When the risk feels personal: self-relative pessimism in AI-related risk perception among people with disabilities

PURPOSE: We examined AI-related risk perceptions among people with disabilities by distinguishing personal risk from societal risk. We further asked whether people with disabilities perceived themselves as more exposed to AI-related risks than society, a pattern this study terms self-relative pessimism. MATERIALS AND METHODS: Data were obtained from the 2023 Digital Divide Survey in South Korea. The analytic sample included 2,200 respondents who self-identified as having disabilities. Linear regression models were used to examine personal risk perception, societal risk perception, and optimism, which was defined as the difference between societal and personal risk perceptions. Multinomial logistic regression was used to assess membership in the equal, optimistic, and pessimistic groups. RESULTS: Most respondents reported equal levels of personal and societal risk perception, whereas a smaller subgroup showed self-relative pessimism. Dread was not significantly associated with personal risk perception but was positively associated with membership in the pessimistic group. Involvement was negatively associated with personal and societal risk perceptions and membership in the pessimistic group. Perceived Controllability showed a dual pattern: it was negatively associated with personal and societal risk perception but positively associated with membership in the pessimistic group. Disability type showed no consistent association, and the degree of disability showed only limited associations. CONCLUSIONS: Among people with disabilities, AI-related risk perception was more clearly associated with selected sociotechnical conditions than with diagnostic disability categories. The findings point to considerations for AI-enabled assistive technology, including accessible risk communication, user control and safeguards, and reporting pathways within the service delivery.