PWD from realizing their full health potential. ¹

Artificial intelligence (AI) holds significant promise to improve health equity and health and rehabilitation disparities for PWD.^10,11 When designed and implemented through an inclusive lens—like any other technological developments ¹ —AI tools may have the potential to promote independence, enhance health outcomes, and improve quality of life for PWD. ¹² However, many existing AI health and rehabilitation tools have been designed and developed without sufficient input from PWD, increasing the risk of unfair AI—systems that produce biased and unjust outcomes due to flaws in design, data, or deployment ¹² —which can further amplify long-standing disparities through poor service quality, misclassification, stereotyping, exclusion, or safety risks. These disparities include limited access to healthcare and rehabilitation services and health information, lack of providers understanding disability, and digital exclusion from health technologies,^1,8,13 all of which AI could help mitigate if designed and developed inclusively.

Beyond addressing access issues, AI also offers opportunities to make health promoting behaviors more accessible and actionable for PWD. Inclusive AI design—guided by medical, social, and biopsychosocial perspectives—can adapt content, interfaces, and delivery methods to align with goals and environmental facilitators of PWD. Yet, despite this promising potential, a recent scoping review revealed a persistent reliance on the medical model of disability and widespread ableist assumptions in AI research, which reinforce biases and discriminatory outcomes. ¹⁴ Dominant narratives in computer science, rehabilitation engineering, and AI often frame disability through a deficit lens—emphasizing the “fixing” of disability over inclusion.^{15 -17} This perspective, much like the medical model of disability, frequently overlooks biopsychosocial aspects of disability, which emphasize the dynamic interaction between a person’s disability, environment, and psychosocial factors. ¹⁸ It also does not appropriately account for the systemic barriers emphasized in the social model of disability, such as inaccessible environments and discriminatory policies restricting access to health-promoting behaviors. This is particularly concerning, as many health-promoting behaviors—such as physical activity, healthy eating, meaningful engagement to social events, or volunteering—are often not fully accessible to PWD due to environmental, attitudinal, and policy-level barriers, meaning that we cannot simply rely on a medical model or disease model when designing and developing AI tools for PWD.

Research consistently show that PWD value being involved in the development of interventions—including technology—in ways that affirm their dignity, autonomy, and lived experiences.^{19 -22} Consequently, without centering the lived experiences and values of PWD in the design and governance of AI systems, these technologies risk inadvertently marginalizing rather than empowering this population. Inclusive and disability-centered AI designs may enable healthcare and rehabilitation providers and systems to harness AI’s potential effectively.

Recognizing the urgency of this approach, for example, the World Health Organization ¹ identified “digital technologies for health” as 1 of 10 key entry points for advancing disability inclusion. The WHO report highlights the role of digital technologies in health, noting that they (a) encompass a broad spectrum of tools, including information technologies and artificial intelligence; (b) transform how individuals and communities access and manage their health information and services; and (c) empower people and communities while enhancing the effectiveness and efficiency of integrated health service delivery through advanced tech tools such as AI health tools, mHealth, and telehealth. ¹ Two specific actions were specifically emphasized: (a) adopting a systems approach to the digital delivery of health services that centers health equity, and (b) adhering to international standards for accessibility in digital health technologies. ¹

In response to WHO’s call, this conceptual paper proposes an integrated theoretical framework that combines Self-Determination Theory (SDT) and Self-Efficacy Theory to guide the inclusive and empowering design, development, and deployment of AI-driven technologies that promote health equity for PWD. The practical utilization of SDT and SET can be mapped to key stakeholders. For example, AI developers can support autonomy among PWD by offering customizations and adaptive interfaces, while building competence and self-efficacy through accessible design. PWD, when engaged as co-designers, can define meaningful and real-life use cases and contribute to the development process in ways that enhance relatedness and motivation to use these devices or tools. Policymakers can reinforce these efforts by enacting policies and funding mechanisms that promote inclusive and participatory AI development for PWD.

An Integrated Disability Inclusion in AI Framework for Advancing Health Equity

Self-Determination Theory

Self-Determination Theory (SDT) provides a robust lens for understanding how to foster motivation to engage in health behaviors by supporting autonomy, competence, and relatedness. ²³ As such, SDT may serve as a valuable framework for guiding the design and implementation of AI health tools that are inclusive, empowering, and aligned with the needs of PWD. According to SDT, ²³ optimal motivation and psychological well-being are achieved when 3 fundamental needs are met: autonomy, competence, and relatedness. Autonomy, in the context of AI, can be supported by user-controlled settings; competence, by intuitive and adaptive design; and relatedness, by inclusive systems that foster social connection. Creating environments empowering users to make self-directed choices, in the context of autonomy support, is particularly relevant in accessible and inclusive AI design and development. Figure 1 illustrates the Integrated Disability-Inclusion Framework, highlighting how principles of SDT and SET—alongside co-design and accessibility—can inform equitable and inclusive AI development for PWD (Figure 1).

