arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2212.13774cs.GT
收藏

交互式迭代同伴评估

Interactive and Iterative Peer Assessment

Lihi Dery

AI总结:

针对传统同伴评分存在分数膨胀、并列多、策略性打分及全排序认知负担重的问题,提出含引导评价算法与中位数投票聚合协议的新型同伴评分模型,经大学课程部署验证,减少了排名并列并降低了学生负担。

摘要:

迭代式同伴评分活动可让学生在课堂项目展示期间保持参与度。收集并聚合同伴评估数据的有效方法至关重要。学生往往会给项目打偏高的分数,因此,要求学生给出数字分数是常见做法,但这通常会导致所有项目的分数都偏高,造成大量最高分并列的情况。此外,学生可能会策略性地给其他人的项目打更低的分数,好让自己的项目更突出。另一种方式是要求学生将所有项目从最好到最差排序,但由于人类认知能力的限制,这种方式存在挑战。为解决这些问题,我们提出了一种新型同伴评分模型,该模型包含两部分:(a)一种用于引导学生给出评价的算法,以及(b)一种基于中位数的投票协议,用于将分数聚合为单一排名顺序,以减少并列情况。基于我们模型的应用程序已在一门大学课程中部署并测试,结果显示备选方案之间的并列情况更少,且学生的认知负担和沟通负担显著降低。

英文摘要:

Iterative peer grading activities may keep students engaged during in-class project presentations. Effective methods for collecting and aggregating peer assessment data are essential. Students tend to grade projects favorably. So, while asking students for numeric grades is a common approach, it often leads to inflated grades across all projects, resulting in numerous ties for the top grades. Additionally, students may strategically assign lower grades to others' projects so that their projects will shine. Alternatively, requesting students to rank all projects from best to worst presents challenges due to limitations in human cognitive capacity. To address these issues, we propose a novel peer grading model consisting of (a) an algorithm designed to elicit student evaluations and (b) a median-based voting protocol for aggregating grades to a single ranked order that reduces ties. An application based on our model was deployed and tested in a university course, demonstrating fewer ties between alternatives and a significant decrease in students' cognitive and communication burdens.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