The relationship between AI and plagiarism is arguably ambiguous as AI-generated text isn't technically authored. Heyward (2023) cites a student who argues that using AI software to write essays isn't "cheating," as companies use it in "the real world" to create content. The student, who believes they have been responsible for their own learning, sees AI software as a tool to level the playing field because they "don’t write eloquently" (Heyward, 2023). Dehouche suggests that AI-generated text be treated similarly to "public domain texts and publicly funded research... belonging to the public" (2023, p.21-22), adding that its use should be fully disclosed "along with the prompts and additional training data submitted to it" (p.22). Cambridge University Press recently issued a statement prohibiting the use of AI as an "author," adding that any use of AI to generate text must be fully disclosed (2023).
Although use of AI-generated text in essays may not constitute plagiarism per se, most agree that its undisclosed use constitutes academic dishonesty. The extent to which it may be allowed in student writing is still open to debate. The University of Colorado offers some sample syllabus statements that range from prohibiting all use of AI, allowing it with permission from the teacher, or even taking it for granted that students will be using AI and evaluating them accordingly. This last statement is especially provocative and may sound like "giving up." However, it would free instructors from the responsibility of endlessly "policing" student writing, and could also lead to critical discussions about how to use AI well.
OpenAI offers the following as an example of the type of statement that could be added to a piece of writing that was created with the help of an AI:
"The author generated this text in part with GPT-3, OpenAI’s large-scale language-generation model. Upon generating draft language, the author reviewed, edited, and revised the language to their own liking and takes ultimate responsibility for the content of this publication."
Identifying the use of AI-generated text can be harder than spotting plagiarism. There is currently no resource as reliable for spotting AI-generated text as turnitin.com is for spotting plagiarism, although the company says they are in the process of creating something. Programs exist that can detect AI-generated text with some reliability, although they can produce false positives and do not always agree with each other.
Confronting a student with suspicions about AI use is thus more difficult than with plagiarism, as there is usually no indisputable source to point to as evidence. Savvy students may even be aware of these limitations and use them to their advantage. Triangulation through mutliple detectors can certainly help. As can frank discussion about the contents of a paper and the tendencies of AI-generated text. The best advice is to design assignments so that AI cannot be easily used as a substitute for work in the first place.
Below are some articles describing what to look for as well as AI detection programs that have proven effective and were recommended by reliable sources. Note that GPT itself can be used to identify text that it generated, although even that is not necessarily reliable. In general, here are some things to watch out for: