美国数学建模竞赛特等奖论文——1996 B O Judging a Mathematics Contest【数学建模】.pdfVIP

Judging a Mathematics Contest 309 Judging a Mathematics Contest Daniel A. Calder´on Brennan Philip J. Darcy David T. Tascione St. Bonaventure University St. Bonaventure, NY 14778 Advisor: Albert G. White Overview Our model is based on breaking the problem down into four main areas and dealing with each: the distribution of papers among the judges, scoring methods, the number of papers to eliminate per round and the number of rounds, and the performance of the model with larger numbers of papers. In each component, we focused on the goals of maintaining fairness and variety in all judging procedures, eliminating as many papers as possible in each round, minimizing the number of rounds, and, most important, seeing that no one goal was attained at the expense of any other. The papers were coded and sent on from judge to judge without any prior knowledge of that particular participant or paper. There is no way to avoid having a judge read a paper twice, but a judge will not read a paper twice in a row until perhaps the final two rounds. Assumptions Budget constraints affect only the number of judges. Time constraints affect only the number of papers that each judge can read. An approximate “absolute” ranking system exists among the judges; i.e., if every paper were to be scored or ranked by each judge, the results of each judge would generally agree with every other (allowing for a few places where consecutive papers may be “flip-flopped”). All papers are eligible to win (none disqualified for cheating, missing sec- tions, etc.). Judges need not be in the same location, but a copy of each paper (electronic or hard copy) is readily available to each judge. Judges remain ignorant of other judges’ opinions on all papers. 310 The UMAP Journal 17.3 There is no way to avoid having a judge read a paper twice (nonconsecu-