Anti-academic Misconduct Initiatives

Academic norms are essential for the community to live healthily. Every post-graduate student must attend a related course before start doing research. Unfortunately, the teaching effect is often not as satisfactory as we expected. We see different cases one after another. My student was also involved in a case that received great attention. I was teaching this course in our computer science college for two years. What an irony? In all these days, I have been investigating this case and re-thinking how to prevent academic misconduct in the future. Unfortunately, simply teaching the student what academic norms are results in very limited effect. Only those people who were intended to follow the academic norms will follow these rules. Is that possible to systematically prevent academic misconduct better will new mechanisms?

The Paper with Plagiarism

On Aug. 16^th, 2021, a friend of mine sent me a Twitter message in which my student was accused of plagiarism in a paper [1] co-authored with another student. Soon later, I also received an email from the original authors who believe their paper [2] has been plagiarized. After carefully checking all the materials I received, as well as related records my students have provided. I believe i) this paper has plagiarized the major contribution from another article; ii) my student knows nothing about plagiarism and did not participate in plagiarism; iii) my student do not have a strong reason to be an author at all as he only provides ideas for improving the paper writing and the experiments.

The first one and the last one are obvious. I believe in the second one, not only because of all the records the student has shown me but also for his previous behaving history. He carefully followed my DOXC initiative [3] to open source all the projects he was leading and put significant efforts to continuedly improving the open-source project, e.g., https://github.com/Res2Net/. I should apologize for not being able to prevent my student from being involved in the plagiarism paper because I’m his supervisor, and I’m teaching the course for academic norms. Soon after I reported this case in my university, the student has received a high punishment for his inappropriate behavior.

This case is an expensive lesson for him to learn. I’m re-thinking why I fail to prevent the scandal. How should I prevent my future students from being involved? Making the rule for authorship clearer is easy. Is it easy to prevent intended and carefully designed plagiarism? If one of the key authors is cheating, all co-authors can easily be kept in the dark. I tried to search Google Scholar using the keywords “Deep Reversible” and “Momentum” today, but I didn’t see the paper [2] with the original idea. I would expect it would be more difficult to find a few weeks after it was initially made publicly available on arXiv.

Three Anti-academic Misconduct Initiatives (AMI)

Although discovering intended plagiarism is difficult. I observed a common feature shared by the plagiarism papers: most of them do not come with source code. None of them comes with full records. Thus, I would like to make the following initiatives to maximally reduce the land for plagiarism.

Open source: Making the source code and data publicly available can save the community lots of time for reproducing the same thing claimed in a paper and give the community more confidence that the paper is highly likely to deserve trust. Thus, the first initiative is open source.

Trustable Evidence: Concurrent work can easily be made clear if there is trustable evidence showing that the main idea of this paper comes earlier than the concurrent research was made publicly available. The second initiative is to keep trustable evidence to avoid indiscernible concurrent work. Any famous third-party repository with fully historical time stamps can be strong evidence, e.g., Overleaf and GitHub.

Safe co-authors: During my early career, if I recommend collaborating with someone unfamiliar to my supervisor, he would ask me “how much do you know about this guy?”. After carefully thinking these days, I’m so worried about my future students’ ability to select safe co-authors. I must teach them a simple rule. Thus, the third initiative is to only co-author with those researchers who made their paper code publicly available and don’t have intended bad reputation records.

Conclusions

Plagiarism happens much more often than we would expect. While almost everyone hates it, we should build systematic rules to prevent them. People should not fall twice in the same place. By the AMI initiatives, I believe academic misconduct could be significantly reduced.

[1] Li, Duo, and Shang-Hua Gao. “m-RevNet: Deep Reversible Neural Networks with Momentum.” arXiv preprint arXiv:2108.05862 (2021).

[2] Sander, Michael E., Pierre Ablin, Mathieu Blondel, and Gabriel Peyré. “Momentum residual neural networks.”, ICML, 2021.

[3] DOCX: Demo, Open source, Chinese version, and eXplain initiative, Ming-Ming Cheng, Oct. 2020, URL: https://mmcheng.net/docx/.