Intra-group Ranking in Peer Assessment: Mechanism, Reliability, and Fairness

Peer assessment is the only scalable mechanism for detailed feedback in higher education, yet meta-analytic evidence shows peer-teacher rating correlations average only 0.63 (r^2 ≈ 0.40). This calibration problem is particularly acute in standard classrooms (N ≈ 20-50), where sparse connectivity prevents noise from averaging out. Existing solutions bifurcate into pedagogical approaches that ignore aggregation mechanics and MOOC algorithms that assume massive redundancy unavailable in routine courses. Neither addresses the structural constraints of small-classroom topology. This study proposes intra-group ranking as a structural alternative and derives three design principles through Monte Carlo simulation. First, even perfect reviewers face a hard accuracy ceiling imposed by sparse graph topology. Second, the "Rule of 9" identifies r=9 as the optimal review load where marginal gains diminish. Third, a "Heterogeneity Trap" disproportionately misclassifies mid-performers, but increasing structural redundancy neutralizes this inequity. We conclude that ranking excels not as a grading instrument but as a triage mechanism (AUC > 0.82), enabling educators to identify outliers while reserving expert attention for ambiguous cases.