How do I use (generative) AI responsibly?

Working in an academic environment, be it as student or as researcher, you have to use the tools available to you responsibly. Generative AI is no different. Hence, there are a few points you should pay attention to when deciding whether to use a certain AI tool in your academic assignments.

1. Check your course guide

First check your course guide to see whether (and in what way) you may use GenAI in your course. In case there is no information in your course guide, using AI to generate the content of your assignments is considered fraud. For theses the same rules apply, though in this case you should talk to your supervisor about what is considered an acceptable use of AI in your project.

2. Do not use AI as a replacement of skills

Don’t use AI as a replacement for essential academic skills. Instead, use AI tools in ways that support and enhance those skills. The same applies to foundational knowledge in your field of research. If you don’t know the basics about your topic, how can you accurately judge whether the output given by an AI is reliable or truthful?

Figure with step-by-step guide to determine when it is safe to use ChatGPT

A list of possible uses is provided here.

3. Think of the limitations

It is crucial to be mindful of an AI tool’s limitations. These will differ depending on the specific tool. What are these limitations, and do they pose a problem for your specific use case? Are you able to work around them? Based on the answers to these questions, make an informed decision on whether to use the tool, in what way, and for what purpose. For example, you should not ask a language model to search for literature or to solve mathematical problems. Nor should you use a model if you don’t have the knowledge or experience to critically review the outcome.

4. Prompting

Do your research on how to write good prompts in order to use your AI tool most effectively. What kind of input does it expect? What can you include in your prompt to improve the quality of the tool’s output?

Check our guides on Large Language Models (LLMs) and Image Generators for more information and tips about prompting.

5. Be critical

Always critically evaluate the (correctness of) the output of your specific AI tool, and compare it to other sources of information. Also validate the source of the information. An answer using company marketing information or a low-grade journal as "source" may not be reliable. The tools you use themselves also may have specific sources of information and therefore biases, which should be critically evaluated before using the tool and its output.

6. Data security

Mind your data security: anything that you feed into external AI tools like ChatGPT may be stored by the company indefinitely, and might be used in unspecified ways. This means that you should both think carefully about what personal information you provide them, and that you should not upload any sensitive data. Be aware that some databases/chair groups/internship companies/etc. might not allow you to share their data with AI tools as this may violate agreements with financers of projects or may even violate intellectual property rights (IPs).

7. Sustainability

Think about sustainability: (generative) AI tools require vast amounts of computer processing power to train, and with frequent use also a significant amount to use. As a result, they consume vast amounts of energy and water, and emit a large amount of CO2. Therefore, consider not using AI when a less energy-intensive alternative, or not using a tool at all, is an option.

8. Reference the AI you're using

For some citation styles (e.g. APA, MLA, and Chicago), guidelines exist on how to properly cite generative AI (McAdoo, 2023; MLA, 2023; The University of Chicago, n.d.). As an example, when using APA it is recommended to include a description of how you used the specific tool, your prompt, and relevant portions of its output in a relevant section of your paper (this could be the methods section, or the introduction). You may want to put the AI tool’s full output in an appendix (or, if applicable, in the online supplementary materials) (McAdoo, 2023). For more information on how to disclose your AI use, see the “How to document your AI use” page.

9. Maintain an archive

When using AI for a project or a thesis it is useful to maintain an archive / data repository in which you store your prompts and subsequent outputs from the AI models you used. This way you can transparently show your use of AI, and reproduce your work later should questions arise about your work.