It is demonstrated that the fully automatic generation of isomorphic questions allows for both repeat assessment, and for this assessment to be individualized. While this does require a substantial up-front effort, once prepared assessments can be reproduced with relative ease and with a near-zero probability of students receiving the same question a second time. We show the effectiveness of this approach using survey and performance data obtained from large year 2 and year 3 computer science classes. A greater than 10-fold increase in online peer discussion was observed, compared to the previous year. Contents analysis of the surveys showed that repeat testing was generally regarded favourably. Quantitative analyses found that prior homework did little to improve initial test performance, but was of vital importance in studying for the follow-up isomorphic test. Moreover, the performance gain on the isomorphic questions was the same for all students regardless of their overall ability, and was retained in the final exam.

事实证明，同构问题的全自动生成既可以进行重复评估，又可以使该评估个性化。尽管这确实需要大量的前期努力，但是一旦准备好的评估可以相对轻松地重现，并且学生第二次收到相同问题的可能性几乎为零。我们使用从大型2年级和3年级计算机科学班获得的调查和性能数据来证明这种方法的有效性。与上一年相比，在线同行讨论的数量增长了10倍以上。调查的内容分析表明，通常认为重复测试是有利的。定量分析发现，以前的作业对提高初始测试成绩没有多大作用，但是对于研究后续同构测试至关重要。此外，同构题的成绩提高对所有学生而言都是相同的，而不管他们的整体能力如何，并保留在期末考试中。