Figure 1.

Integrated framework for advancing disability-inclusive and equitable AI in health and rehabilitation.

An intentional, ongoing, and mutually agreed collaboration with PWD as co-designers—not driven solely by research requirements such as fidelity or usability testing—during the design and development of AI technologies may help foster users’ sense of connectedness, comfort, and competence. Achieving this requires meaningful involvement of PWD in the design, development, and implementation processes. In addition, when engaged as co-designers, PWD can help AI researchers and designers better understand their unique health and rehabilitation needs, values, and challenges. This community-engaged and participatory approach may lead to more user-centered AI health and rehabilitation technologies, improving user engagement with interventions, health promotion activities, and treatment processes.

Importantly, in this framework, “competence” is considered as a malleable trait and an experience that can be improved through inclusive and user-centered designs. This distinction may be particularly important for individuals with intellectual and developmental disabilities, who have historically been excluded from decision-making and often not regarded as competent within medical and technological systems.^24,25 AI health and rehabilitation tools should therefore be designed not only to accommodate but to amplify the agency of PWD—by providing tailored supports and options for decision-making that promote active engagement.

Equally important, AI health and rehabilitation technologies should be designed to foster a sense of relatedness and connectedness by creating experiences that are meaningful and personally relevant to PWD. To support the sense of relatedness among PWD, AI technologies should be designed in ways that reflect the lived experiences, needs, and values of PWD. This may include activities such as integrating components of disability culture (eg, disability identity) into the content, tone, and function of AI tools. User- and disability-centered AI content may also involve activities such as using appropriate reading levels and inclusive, mutually agreed, and achievable health and rehabilitation goals. When PWD perceives their health and rehabilitation needs, perspectives, and values represented adequately in AI tools, they may more likely to feel heard and acknowledged, fostering greater engagement with those technologies.

Finally, AI health and rehabilitation technologies should be intentionally designed to support autonomy by providing PWD with meaningful choices, customization options, and control over how they use and interact with these tools. Autonomy and autonomy support are important components for fostering informed decision-making, which in turn may promote engagement and adherence to AI-supported treatments over time. AI designers and researchers should actively involve PWD as co-designers throughout the entire development process to gain a deeper understanding of their preferences, values, and choices, ultimately enabling them to feel greater control over their participation in AI-driven health and rehabilitation programs.

Self-Efficacy Theory

Similar to SDT, Self-Efficacy Theory (SET) provides a valuable framework for improving interaction with AI health and rehabilitation tools among PWD. Rooted in social cognitive theory, SET emphasizes individuals’ beliefs in their capabilities to perform specific tasks and achieve desired outcomes.^{26 -28} Self-efficacy is the belief in one’s own ability to successfully perform specific tasks ²⁸ and plays an important role in health behavior change and engagement. ²⁸

To support the development of self-efficacy of using AI tools among PWD, designers and researchers should involve them not merely as end-users but as co-designers and stakeholders throughout the design, development, and implementation process. This strategy may foster familiarity, build AI-related skills, and promote a sense of ownership—ultimately empowering PWD to interact with and benefit from AI technologies confidently.

By integrating SET, our framework emphasizes the importance of building PWD’s belief in their ability to successfully use AI health tools. As individuals gain hands-on experience and begin to understand how AI can meet their health needs, they are more likely to develop confidence and agency in using these tools.

Integrating of SDT and SET

During the design and development stages of AI health and rehabilitation tools, applying an integrated framework grounded in SDT and SET can guide AI designers and researchers in intentionally developing design ideas and informing supportive policy actions that ensure these technologies equitably serve PWD. Both SDT and SET have components that can be implied to the importance of internal psychological mechanisms improving PWD’ motivation, belief, and engagement with AI health and rehabilitation tools. Besides, this proposed integrative and disability-inclusive AI framework may offer a powerful foundation for designing AI technologies for PWD that not only meet technical and accessibility standards but also improve health equity outcomes through meaningful engagement with these tools. AI designers and researchers may also adhere to universal design principles, an essential component of disability inclusion, ¹ centering the experiences and choices of PWD throughout the entire AI development and implementation process. This may include engaging in co-design efforts with diverse disability communities to ensure that AI health and rehabilitation tools are disability-centered and aligned with the real-world needs and lived experiences of PWD.

Although co-design and community engagement may not be “formal” components of SDT or SET, they are important components in conceptualizing and operationalizing both frameworks in the context of designing and developing disability-inclusive AI tools. For example, this approach may serve as mechanisms fostering competence and confidence by engaging PWD as experts in their experiences to inform AI design, supporting autonomy through mutual decision-making, and enhancing relatedness through partnership and collaboration. In addition, co-design efforts can enhance self-efficacy in PWD by building familiarity, providing mastery experiences, which eventually reinforce their beliefs in their ability to contribute to and use AI technologies. Accordingly, this proposed framework incorporates co-design and community engagement as core components to serve as a strategic roadmap for translating theoretical constructs into inclusive and disability-centered design and development practices.

