Predictors of AI Knowledge Among Social Work and Allied Social Science Students: A Multi-University Analysis

Background and Purpose: The rapid integration of artificial intelligence (AI) into higher education has generated widespread calls for AI literacy among students in professional fields (Benton, 2023). Within social work and allied social science disciplines, AI literacy encompasses not only technical competency but also the capacity to evaluate AI’s impact on social inequalities, ethical practice, and service delivery (Ahn et al., 2025). Yet recent qualitative research indicates that students in these fields acquire AI knowledge primarily through informal, self-directed pathways rather than through institutional support or guidance (Lee et al., under review). While descriptive studies have documented students' AI use patterns and attitudes (e.g., Meinhardt-Injac et al., 2025), fewer have examined what factors are associated with AI knowledge. This study examines the demographic and learning context factors associated with self-rated AI knowledge among social work and allied social science students.

Methods: Data were drawn from a cross-sectional survey administered from May 2025 to March 2026 across three U.S. universities. The sample (N = 396) was predominantly graduate-level (76.3%), female (80.3%), and White (67.4%), with a mean age of 26.9 years (SD = 8.3). The largest disciplinary representation was social work (59.1%), followed by public health (12.1%), criminal justice (10.6%), and psychology (7.6%). Nearly 43% reported experiencing financial difficulties, and only 17.2% reported being financially comfortable. A hierarchical multiple regression was conducted to predict students' self-rated AI knowledge composite scores. Demographic characteristics (academic level, age, gender, race/ethnicity, perceived financial situation) were entered in Block 1, followed by AI learning context variables (self-study via apps/tools, formal coursework, social media, on-campus AI learning exposure) in Block 2.

Results: The full model was statistically significant, F(9, 324) = 4.37, p < .001, R² = .11. The demographic block alone accounted for 7.8% of variance, and the addition of learning context variables produced a significant R² change of .03 (p = .028). Self-directed learning via apps and tools was the strongest predictor in the final model (B = 1.60, β = .158, p = .003). Age was negatively associated with AI self-knowledge (B = -.10, β = -.156, p = .006), and perceived financial situation was positively associated (B = .68, β = .120, p = .026). Formal coursework, on-campus AI learning exposure, social media, academic level, gender, and race/ethnicity were not significant predictors.

Conclusions and Implications: The prominence of self-directed learning as a predictor, alongside the non-significance of formal coursework and on-campus exposure, suggests that current educational structures in social work and allied social science disciplines may not be adequately building AI knowledge for their students. The association between financial situation and self-rated AI knowledge raises equity concerns that are particularly salient in a field committed to social justice; that is, if AI knowledge is acquired primarily through informal, resource-dependent pathways, students with fewer financial resources may be systematically disadvantaged. Social work programs and allied disciplines should consider developing structured, discipline-specific AI learning opportunities that build on students’ existing self-directed behaviors while reducing dependence on resources students may not equally possess.