Skip to Main Content

Generative AI: Prompting

This LibGuide offers information on how to use GenAI responibly, and more.

Prompting

Prompting (or Prompt Engineering) involves giving a command (prompt) to the language model behind an AI application. Prompting is a whole new way of retrieving information and it requires different skills from searching a database or search engine.
 

Picture from Flaticon.com

The special thing about generative AI is that, no matter the exact input, you always get information back. This is not to say that the exact input doesn't matter. A well-known saying about GenAI is: Garbage In = Garbage Out.

On this page, you will find several frameworks through which you can be inspired when prompt engineering.

Prompt Library

Using a good prompt is crucial, but you don't (always) have to build it from scratch yourself. Various organisations have now built their own Prompt Library. A very useful one is that of Maastricht University. These contain a kind of ‘template’ that you can edit to suit your own information needs.

Nationwide, work is underway on a Prompt Library for all higher education institutions.

Four sentences (basics)

Following the four-sentence principle, build your prompt based on the components:

  • Persona (What role should GenAI take?): You are an expert in ...
  • Goal (What is GenAI aiming for?): Your goal is to ...
  • Task (What should GenAI do for you?): Your task is to ...
  • Context (What else does GenAI need to know?): Include a table in the answer and make sure you don't use more than 3 columns.

Source: MacNeil, C. (2024,17 May). Write better AI prompts: A 4-sentence framework. Asana. Accessed 12 December 2024, from https://asana.com/nl/resources/ai-prompting-basics

PREPARE

Dan Fitzpatrick (The AI Educator) has developed the PREPARE framework. PREPARE stands for:

  • Prompt (What do you want from GenAI?): Write a roadmap for undergraduate Health students to conduct thesis research on ... using qualitative interviews.
  • Role (What role should GenAI assume?): You are a thesis supervisor at hbo and have a lot of experience doing practice-based research yourself. 
  • Explicit (Do you e.g. have a specific methodology in mind?): I want to do practice-oriented qualitative research using interviews.
  • Parameters (What else should the answer satisfy?): Answer using playful language in English, maximum 200 words and maximum 12 words per sentence. The use of language should match the target audience: ...
  • Ask (Should GenAI ask you questions back?): Ask me reflection questions, with the aim of further fine-tuning the questionnaire.
  • Rate (Let GenAI rate its own answer with a point): Give your own answer a score from 1 to 10.
  • Emotions (Does anything depend on the answer?): My graduation depends on this questionnaire. I am really struggling with my thesis and with a good set of interview questions, hopefully I can still graduate this year.

Source: Fitzpatrick, D. (2023, 29 November). AI Prompt Prepare Update. Digital Learning. Accessed 9 December 2024, from https://blogs.loucoll.ac.uk/learningtechnology/2023/11/29/ai-prompt-engineering-framework-update/

"Chatten met Napoleon"

Barend Last (educationist) and Thijmen Sprakel (English & music teacher) recommend six steps:

  1. Describe the task: Describe specifically and unambiguously what kind of answer you want to see. Eg.: Describe the procedure of inserting an indwelling catheter.
  2. Encourage a persona: In other words: Tell the chatbot ‘who’ it is. Eg.: You are a fourth-year nursing student.
  3. Outline a roadmap: Through which steps should the chatbot perform your task? Eg.: First, explain how the nurse prepares the patient properly, then present through which steps the catheter is inserted and finally name what is needed for proper aftercare.
  4. Provide context and constraints: Are there things the chatbot needs to take into account? Eg.: Describe how the patient will be as uncomfortable as possible with the catheter, but limit this to tricks to minimise physical discomfort.
  5. Specify the purpose and target audience: These are very decisive for the output generated by the chatbot. Eg.: The goal is to make elderly people living at home experience as little disruption as possible from an indwelling catheter.
  6. Format the output: What form should the output take? What length, structure and formatting? Eg.: Write a roadmap of 250 words and use simple and accessible language. Limit each step to two to three short, clear sentences. Make the most important terms bold.

Source: Last, B., & Sprakel, T. (2024). Chatten met Napoleon: Werken met generatieve AI in het onderwijs (2e herziene editie). Boom. (Catalogue)

Meta Prompting

Meta Prompting means using an AI chatbot to create a prompt that helps the chatbot ask itself a question. This method is fast and can be much more effective. Several scientific methodologies rely on the principle of meta prompting. For example:

  • Summarizing: What is a good prompt to summarize these documents? - instead of: Summarize this. 

Source: Van Ess, H. (2025, 16 July). Google vs. AI: When to use which? Digital Digging. Accessed 26 August 2025, from https://www.digitaldigging.org/p/google-vs-ai-when-to-use-which

www.zuyd.nl | Disclaimer | Over Zuyd Bibliotheek