AI has not triggered mass academic layoffs in Canadian universities, but may be narrowing access for low-income applicants; institutions report role changes and employment growth rather than large-scale displacement, while enrollment patterns suggest a small negative association between AI adoption and low-income student access.

Main Finding

AI adoption in Canada appears to have two primary effects: (1) it is not a major driver of job replacement within Canadian universities (i.e., limited direct displacement of academic and institutional staff), and (2) it is associated with a small but measurable reduction in higher-education access for students from low-income families. Overall, AI has been an important driver of employment growth more broadly, but the study finds little evidence that recent AI advances have substantially raised unemployment in the postsecondary sector.

Key Points

• Research focus: employment dynamics and higher education in Canada, with attention to whether AI drives unemployment and whether AI affects access to higher education for low-income families.
• Hypotheses evaluated (summary):
  • AI adoption contributes to employment growth in Canada.
  • AI advances increase unemployment (overall or in higher education).
  • Postsecondary institutions use AI positively in operations and teaching.
  • AI adoption reduces access to higher education for low-income students.
• Main empirical conclusions:
  • AI use in postsecondary institutions is widespread and generally positive for institutional functions and pedagogy.
  • No strong evidence of large-scale job replacement within Canadian universities attributable to AI (displacement effects are limited or offset by role changes).
  • A modest negative association exists between AI adoption and enrollment/acceptance rates for low-income students, suggesting AI may exacerbate access barriers for vulnerable groups.
• Modeling approach: a Random Forest classifier was used as the primary predictive model to classify outcomes of interest (e.g., enrollment, job displacement signals).
• Policy relevance: results inform workforce planning, financial aid targeting, and institutional AI governance.

Data & Methods

• Data sources: the study relies on secondary data—national and institutional-level datasets—covering employment, higher-education enrollment, and indicators of AI adoption. (Examples of such sources include national labour statistics, institutional administrative records, and publicly available reports; specific datasets were not enumerated here.)
• Unit(s) of analysis: combinations of individuals (students, employees), institutions (postsecondary institutions), and regional labor-market aggregates, depending on the hypothesis tested.
• Outcomes analyzed:
  • Employment outcomes and job-change signals within universities (e.g., separations, role reclassification).
  • Enrollment/access measures for lower-income students (e.g., application, acceptance, matriculation rates).
  • Measures of institutional AI use (administration, teaching tools, automated services).
• Main model:
  • Random Forest classification as the primary predictive algorithm, chosen for robustness to nonlinearity, interactions, and mixed-variable types.
  • Standard supervised-learning workflow (feature engineering, train/test splits or cross-validation, tuning of hyperparameters).
  • Interpretability: feature-importance metrics and partial-dependence analyses were used to assess which covariates (including AI-adoption indicators and income) most strongly predict outcomes.
• Controls and identification caveats:
  • Analyses control for observable covariates (demographics, field of study, institutional size/type, regional labor-market conditions).
  • Causal interpretation is limited by observational secondary data; confounding and selection remain possible. The study frames findings as associations and suggests designs for stronger causal inference in future work (e.g., longitudinal or quasi-experimental approaches).

Implications for AI Economics

• Labor-market dynamics:
  • AI’s net effect on employment is heterogeneous: it can complement some occupations (supporting employment growth) while changing job content in others. In Canadian universities, displacement risk appears limited; policy focus should be on retraining and role redesign rather than large-scale layoffs.
• Human capital and inequality:
  • The modest negative association between AI adoption and low-income students’ access implies that AI could exacerbate educational inequality if left unchecked (digital divides, differential access to preparation, automated admissions tools that incorporate biased proxies).
  • Economists should incorporate technology-driven access mechanisms into models of human-capital accumulation and intergenerational mobility.
• Policy recommendations:
  • Targeted mitigation: expand financial aid, digital-access programs, and bridge programs to prevent AI-driven widening of access gaps.
  • Workforce adaptation: invest in upskilling/reskilling of university staff and redesign of roles to leverage complementary human-AI capabilities.
  • Governance and ethics: develop institutional AI policies addressing fairness, transparency, and bias (especially in admissions or automated advising systems).
• Measurement and research agenda for AI economics:
  • Prioritize longitudinal linked administrative datasets that connect education records to labor-market outcomes to trace long-term effects.
  • Use quasi-experimental methods (e.g., staggered adoption diff-in-diff, instrumental variables) to isolate causal impacts of AI adoption.
  • Examine distributional effects: which subgroups gain or lose, and through what mechanisms (costs, screening tools, preparatory gaps).
  • Incorporate model-based counterfactuals and scenario analysis to project long-run workforce trends under different AI-adoption pathways.
• Ethical and institutional considerations:
  • Monitor algorithmic bias in tools used for admissions, advising, and assessment.
  • Ensure transparency and auditability of AI systems that influence access and employment decisions.

Keywords: artificial intelligence, higher education, employment in Canada, job replacement, low-income families, Random Forest, secondary data, educational access, labor economics.