Institutional Structures, Digital Inequality, and AI Integration in Higher Education

Synthesis: Adeniranye et al. (2026) provide one of the first systematic comparative assessments of AI integration across institution types within a single national higher education system. Analyzing 45 Nigerian universities (15 federal, 15 state, 15 private) across six dimensions — infrastructure, curriculum, research, industry partnerships, international collaborations, and policy frameworks — they find only moderate overall integration (M = 4.79 on a 10-point scale) with meaningful variation. Critically, governance type did not predict integration once institution age and geographic location were controlled: older universities and those in Nigeria's South-West corridor scored significantly higher. Strong correlations among internal capabilities and between international collaborations and industry partnerships show that capability and network ties are mutually reinforcing, so well-connected institutions accumulate compounding advantages while others risk deepening digital inequality.

Key Findings

1. Moderate overall integration with wide variation. Across 45 universities the mean AI integration was M = 4.79 (SD = 1.71), range 1.83–7.83 on a 10-point scale — a baseline of substantial but uneven adoption.
2. Institution type does not determine integration. Federal universities had the highest mean (M = 5.25), followed by private (M = 4.76) and state (M = 4.36), but one-way ANOVA found no significant overall difference by type (F(2,42) = 1.01, p = 0.372).
3. Age and geography, not governance, predict integration. Institution age was the strongest predictor (β = 0.43, p = 0.016) and South-West location was also significant (β = 0.31, p = 0.029); institution type was not significant when controlling for these (full model R² = 0.30, F(4,40) = 4.29, p = 0.006).
4. Curriculum is the strongest dimension. Curriculum integration led overall (M = 5.73), with private universities recording their strongest score there (M = 5.87); policy frameworks were weakest overall (M = 4.09), with only 27% of institutions scoring 6+ on formal AI strategies.
5. Internal capabilities reinforce each other. Infrastructure, curriculum, and research intercorrelated strongly (r = 0.79–0.80, all p < 0.001), indicating mutually reinforcing institutional capacity.
6. Network position compounds advantage. The strongest external relationship — international collaborations × industry partnerships (r = 0.74, p < 0.001) — supports a "connections beget connections" dynamic in which globally connected institutions accrue prestige, funding, and further ties.
7. Policy lags practice. Policy frameworks showed the weakest correlation with other dimensions (notably curriculum, r = 0.43), signaling a disconnect between formal AI strategy and operational curriculum activity.

Methods in Brief

Three trained researchers conducted a comparative content analysis of public university websites, strategic plans, course catalogs, and partnership announcements for 45 Nigerian universities (equal-allocation stratified sample of 15 each across federal/state/private). Each institution was scored on six theoretically grounded dimensions using structured 10-point anchored rating scales, refined through pilot coding; the overall score is the unweighted mean. Analysis used one-way ANOVA, Pearson correlation (Bonferroni-corrected), and multiple regression. Inter-rater reliability on 20% of the sample was strong (Cohen's κ = 0.79–0.85), and external validation rates varied by source type (academic publications 78%, news 67%). The integrated framework combined institutional theory, the resource-based view, and network theory.

What this means for practice

• Administrators. Audit your institution against all six dimensions before setting AI priorities: curriculum led at M = 5.73 while policy frameworks trailed at M = 4.09, so the weakest area is usually formal strategy, not the teaching program.
• Institutions. Fund infrastructure, faculty AI Pedagogies and Teaching Strategies training, and research support as one package, since those three internal capabilities intercorrelate at r = 0.79–0.80 and single-dimension investment leaves the others unsupported.
• Policymakers. Target benchmarking and incentive mechanisms at newer and non-South-West universities, because institution age (β = 0.43, p = 0.016) and South-West location (β = 0.31, p = 0.029) — not governance type — predicted integration.
• Researchers. Report the evidential basis for every score when adapting the instrument for another system, since external validation rates differ by source type (academic publications 78%, news 67%).
• Partnership leads. Treat international and industry partnerships as complementary rather than competing, and choose partners by history, geography, networks and dimensional strengths rather than by institutional category; deliberately include less-connected institutions to counter compounding advantage, and diversify funding to avoid the dependency risk the authors tie to the 2025 USAID funding freeze.

Limitations

• The study relies on publicly observable documentary and web indicators, which may capture institutional signaling alongside practice and may undercount institutions with a limited web presence.
• The equal-allocation sample is not proportionally representative of the wider system.
• A modest N = 45 limits statistical power and yields wide confidence intervals, so the reported relationships are associations rather than causal estimates.
• The single cross-sectional time point cannot track the pace at which AI adoption is changing.

Citation

Adeniranye, D. I., Lunn, S. J., Mosobalaje, O., Eze, P., Berhane, B., & Adeniyi, D. (2026). Institutional structures, digital inequality, and AI integration in higher education. Computers and Education Open, 11, 100400.