Diversity regularization λ = 0.30 was selected from a sensitivity analysis as balancing recommendation relevance and content diversity¶

Claim · All claims
Evidence · 1 study · associational r1 · q1
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? A sensitivity analysis over λ ∈ {0.10, 0.20, 0.30, 0.40, 0.50} selected λ = 0.30, which maximized intra-list diversity while keeping mean reciprocal affinity rank at 1.4 and intra-list cosine similarity at 0.41. → Najem 2026

Evidence¶

Najem 2026¶

Najem, K., Zaoui Seghroucheni, Y., & Ziti, S. (2026). Behavioral clustering for adaptive learning object recommendation in higher education using learning analytics for personalized student success. Discover Education. https://doi.org/10.1007/s44217-026-02181-7

q1 · i? · associational · r1

Numerical sensitivity analysis of the diversity regularization parameter over five λ values (Table 6); λ = 0.30 gave mean reciprocal affinity rank 1.4, intra-list cosine similarity 0.41, and 3 content types covered, versus rank 1.2 and similarity 0.92 at λ = 0.00.

"The value λ = 0.30 was selected because it maximized intra-list diversity (measured by average pairwise cosine distance among the top-5 recommendations) while maintaining a favorable balance between recommendation relevance (mean reciprocal affinity rank = 1.4) and content diversity (intra-list cosine similarity = 0.41, covering 3 content types)."

Discussion¶

Related Claims¶

• Classification threshold trades sensitivity against specificity: 0.30 yields 91.3% sensitivity, 0.60 yields 89.7% specificity — related
• The verification threshold trades hallucination reduction against over-refusal, with the default tau of 0.20 providing the best balance — related
• Research-question responses score higher Answer Relevancy (0.87) than follow-ups (0.72), while Faithfulness, Context Relevance and Response Groundedness are lower — related