ChatGPT took the world by storm in the late end of 2022. Millions were using the tool within a week, and by now, it has garnered hundreds of millions of active users.
There are possibly hundreds of stereotypes surrounding ChatGPT, but the largest and most influential are those who claim ChatGPT is allowing students to get by in their English classes without actually writing any essays. Many school districts responded to these claims by immediately banning the tool, denouncing it as plagiarism and cheating.
The concerns around such cheating are well-placed. ChatGPT is known to produce extremely well essays in a short time period. A well-constructed prompt can deliver C, or even B-level work in seconds. Beyond just writing essays, ChatGPT can give users information about practically any topic it is asked about, from the history of Rome to the intricate details of string theory.
However, a research study conducted at Stanford has shown that cheating has hardly changed ever since the drop of ChatGPT, and perhaps has even gone down. The study found that before ChatGPT was released, around 70% of students had participated in some form of cheating over a one-month period. These forms of cheating ranged anywhere from cheating on a final exam to copying an answer or two from a friend. After ChatGPT was released, that figure remained the same, even slightly dropping to be closer to 60%. The only thing that ChatGPT has done for students is make cheating faster and/or easier. It has had no part in increasing the number of cheating students. A student who didn’t cheat before ChatGPT was around isn’t really likely to change their ways simply because ChatGPT is there.
Many people have come to take ChatGPT’s word as fact, something that can never be wrong. But these assumptions are proven to be incorrect. According to Josh Howarth – a senior developer with a Masters in Computing Science – ChatGPT, depending on the topic you are asking about, has an error rate of about 12%. Obviously, simpler, fact-based questions will have much higher accuracy, extremely close to what a human would find from research. However, when ChatGPT is asked about novels or other related topics, its accuracy drops tremendously. It has a tendency to fabricate information for no reason, where a student would be able to give a reasonable answer in seconds.
These sorts of inaccuracies are extremely easy to fact check with small questions, but it becomes much more sinister when regarding large essays. In fact, in early 2023, a lawyer used ChatGPT to write their argument. His case cited previous cases related to his client. However, the cases that ChatGPT cited were entirely fabricated. This is just one example of the multitude of mistakes that ChatGPT makes.
Despite its inaccuracies, ChatGPT is well known for its ability to summarize text that you directly feed it. Many students have used this to quickly scan through long texts. Unlike depending on it to research your topic for you, ChatGPT is much less likely to make incorrect claims in its response, making it very good at summarizing. It’s main weakness is its inability to fact-check itself, it is also good at both editing and creating templates.
ChatGPT shouldn’t be seen as an evil tool for cheating. Yes, it makes cheating much easier, but we have already seen that students often shy away from cheating with ChatGPT anyway. Instead of villainizing it, we should be promoting it as a tool to be used. ChatGPT makes the text easy to digest, it can help you get a start on your work, and it can even point out mistakes in writing. It makes information more accessible.
But these benefits can only be reaped if we use ChatGPT properly. If students continue to cheat with ChatGPT, to use it to write their essays, the stereotypes will only grow stronger surrounding it. If we want the privilege of being able to use ChatGPT, we have to prove that we can use it correctly.
