Skip to main content
To KTH's start page

Use generative AI efficiently and ethically

There are questions regarding the use of generative AI applications that needs answers before they can join the other AI-powered tools we use at KTH, such as Grammarly. How can we effectively use generative AI, while reducing the risk of misuse? How can we learn to use it in a way that is ethical and academically honest, and help students do the same?

The number of generative AI applications that are available online is huge and are currently increasing every day. These applications can generate all sorts of materials: text, pictures, code, sound, video, music, etc.

We will use ChatGPT as an example here, which is an AI application that generates text. It seems fitting since most people use this generative AI first, and many other generative AIs work similarly.

Efficient ways to use generative AI such as ChatGPT

To use a generative AI, you need to give it instructions in some way, usually in the form of questions or statements. These instructions are called prompts. For chatbots like ChatGPT you enter the prompts in a textbox and get the response in a chat. Most generative AIs accept prompts in similar ways, even if the result may be displayed differently.

Write purposeful prompts

You should write prompts with your idea of the output in mind, otherwise, you risk getting generic answers from the generative AI. When writing prompts, do the following to get more focused and relevant outputs:

  • Specify constraints. For example, “only list the three most important points” or “do not repeat yourself”.
  • State your purpose, either explicitly or implicitly. For example, “quiz me on X, I intend to learn the subject” or “suggest a lesson plan to best teach X”.
  • Include longer texts and focus the rest of the prompt on those texts. For example, “create feedback on how to be a better writer based on the following essay”.
  • Define a persona or role for the AI. Some simple examples: “you are an expert logician” or “answer as a confused student”. You can also add much more detail, see examples on the page AI Personas to Talk to in ChatGpt (aidemos.info) .

If you’re not happy with the result, ask the generative AI to try again with the same prompt or write a different prompt. You can also refine the output by iterating it through interactive dialogue with the generative AI, as explained under the next heading.

Iterate through interactive dialogue

Using a generative AI efficiently is more than just writing a great prompt, it is about working with the AI to iterate. After getting a response to your first prompt you can act as if you are interacting with a person. For example, ask it to explain something in more detail or by using other words, or ask it to correct something it got wrong. This way, you do not need to write a “perfect” prompt, you simply chat and guide the AI to the result you want.

As an example, we asked ChatGPT (in May 2023): “How would you explain how to use ChatGPT to someone who has never used it before?” However, the answer was too long to cite here so we tried refining it by asking: “Can you please write this more summarized?”. So, we got an answer that was shorter and more on point:

ChatGPT med en aktiv konversation och en chattbox på engelska. Vänstra palnelen visar tidigare chatt

“Certainly! Here's a more summarized version:

  1. Enter your query or statement in the chatbox.

  2. Read the response from ChatGPT.

  3. Continue the conversation by asking follow-up questions or providing more details.

  4. Experiment with different queries and variations.

  5. Be clear and specific in your instructions or context.

  6. Remember that ChatGPT has limitations and may occasionally provide incorrect answers.

  7. Enjoy the conversation and provide feedback for improvements.

Hope that helps! Let me know if you have any more questions.”

If we wanted to, we could have kept improving the output. For example, by asking it to elaborate on points 3 and 4, remove point 7 and explain what we need to do about point 6. In this way, you can chat with the generative AI until you have the output you want.

Ethical use of generative AI as a teacher

How can we use generative AI in a way that is ethically justifiable? This question is not easily answered. There is a lot to take into consideration when deciding if you should incorporate it into your course. Here we have gathered support regarding ethical use, both practical hands-on advice and more general guiding questions.

Dos and don’ts for ethical use

  • Do not enter any personal information about yourself or your students, such as when writing prompts. The privacy and confidentiality of your students are important to consider.
  • Be transparent about when and why you are using AI. By this you are providing a good practice to your students on how to use it honestly. 
  • Be aware of and have a plan to handle biases and limitations of the AI and ensure that your students understand how to do it as well. AI tools are not free from bias and therefore we should use them with caution. 
  • Do give students the choice to opt out of using AI or having it used on their material. You cannot ensure that the students' material will not be used to train the AI, and they should get to decide how their material is used. 
  • If possible, opt out of giving your data to the AI application for training, to protect any potential problems with your copyright. For some AI applications this is possible to do under settings.

Four key ethical considerations as an underlining guidance

The European Commission has published Ethical guidelines on the use of artificial intelligence (AI) and data in teaching and learning for educators (European Union, 2022, education.ec.europa.eu) . These guidelines identified four key ethical considerations as underlining guidance when using AI in teaching.

  • Human agency. Consider how using AI will impact your and the student's capability to be competent members of society.
  • Fairness. Make sure everyone can use the generative AI application and that the AI does not discriminate or exclude people due to biased algorithms.
  • Humanity. You should have a human-centric approach to AI where there is respect for people's identity, integrity and dignity.
  • Justified choice. Use knowledge and data to justify appropriate collective choices with multiple stakeholders. Do not over-rely on data that might be biased or insufficient algorithms.

Selected guiding questions

The following is a list of guiding questions to help you reflect on the ethical use of generative AI in your course. Use them to identify and address possible challenges for each learning activity involving generative AI. We recommend you revisit them as the course progresses, especially if generative AI is used differently than you expected before the course started. The questions are based on the “guiding questions for educators” in the Ethical guidelines on the use of artificial intelligence (AI) and data in teaching and learning for educators (European Union, 2022, education.ec.europa.eu) .

  • How does the AI system affect the didactical role of the teacher? Should any learning activities be adapted accordingly?
  • How can students opt out and is it possible to do without it affecting their grades?
  • How do you prevent overconfidence in or overreliance on the AI application?
  • Are all teachers sufficiently trained to safely use the AI application without causing harm or violating the rights of students?
  • Does the use of the system create any harm or fear for individuals or for society?
  • Are the instructions and information accessible and presented in a way that is clear both for teachers and students?
  • Is the system accessible to everyone in the same way without any barriers? How is this ensured?
  • Who is responsible for the ongoing monitoring of content produced by the AI application?
  • Who is responsible and accountable for final decisions made regarding the procurement and implementation of the AI application?

Can you require students to use AI?

No, we cannot require students to use generative AI at this moment (written in august 2023). The reason is that we cannot yet ensure that generative AI applications are legal according to data privacy, copyright and GDPR laws. This topic is widely discussed and is currently being investigated in the EU, therefore it is possible that this recommendation will change in the future. Read more about ethical issues and legal matters .

Until we are able to fully control how our data is being used by generative AI applications, we should be extra careful with how we use them and only require useage on a voluntary basis.

Further reading 

General usage:

ChatGPT and artificial intelligence in higher education: quick start guide (unesco.org)

Writing prompts: 

Generative AI and ethics:

Recommendation on the Ethics of Artificial Intelligence (unesco.org)