ChatGPT does more harm than good to learning
Photo Credit: Ilya Pavlov @ UnplashAs a former student, I understand how AI-generated text applications like ChatGPT can assist students. ChatGPT can help students quickly brainstorm ideas, find examples for their arguments, and even draft essays. I tested ChatGPT with old questions from my MBA degree at UTAS. In multiple choices questions, the results were hit and missed. However, it excelled at essay questions like "discuss the pros and cons of entrepreneurship" or "list and explain the types of monopolies". These questions were typically worth a lot more points. It could even analyse short-case facts to respond to short-answer questions. There were some made-up facts here and there, but the overall answers were passable. I achieved a 6.9/7 GPA, so I know what passable solutions look like.
However, the adverse effects of using such an application outweigh its benefits. To obtain such a high GPA, I spent months researching journal articles, drafting and redrafting my essays, absorbing knowledge and honing my writing skills simultaneously. ChatGPT could speed up this process, but it also takes away the opportunities to consolidate knowledge, which is a crucial part of learning. In addition, relying on ChatGPT discourage students from practising researching for information, taking the info spilled out from the machine as granted. Not only does this harm students' ability to fact-check information but also their ability to articulate ideas. Last time I checked, communication skills are in almost every job advertisement. Another critical point of using ChatGPT or similar apps is the potential for academic dishonesty - presenting others' works as your own. It is too simple to copy the complete responses from the AI, throw in a random reference and call it a day. The worst part is that I see situations like this can easily occur.
All factors in the triangle of fraud are present with ChatGPT. For those unfamiliar with this concept, the fraud triangle (Cressey 1953) explains three factors that encourage someone to commit fraud. They are opportunity, rationalisation, and motive/ pressure. With most take-home essays, there is no supervision on how a student may complete them. There is a clear opportunity for students to access ChatGPT and start typing prompt keywords. Even with a ban on IP, a VPN (Virtual Private Network) could be a workaround. OpenAI, the company behind ChatGPT, will put its product behind a paywall. However, a group of students can share the subscription fee. There are several rationalisations students convince themselves for using such a tool. It could be 'I just do it this time and do real study next time', 'no one will know', or 'this is not an interesting subject, I do not want to spend time on it'. There are also numerous motives for a student who wants to use ChatGPT instead of actual learning. It could be time pressure, fear of a low GPA or the fact that they know someone else is using the tool.
Photo Credit: Andy Kelly @ UnplashPreventing students from using ChatGPT unethically will incur costs to universities and other higher education organisations. Further investment in AI-text-detecting software is needed. Multiple choice test banks must be frequently refreshed to avoid feeding into the algorithm behind the AI tools. Test banks will take time and money to develop or purchase. Furthermore, a change in the assessment methods may be necessary. More oral tests should be implemented, replacing traditional written assignments. This strategy will require more human resources, leading to increased salary costs. These costs are justifiable, though, for maintaining the quality of graduates.
Reference
Cressey, R (1953), Other people's money: a study in the social psychology of embezzlement, Glencoe, IL: The Free Press.