A Quantitative Revision Method to Improve Usability of Self-and Peer Assessment in MOOCs
With the flourishing of Massive Open Online Courses(MOOCs),a course is no longer limited by any fix times and places.MOOCs enable a learner to access the contents of a world-wide course free from time and locations constrains,which lead to an unimaginable increasing amount of learners.Hence,it becomes a challenging problem in MOOCs that how to track all the students'learning pro-gresses effectively and provide them timely feedbacks.To address this problem,self-and peer assessments have gained considerable momentum recently,as they can involve learners in their learning and the evaluation.However,as different students have different personalities,knowledge levels,senses of responsibility,and grad-ing tactics,there are often many noisy data in the self-and peer grades that reduce their usability for giving an accurate and fair-ness final grade.In this paper,we study a self-and peer assessment dataset about a course with such an undesirable phenomenon that the grade inconsistency distributed over most of the students.To mitigate the observed problems,we propose a quantitative revi-sion method by reducing the grading deviation gradually for all the students.We design an adjustment matrix to revise all the self-and peer grades in a global view and propose an algorithm to ob-tain the adjustment matrix based solely on the given noisy dataset.Moreover,we provide a method to measure the usability of a self-and peer grade dataset.Finally,we evaluate our approach in the collected self-and peer grading dataset.Experiment results show that our approach improves the usability of the obtained dataset at least 33%.
Self-and peer assessment grades usability quantitative revision
Jun Na Yuan Liu
Northeastern University Shenyang,China
国际会议
2019国图灵大会(ACM Turing Celebration conference-China 2019 )
成都
英文
105-110
2019-05-17（万方平台首次上网日期，不代表论文的发表时间）