Efforts to safeguard academic integrity and fair assessment received a new challenge in November 2022, with ChatGPT, artificial intelligence entering the market in a usable form as a chatbot, which no one at educational institutions had expected until recently. I still remember Elisabeth Bik’s wonderful keynote at the European Conference on Academic Integrity and Plagiarism (ECAIP 2022) in Porto in the spring of 2022 and her examples of ‘tortured’ paraphrases and weird synonyms, which were repeatedly copied unreflectively from automatic tools into academic texts. This regularly caused a laugh among readers of such texts. Computer programmes that produced grammatically correct and reasonably meaningful sentences for academic use were not available to the wider public until recently.
That has changed with ChatGPT. The bot produces grammatically correct sentences with correct spelling and delivers seemingly plausible content, at least superficially. The result is hardly distinguishable from superficial texts written by humans, and ChatGPT can thus pass exams.
Educational institutions react to this in very different ways. Some educators bury their heads in the sand and do not tell students ChatGPT exists. Others show panic reactions because now we have to change everything.
Term papers, theses, plagiarism – all yesterday’s news and no longer relevant. We can continue to extend this list.
Does that really change everything now? I think not. First, the good news.
The vast majority of students do not spend all day thinking about how to cheat and get their degree without any increase in knowledge. There are actually some who are interested in their field of study.
This majority will not cause any problems in the future and we should support them in their motivation and development. Nevertheless, there was and always will be a small part of the student community to whom this does not apply. This part is, of course, far too large and has now been given another option in addition to contract cheating services or plagiarism. There are many reasons why someone cheats. For example, incompetence, lack of time, unwillingness, laziness, pressure to perform, career lust, etc.
Tools like ChatGPT used for cheating have much in common with the previous possibilities.
Just like a perfectly disguised plagiarism, a ChatGPT text is actually undetectable after several revision phases in which one specifically asks back with original research questions, gives qualified input, incorporates suitable literature and sources and asks for linguistic revision in the right places.
It is just that this requires expertise, methodological competence, time and interest in the topic of the text. To exaggerate, one could say that if the learning goal has already been achieved before the exam, then it is no longer so dramatic if the use of ChatGPT is not detected in the process.
However, experience with plagiarism shows that it is very often created by people who have no time, no motivation, no competence or no interest in the topic. This is then noticeable when reading.
ChatGPT is currently very good at inventing plausible-sounding references and sources and also likes to present false facts with great self-confidence that are only apparently plausible. In addition, ChatGPT does not know what happened in class, unless someone tells it.
If the user does not put any energy into revising the generated text, due to lack of time or competence, the chance of ‘stupid’ errors in the text is quite high. As has been the case up to now, assigning written papers is therefore only useful if they are also accompanied, read and provided with feedback by the teachers. The large number of fraud attempts that simulate competence then have a good chance of being discovered because they often contain — sometimes absurd — computer-generated errors with regard to content. So not everything does change after all.
Technical tools like this are currently freely available in the world. As with calculators or computers, soon, no one will want to do without them and delegate certain writing tasks to the AI on a regular basis.
ChatGPT and its kind must therefore be didactically integrated into teaching. There are scenarios for developing text and methodological skills with the help of chatbots.
For example, students can analyse machine-generated text in class and compare it with texts written by researchers. My wish would be that they would soon be used on a large scale.
About the author:
Dr Oliver Trevisiol is a subject librarian for history, education and theology at the Communication, Information, Media Centre (KIM) at the University of Konstanz (Germany)
Contact: oliver.trevisiol@uni-konstanz.de