ChatGPT in Academic Writing
Recently ChatGPT causes a lot of discussions how and if it should be allowed in academic writing. Especially for students this seems to be a shortcut to get an assignment done with very little effort.
Let's review what is the main purpose of an academic paper: "advance the understanding of a particular topic or field, and to help drive further research and innovation" (ChatGPT, 2023).
If your paper achieves this goal by using ChatGPT, then ChatGPT or Prompt Engineering is the research method and should be mentioned and described as such in the "Research Method" section. Unfortunately ChatGPT is not yet very helpful when it comes to producing new insights in a field1 but it can yield good results in a couple of low lever tasks, for example:
- Finding an existing name of a concept you have in mind but cannot name. Using this name you can then conduct further research.
- Summarizing text in context of a research question2 for a quick liturature review.
- Grammar checking and sentence simplification.
If all this helps to achieve the goal of the paper faster or makes it easier to understand, then this new technology should be welcomed with open arms.
If a students asks ChatGPT to write a complete assignment by just providing a topic, this assignment should fail in the review. If not, the review process is flawed in the first place. It is not sufficient anymore to check the punctuation and the citation format. Those times should have ended a long time ago. At the moment, ChatGPT is not good at citing sources and makes up references if asked to do so. Therefore, the validity of the references should be the first thing to check.
Despite the shortcomings when citing sources or coming up with not yet seen insights, ChatGPT can be a gamechanger for students who are looking for a short cut. The strong paraphrasing capablities allow students to simply paste the content of relevant papers into the chat and ask ChatGPT to paraphrase them. Using this tactic students will basically train their own mini model which they feed with the relevant information to produce the assigment automatically.
OpenAI - the creators of ChatGPT - seem to be aware of this fact as they are selling the antidote to the problem they have created (OpenAI, 2023).
OpenAI released a text classifier which try to tell if a text was written by AI or by an Human. Given this article for example - ChatGPT can not says for certain if it was written by a human or an AI (see figure 1).
Would you like to take a guess?... Certain sections were verified using a basic prompt (see figure 2 for different rewriting prompts and their results).
Given two students, how does the amount of "higher thinking" differ when one uses ChatGPT but the other not?
Is it really cheating, or is it using a calculator to solve a complex math problem? While the answer to this question will possibly be obvious in five years, for now, ChatGPT is forcing us to rethink the education system.
The emphasis should be on the education system, as this is the time where students do not create new insights but to learn the basics. In academics, ChatGPT should certenly been seen as a new legit tool.