Mapping Applications of Artificial Intelligence in Social Support for Persons with Disabilities: A Systematic Scoping Review

The review maps AI applications in social support for persons with disabilities into five key areas and reports the number and proportion of studies in each. Relative to prior scattered single-point studies, this work integrates 72 studies through a systematic scoping review and organizes application areas using the lens of participation, autonomy, and environmental fit. A systematic search was conducted in PubMed and Web of Science in accordance with PRISMA-ScR guidelines, identifying 72 studies, with sample sizes and percentages reported for each area under an explicit methodological framework.

The review identifies a structural imbalance in the field: applications addressing functional support far outnumber those addressing psychosocial participation. This finding makes explicit the pattern that domains addressing psychosocial participation are represented by very few studies, notes that this limits generalizability, and provides a basis for future research directions. Based on comparison of study counts across areas, with Mental Health Support and Social Inclusion represented by only 3 studies (6.3%) versus 20 (41.7%) for Mobility and Navigation Assistance, a marked numerical difference.

The review summarizes key challenges in the field, including ethical concerns regarding data privacy, inadequate training datasets, high implementation costs, algorithmic biases, and limited real-world evidence of benefits. Relative to studies focused on a single technology or application, this review aggregates cross-study common barriers at the review level and reports the proportion of included studies in which each challenge appears. Challenge proportions come from aggregated statistics across the 72 included studies, with data privacy concerns at 58.3%, inadequate training datasets and high implementation costs each at 25.0%, and algorithmic biases and limited real-world evidence each at 20.8%.

The review indicates a predominantly medical-model orientation in AI development and systematic underrepresentation of persons with intellectual and psychosocial disabilities. This judgment links technology development orientation with the underrepresentation of specific disability groups, pointing to the need for social-model approaches, real-world validation, and inclusive codesign. The conclusion is based on synthesis of the overall orientation and covered populations of the included studies; it is a review-level pattern identification, which the original text frames as a field-level need to be addressed.

The review's conclusions apply to an overall understanding of the current state of AI applications in social support for persons with disabilities, primarily for researchers, technology developers, and relevant policy practitioners, to identify research distribution, common challenges, and gap areas; its analytical framework is participation, autonomy, and environmental fit, and its search scope is limited to PubMed and Web of Science.

The original text notes that domains addressing psychosocial participation are represented by very few studies, which limits the generalizability of related conclusions; additionally, limited real-world evidence of benefits (20.8%) suggests that the effectiveness of AI applications in actual settings still awaits further validation. When referencing the proportions of each area and the challenge rates, readers should consider them alongside the specific disability types and application scenarios they focus on.