The philosophy of cognitive diversity: Rethinking ethical AI design through the lens of neurodiversity

This paper redefines fairness in artificial intelligence by centering neurodiversity as an axis of justice, coequal with race, gender, and class. Conventional fairness metrics—such as demographic parity and equalised odds—fail to account for the cognitive variability, fluctuating access needs, and intersectional harms experienced by neurodivergent users. Through a normative-conceptual framework grounded in Rawlsian redistribution, Kantian autonomy, and intersectional epistemology, we diagnose ethical failures in three AI domains: hiring systems, language models, and assistive technologies. Drawing on feminist, critical-race, and disability-studies HCI, we propose a dual design paradigm: (1) Participatory design as epistemic and distributive justice, and (2) Adaptive fairness as a dynamic, user-configurable model governance logic. Each case is analysed using a structured traceability method that maps observed harms to violated principles and concrete design responses. We conclude with enforceable governance mechanisms—including procurement standards, cognitive-inclusion audits, and participatory certification—to shift AI from structural exclusion toward cognitive justice.