- 期刊
大學申請入學綜合學習表現分組評量的計分法誤差
Method Errors in Scoring Holistic Learning for University Admissions
摘要
申請入學管道已成為大學最主要的入學管道。該管道第2階段學生資料審查是一場大型的分組表現性評量，必須面對表現性評量自帶的評分者偏誤問題和分組評分的等化問題。評分者偏誤歷來已有許多文獻探討，分組評分的等化做法探討則較為少見；然而，它卻是非常實際而且必要的。因為不同的算法，受評者的總分和排名差異巨大，影響了公平性和管道的信任度。本研究探討不同計分法在大型分組表現性評量下，受評者真實表現可能受到的扭曲情形，藉此了解計分法的特質、效能及特性。本研究依據申請入學管道的招生情境，建構了統計模型，透過多個變項的參數控制，分析5種常見的計分法，發現了以往未曾發現的「計分法分組系統性誤差」的存在，另外亦解析出干擾計分法效能的因素以及其對申請入學者真實表現的扭曲程度，整理出「計分法對影響因素的敏感度」。研究成果除可提供招生單位參考外，亦補充了申請入學機制中研究者較少論及的一隅，未來或可藉此延伸發展出更佳的等化計分法，降低誤差的發生。
並列摘要
The Application for Admission Channel has become the primary admission method for universities in Taiwan. In the second stage, students' high school Holistic Learning reviews, which inherently involve large-scale performance assessments, necessitate the separation of ratees into different groups. This stage faces challenges of rater bias and the equalization of different group scores. While rater bias has been widely discussed, group score equalization methods have not received as much attention. However, the study of scoring methods is crucial as it addresses scoring biases across groups, impacting total scores, rankings, and the overall fairness and trustworthiness of the admission system. This study investigates the potential distortions in the real grades of ratees' performances by examining five commonly used scoring methods by colleges. It identifies the presence of systematic errors in scoring methods for grouping performance assessments. Factors affecting the performance of scoring methods are classified under the "Sensitivity of scoring methods to influencing factors." This research provides valuable insights for admissions units and addresses a neglected aspect of the Application for Admission mechanism, potentially laying the groundwork for improved group score equalization methods.