Interestingly, one of the most searched questions about ChatGPT on the Internet is whether it poses a risk as a tool for plagiarism.
We interacted with ChatGPT to discuss this issue in depth, and here are the key takeaways from our chit-chat:
- It makes plagiarism “easier to conduct”: For example, a full essay can be generated in seconds with the push of a button.
- It also makes it “harder to detect”: It becomes more difficult to tell plagiarised pieces apart from original ones.
- These factors can lead to an increase in plagiarism and cause damage to honest students and professionals who put effort into producing original content, products, etc.
- Software trained on machine-generated text can be a potential solution to these issues.
In the short video below, you can follow ChatGPT as it responds to our query about it being used as a tool for plagiarism. In the rest of the post, we discuss both this video and the key takeaways highlighted above in detail.
The risks of ChatGPT being used as a tool for plagiarism
To understand the potential risk ChatGPT poses as a tool for plagiarism, we entered the following prompt:
Write me a 400-word blog post on the risks of ChatGPT being used as a tool for plagiarism.
In fact, we asked the same question twice. The short video above was based on the second response. We failed to record the first response but, luckily, have the screenshots. The discussion below is based on both responses provided.
In the two posts generated, ChatGPT recognizes that its ability to “generate text that is indistinguishable from human writing” can give motivation to people to use it as a tool for plagiarism and gives the following examples:
A student could use the model to generate an entire essay, and then turn it in as their own work. Or, a professional could use the model to generate a report or article, and then present it as their own original work. (see Figure 1)
With ChatGPT, plagiarism becomes “easier to conduct”
It admits it can facilitate plagiarism by making it easier for people to “pass off someone else’s work as their own.” All you need to do is type in your query and hit the submit button! In seconds, you get a well-prepared response that reads like it’s written by a real human.
As a result of this “ease” of the act, it conjectures that:
This could lead to an increase in plagiarism in the academic and professional worlds. (See Figure 2)
It also becomes “harder to detect”
Advanced language models such as ChatGPT can mimic human writing in their responses and can generate original answers. This makes it harder for traditional plagiarism detection software to flag work content generated by them as the former focus on the matched or similar text. A natural consequence of this is that more cases of plagiarism would go undetected, making it easier for individuals to “get away with it.” An unwanted side effect of this is that:
This would be particularly damaging in the academic world, where students’ hard work and original research is what sets them apart from their peers, and gives them a competitive edge in the job market. (See Figure 3)
Potential Solution: Advanced plagiarism detection software
It is common for organizations such as universities to use specialized software (e.g., Turnitin, Quetext, Unicheck, etc.) to detect cases of plagiarism. An interesting recommendation is that such software can be improved to tackle the increased risk of plagiarism posed by language models:
… use plagiarism detection software that is specifically designed to detect text generated by language models. These tools use advanced algorithms to analyze text and identify patterns that are characteristic of machine-generated text… (See Figure 4)
In this post, we have offered a detailed discussion on “ChatGPT and plagiarism”. It is clear that students or professionals can easily be tempted to use language models or artificial intelligence in general as a shortcut in their studies or tasks. Does it mean that we should abandon these helpful AI tools? The answer seems to be no. A potential solution lies ahead where software for plagiarism detection can be trained to recognize content created by such tools and help tackle this issue.