Answers are grounded in this paper and cite their sources.
Examination script vetting, as the post-marking review of answer scripts by independent academic staff, is a critical quality assurance process in tertiary institutions. Its allocation to eligible lecturers is typically performed manually in many institutions by coordinators without algorithmic support, resulting in inequitable workload distributions and frequent constraint violations. To the best of our knowledge, no formal computational model exists for the Examination Script Vetting Allocation Problem (ESVAP), leaving institutions without principled, automated allocation support. This study aimed to design and implement a conflict-aware load-balancing memetic algorithm (CALBMA) for optimal examination script vetting allocation and deploy it within a web-based system. The ESVAP is formally modelled as a Binary Quadratic Program with four hard constraints (C1–C3 and C5), including the novel single-program-per-envelope constraint (C5). CALBMA hybridises a genetic algorithm with a constrained swap-based local search. It was implemented in Python and benchmarked against greedy and random allocation baselines on three synthetic datasets using five metrics: Jain's fairness index, within-rank workload variance (WRV), constraint violation rate (CVR), computation time, and Solution Quality Score. CALBMA achieved a CVR of 0 across all datasets, a 34.4-fold WRV reduction over greedy on the medium dataset (2.50 vs. 86.11), and statistically significant superiority confirmed by the Wilcoxon signed-rank test ( p = 0.000002). Evaluation relied solely on synthetic data because no public ESVAP benchmark exists, and real institutional vetting records require institutional approval and data-sharing agreements beyond the scope of this study; the web prototype lacks enterprise-grade security. Institutions should adopt algorithm-driven allocation to eliminate inequity, while future work should extend CALBMA to multi-objective formulations and validate it on real institutional data.
No takes yet. Share an insight, caveat, or question.
Osofuye et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: