Prompt Workbook – English
The Prompt Workbook is a collection of practical prompt examples for higher education teaching and learning that we have developed or further refined ourselves or found elsewhere. It is intended to support educators and students in exploring meaningful uses of chatbots in specific academic contexts. The prompts are intended as adaptable examples that encourage users to develop their own prompts suited to the respective context. The aim is to promote forms of AI use that meaningfully support the achievement of teaching and learning goals without displacing independent thinking, judgment, and learning.
Under “All Prompts”, you will find filters that help you find prompts suited to your needs. From the filter results, you can click through to the individual prompt pages. Each prompt has a short description of possible uses, followed by the example prompt to copy.
Please note
The prompts in this workbook are not perfect and are not suitable for every context. They should therefore be checked before use and adapted to your own needs. The prompts are not tied to a specific chatbot model. Each prompt indicates the chatbot model with which it was tested. Not every prompt works equally well with every chatbot model, and a prompt that previously worked without problems may become unreliable after an update to the chatbot model.
Always verify AI outputs
Chatbots generate responses based on statistical patterns in their training data. They do not compare their outputs against reality and can formulate false statements just as fluently and confidently as correct ones. Although models can sometimes provide useful self-assessments of their confidence, these are not reliable enough to replace your own verification. Treat all results as hypotheses that you need to verify yourself.
Make a conscious decision about using chatbots
Whether using a chatbot is helpful or not depends on several factors. If the following guiding questions can be answered with yes, the use of a chatbot can be beneficial; if some or all answers are no, the use of a chatbot is not advisable:
- Does it not matter significantly whether the chatbot’s results are right or wrong?
- Does my input contain no sensitive or personal data?
- Is my input free of copyrighted works?
- Is it likely that the chatbot’s training dataset contains high-quality and suitable information for my question?
- Am I able to judge the correctness and quality of the chatbot’s response myself?
- If my own knowledge is insufficient, do I have other options to check the quality of the chatbot’s response, and the time needed to do so?
- Is the damage caused by a factually incorrect answer from the chatbot limited and, if it occurs, can it be remedied?
- Am I willing to take responsibility for the results?
Find out more in the guide to achieving good UHHGPT results.
Contact and further information
Gunda Mohr, Hamburg Center for University Teaching and Learning (HUL)
If you have developed prompts in the context of teaching and learning that might also be of interest to others, we would be pleased if you sent them to us so that we can include them here. We will, of course, cite the source.