Adaptive AHP-Entropy Combined Weighting Model for Teachers’ Digital Literacy Assessment

To address the widespread reliance on equal-weight assumptions in the assessment of teachers’ digital literacy and the lack of a mechanism for diagnosing disagreements between subjective and objective weighting, this study proposes a conflict-aware and adaptively balanced combined weighting model (CAW-AE). The model jointly measures structural conflicts between subjective and objective weights using a geometrically weighted combination of cosine similarity and rank displacement. Weighting states are classified into low-, medium-, and high-conflict intervals according to the degree of conflict. Differentiated indicator-level reconciliation coefficients incorporating a conflict penalty are then constructed. Their reconciliation intensity is determined by an analytical solution that minimizes overall deviation and is subject to truncation constraints ensuring a convex combination. Empirical results based on assessment data from 347 teachers and 20 indicators show a global conflict degree of 0.271, with conflict concentrated in digital social responsibility. CAW-AE has a Herfindahl–Hirschman index of 0.066 and a coefficient of variation of 0.334, occupying a justified position between the individual methods and a fixed-coefficient compromise. Tests show a rank reversal rate of 8.2% under ±20% perturbations of expert judgments, lower than the 14.8% of AHP alone and the 9.7% of game-theoretic combination. Its group discrimination (Kruskal-Wallis H = 33.76) is the highest among the six schemes. Its rank correlation with the external criterion is 0.428, significantly higher than those of equal weighting and entropy weighting alone. These results indicate that CAW-AE improves ranking stability and differentiation between groups without sacrificing the reasonable contribution of domain knowledge.