Good academic practice
Good academic practice
There are various norms and practices when it comes to documenting and citing use of generative AI. While otherwise complying with the norms of good academic practice, plagiarism rules, etc., the overriding principle is to describe/document clearly where, when, and for what purpose you have used generative AI. Transparency is a key criterion when it comes to ensuring the integrity of your work. This is in line with The Danish Code of Conduct for Research Integrity that advocates honesty, transparency, and accountability.
Furthermore, the above is in accordance with 'The European Code of Conduct for Research Integrity' from Allea, which was revised in 2023.
Document your use of generative AI
Any use of generative AI must be documented clearly, and possibly be attached to your assignment. This is, of course, assuming that it is an allowed aid; if not, it falls in to the category of cheating!
It is essential to capture all relevant information. Your documentation should include the following key elements:
- The platform's name and, if applicable, a version number or similar, as well as the date of the training data, if available.
- Input: What you fed into the model.
- Output: What the model responded with.
- If possible, generate a public link to the specific dialogue and include it. Note that this link may become invalid at some point if there are technical changes.
- When you create multiple prompts with input/output, make sure to include the entire history in your documentation. Output often depends on the context of previous input/output.
- Pay attention to whether you asked the platform to generate output multiple times. If so, include all outputs, and consider explaining in your attachment, or assignment, why a particular output was not sufficient in your context.
You can use copy/paste or make screenshots to document use.
Generative AI as method
If you use generative AI to generate ideas, create content etc. (where it does not violate any rules), you can state your writing e.g.:
'During the creation of the interview questionnaire, ChatGPT version 3.5 was used to qualify and identify relevant questions. Documentation of use, including prompts and output, can be found in Appendix 1.'
Additionally, it is always relevant to justify why the changes, additions, etc., supplemented by generative AI, are correct and valuable (based on your theories and expertise). Also, be aware if your academic programme has guidelines and/or rules that explicitly exclude the use of generative AI. This is often based on the specific learning objectives of a given course.
Can generative AI be a cited source?
Academic practice and integrity rest on the recognition and reference to the works of others. This applies not only to what is expressed in writing but also to ideas, results, data, code, etc. The Danish Code of Conduct for Research Integrity states, 'When using one’s own work and the work of other researchers in a publication, appropriate and accurate references to such work should be provided.' The IA language models always rest on the work of others. Therefore, you must to support you AI reference with a reference to the original source.
For example, a statement like the following would not be sufficient or in accordance with good academic practice:
'According to ChatGPT, Anthony Giddens believes that ...' unless your assignment is specifically about investigating ChatGPT's perception. It is essential to consider what is considered good academic argumentation in this context.
Avoid plagiarism
Generative AI's language models are built on the work of others, as the model has no knowledge without having accessed data from books, articles, Reddit forums, etc.
Without references to the work of others, any statement, result, image etc., will appear as a result of your work if you reproduce it either verbatim or in a paraphrased/rewritten form. Broadly speaking, you are committing plagiarism if you present others' work or ideas as your own! On the other hand, when you refer to the work of others, you acknowledge their efforts, and your citations serve as links between articles, books, data, code, etc., allowing others to trace who is referring to whom.
Furthermore, openly recognizing the work and influence of others through correct referencing also ensures the transparency of your own work. Your references will enable others to verify the sources of your work, to appreciate thoroughness and academic outlook, and to track potential errors in, for example, the source's conclusions. References ensures traceability between your work and the original source.
If you use generative AI to form the background of your approach to a subject and cite ChatGPT, you will be going against current academic norms and practices!
Cite generative AI
When you ask ChatGPT with the prompt: 'Should I always cite generative AI in an assignment, or can I use text without citing it?' part of the answer contains the following passage, which is in line with other recommendations for practice:
If you use AI-generated text as a source to gain insights or ideas, you should probably cite the AI model to show where the information is coming from. However, if you use generative AI to generate text as part of your own thought process or brainstorming without relying on or reproducing specific details, it may be less necessary to cite the AI model.
The best advice is to be clear and explicit always about your use of generative AI - better safe than sorry.
Make correct references to use of generative AI
If you need to reproduce a statement in text that is a direct quote from generative AI, you can use one of these reference styles that matches what you are already using in your text.
With regards to images, you should likewise provide information about the platform, prompt, date, and the license associated with the image, for example, CC0.