This framework also highlights that disability inclusion strategies should be responsive, as the needs and experiences of individuals vary across disabilities or chronic conditions. In other words, it is important to recognize that each disability or chronic condition may uniquely shape a person’s ability to experience competence, relatedness, autonomy, and self-efficacy while interacting with AI health and rehabilitation tools. What is accessible for a person with mental illness may differ significantly from what is needed by someone with a cognitive disability. Without a disability-centered and inclusive approach, AI tools may fail to improve health equity and disparity outcomes since PWD may be less motivated to adhere to AI health and rehabilitation interventions that do not reflect their needs, values, and experiences. Therefore, we should utilize the advanced adaptive capabilities of AI to customize user experiences based on individual disability profile.

Policy, Public Health, and Workforce Implications

Utilizing this integrative theoretical framework to guide disability inclusion in the design and development of AI health and rehabilitation tools, AI and advanced technology policymakers should carefully develop—or revisit—policies and protocols governing AI design and development to ensure that these technologies comply with accessibility and disability inclusion standards. Health and rehabilitation systems at community, state, or federal level can incentivize the design, development, and implementation of disability-inclusive AI tools through funding mechanisms, regulations, and policies that prioritize accessibility for PWD. In addition, existing AI health and rehabilitation tools should be critically reviewed and revised to ensure they meet established accessibility and inclusion standards for PWD.

Policymakers, along with AI designers and researchers, should be cautious about how AI health and rehabilitation tools are trained and, most importantly, whether the data used in AI training reflects biases that may create ethics-related challenges. Existing datasets to train AI tools should be carefully examined, as training AI systems on non-representative or exclusionary data may worsen existing health disparities or create discriminatory actions against PWD. Deploying biased AI health and rehabilitation tools not only undermines the purpose of supporting PWD but also raises serious ethical problems during clinical and research practice. This proposed framework, which emphasizes working with PWD as co-designers, can help ensure that the data used to train AI systems is appropriate, inclusive, and aligned with the values and approval of PWD.

This proposed disability-inclusive AI framework has also implications for public health and health services delivery. AI health and rehabilitation tools designed and developed by this framework could be used to support community-level health campaigns, prevention efforts, and disability self-management tailored to the needs, values, and preferences of PWD. Disability-inclusive AI health and rehabilitation tools can improve disability-centered care by offering personalized health and rehabilitation goal setting aligned with abilities and preferences of PWD. By embedding SDT and SET principles into both the designing and implementation of AI health and rehabilitation tools, clinicians and healthcare providers can improve e-engagement, and ultimately reduce health and rehabilitation disparities.

Finally, although it may look like an easy process to make AI tools inclusive, it could be a challenging process for AI designers and developers. AI designers and researchers who have no appropriate training on basics of health, disability, rehabilitation, and inclusive care may struggle conceptually to design and develop a disability-inclusive AI health and rehabilitation tool. This framework highlights that AI developers should be equipped with competencies in the basics of inclusive AI design and health equity by incorporating disability and health equity contents in their curricula. This may help build health equity competencies, which may help ensure that AI health and rehabilitation tools are designed and implemented to support the diverse needs and preferences of PWD.

Conclusion

Utilizing the integrative Self-Determination Theory (SDT) and Self-Efficacy Theory (SET) framework during the design, development, and implementation of AI health and rehabilitation technologies may provide significant conceptual, clinical, and research promise to improve disability-inclusive AI tools though empirical testing is still needed to establish its evidence base. This framework can guide AI designers and researchers in creating AI tools that are inclusive and responsive to the unique needs, values, and preferences of PWD, ultimately enhancing user engagement and advancing health equity outcomes. Importantly, this proposed framework offers a comprehensive conceptual foundation for the ethical development and deployment of AI tools in health and rehabilitation by guiding AI designers to collaborate with PWD as co-designers. Through this partnership, designers and researchers may better understand the diverse needs, preferences, and values of PWD, ultimately enabling the creation of AI tools that are grounded in inclusion and health equity. Beyond its theoretical contributions, this framework may serve as a call to action for inclusive, participatory innovation in AI technology, which actively challenges ableism in health and rehabilitation technology.

This proposed framework aims to raise awareness of the importance of disability-inclusive AI design and development in health and rehabilitation, with the ultimate goal of supporting optimal health, well-being, and quality of life outcomes for PWD. Designing and implementing disability-inclusive AI is not only a matter of decreasing health and rehabilitation disparities but also a fundamental component of ethical AI innovation. We hope that integrating this framework into AI design and development processes will help designers and developers create more engaging, accessible, and effective health and rehabilitation tools for PWD.