Practising Prompting

The quality of the content generated by AI depends on the quality of the questions asked, i.e., the prompts that users input. Careless or insufficiently detailed prompts result in imprecise or even misleading answers.

So-called prompt engineering is therefore developing into an important skill. Much can be learned through trial and error, but for specific purposes within various subject areas, it pays to find strategies for effective prompting together with students. By comparing and analysing the outcomes of different prompts (or, more often, sequences of increasingly refined prompts), students become more adept at subject-specific working methods while gaining deeper knowledge and experience of the possibilities and limitations of using generative AI.

Prompting Tasks

Here are some suggestions on how prompt exercises can be structured. They can, of course, be varied in many ways and may require more or less detailed introductions, depending on prior knowledge and difficulty level.

Example of a Classroom Exercise

Students are divided into groups. Each group is given a task to be answered during the class session, preferably with some time pressure. The groups organise their work themselves. The use of generative AI is explicitly allowed (but does not have to be prescribed!). All prompts and generated answers must be saved. Each group then briefly presents their proposed solution to all other groups and how they arrived at it. In the concluding discussion, the solutions and the importance of the prompt work for their quality are compared.

Example of an Online Exercise

Students are given the same task, to be answered with the help of generative AI. Students work individually and are encouraged to save the entire conversation they have with Copilot. Students upload both their answer and the conversation in Studium, where the teacher has activated peer review. Each student reads and comments on the answers and prompt handling of 2-3 other students. In a follow-up discussion (synchronously in Zoom, or asynchronously in a discussion forum), the teacher leads a review where the results of different prompt strategies are compared.

More About Prompting

A very short introductiory video on writing good prompts in CopilotExamples of prompts for educational purposesPrompt Engineering Guide - a comprehensive overviewAn interesting article, AI Prompt Engineering Isn’t the Future, which distinguishes between problem formulation and prompt engineering