A prompt is essentially a set of instructions or a question you provide to a GenAI model to extract a specific response. When you interact with GenAI tools, such as the UvA AI Chat, the prompt acts as the input that guides the GenAI to generate relevant output.

Prompts are important because they directly influence the quality and relevance of the GenAI’s response. A well-crafted prompt can lead to insightful, accurate, and useful answers, whereas a vague or poorly structured prompt may result in irrelevant or confusing responses. Understanding how to create effective prompts can significantly enhance your interaction with GenAI tools.

How to formulate good prompts?

GenAI systems need clear instructions to work properly. The better you formulate your prompt, the better the response. The practice of designing good prompts is called prompt engineering. But how can you formulate good prompts?

Tips & tricks to formulate a good prompt

To easily recall the elements of a good prompt, we use the mnemonic ‘CLEAR’:

Context: Provide as much detail and context about the task you want the tool to perform as possible in your prompt.
Length and style: Be specific about the response you want to receive, such as the length of the response in words, bullet points, or paragraphs. You can also specify a particular writing style, tone, and audience, or even that you want it to think ‘outside the box’ or from a particular perspective.
Examples included: Include examples of the desired response. For example if you are asking GenAI to create test questions, give an example of a test question about another topic.
Active verbs: Start with an active verb, such as “Create …”, “Summarise …”, “Structure …”, “Generate …” or “Translate …”. This clearly defines the action you want the GenAI to perform, making the intention and direction of the task unmistakable. Active verbs help in creating precise, actionable instructions.
Role: Give GenAI a role. Specify in which capacity it should respond. For example, “You are an expert in computer sciences”, or “You are a master student who asks me for practical information about the master thesis project”, or “You are an educational expert who gives me feedback on my learning outcomes”.

Iterate!

If you’re not satisfied with the initial response, you can ask follow-up questions to make it more specific or provide new details for your instructions. Have a conversation with GenAI! It will remember the context of your initial question. Still not happy with the outcome? Ask the GenAI model to make your prompt better, so it can also give you better output.

Step-by-step reasoning

Next to this iterative process, you can ask GenAI to take you through each step of its process. Let the GenAI system walk you through its answer, so you can see how it got there. You can also provide steps for the model that represent the chain of reasoning or particular actions it should engage in.

Example prompts

Generate a lesson plan

“You are a University teacher. Create a comprehensive lesson plan for a university-level course on Environmental Science focusing on ‘The Impact of Climate Change on Marine Ecosystems.’ The lesson plan should include the following: clear objectives, detailed key concepts, step-by-step activities, essential readings, and specific assignments. Ensure the activities promote critical thinking and include options for group work and individual study. Additionally, provide assessment methods to evaluate student understanding and engagement.”

Design an interactive activity

“You are a didactical expert. Design an engaging interactive classroom activity for university computer science students focused on ‘Fundamentals of Machine Learning.’ The activity should include the following: precise learning objectives, detailed step-by-step instructions, required materials and software, and techniques for assessing student participation and comprehension. Encourage hands-on practice with coding exercises or simulations, foster collaborative group work, and incorporate real-world machine learning scenarios to contextualize the learning experience.”

Create a quiz

“You are a University teacher. Create an in-depth multiple-choice quiz for university students on ‘Macroeconomic Theory.’ The quiz should cover a range of topics including GDP calculation, inflation rates, fiscal policy, and supply and demand curves. Ensure questions vary in difficulty and include comprehensive answer explanations to reinforce learning. Structure the quiz to assess not only factual knowledge but also analytical skills, and provide feedback on common mistakes to support student improvement.”

Create exam questions

“You are a university teacher of a third-year course called Corporate Social Responsibility at a business school. Create two versions of exam questions covering the topics of sustainability, circular economy, and climate change (based on the textbook by Leleux, B. and Van der Kaaij, J. (2019). Winning Sustainability Strategies. Palgrave Macmillan, London, first edition, ISBN: 9783319974446). Make sure to include 7 multiple choice questions (with 4 provided answer options each) and 3 open ended questions.Make sure the questions are similar in difficulty and length to the example below:[enter the version of the exam you have already created, e.g., for the previous semester]Provide correct answers to the questions.”

Explain a difficult concept

“You are a University Physics teacher. Explain the concept of ‘Quantum Entanglement’ in simple terms suitable for university physics students. Provide a concise overview, highlight key principles, and detail real-world applications in technology and research. Include examples to illustrate complex ideas, and address common misconceptions to enhance understanding. Aim for clarity and coherence to ensure students grasp this challenging topic.”

Talk to a historical figure

“You are Ruth Bader Ginsburg. I will ask you several questions about gender-equality and discrimination. Please answer them from the perspective of Ruth Bader Ginsburg, as if she was the one providing the answers. Answer those questions in the style of language that Ruth Bader Ginsburg woud use. In your answer, be precise and provide examples from Ruth’s speeches and writings. Limit the answers to maximum two paragraphs. Make sure to use language that advanced non-native English speaker would understand.

Also take a look at the persona examples for already-used system prompts.

Conclusion

Make your prompt specific, give it a role, and specify which content it can use.

Is it not at the level that you want it to be? Iterate and give the GenAI tool more directions and instructions to create a better output. Or give the GenAI tool more information! Think about uploading your old exams to show how to construct good exam questions, uploading a piece of text to show your own writing style, or sharing background information that the tool should take into account while performing the task (e.g., a TLC page with information about Constructive Alignment).

More information?

Do you want more information about prompt engineering? Visit our GenAI e-learning for lecturers, which includes a chapter about prompt engineering.

This content is created by TLC Science for the UvA Science Faculty. Do you work at another UvA Faculty? Please follow your faculty’s specific guidelines.