How automated feedback through text mining changes plagiaristic behavior in online assignments

This study is intended to decrease learners' plagiaristic behaviour in online assignments by providing automated feedback based on text mining analysis. Document similarity analysis was done at the middle and end of the semester on 4268 reflection texts (max. 500 characters) written by the participating university students (N 1⁄4 59) about concepts which they had learnt in the computer science course. In the analysis conducted in the middle of the semester, the plagiarism ratios in the posts that students wrote during the first six weeks were calculated, and these ratios were used to give the students feedback. The analysis conducted at the end of the semester was used to test whether there was a change in the plagiaristic behaviours of the students after they had received feedback. To calculate the plagiarism ratio, the students' posts were compared to internet search results and posts written by other students. A paired samples t-test was used to identify whether there was a difference in the plagiarised post ratios before and after feedback was provided. In addition, the plagiaristic behaviours of the students before and after feedback were compared proportionally using the McNemar's Chi-square test. The results of the analyses revealed a statistically significant difference in the plagiarised post ratios and the ratios of students performing plagiaristic behaviour before and after feedback. It was found that after feedback was provided, while the ratio of plagiarised posts went down 21.07% on average, 83% of the students had lower plagiarised post ratios. The number of students who did not commit plagiarism increased by 42.37% after feedback was provided. © 2015 Elsevier Ltd. All rights reserved.