Prompts: decalogue

Prompts: decalogue

Prompts are the main way to interact with generative Artificial Intelligence tools. A prompt is not just a question: it is the way we indicate to AI what we need, what we want it for, and how it should present it.

Mastering this skill makes it possible to obtain more precise, useful responses that are better aligned with the university’s teaching, research, and academic needs. It is already an essential digital competence for the entire university community.

Decalogue for writing good prompts

Fundamental practices to guide AI appropriately in educational, research, and professional contexts.

1. Be clear and direct

State exactly what you need, without ambiguity.
   ⇾“Explain in simple terms…” works better than “Explain this to me.”

2. Provide relevant context

Specify who the response is for and for what purpose: educational level, subject, type of activity.
   ⇾“For first-year Engineering students…”

3. Define the objective of the task

Indicate what action you expect: summarise, explain, compare, analyse, transform, generate ideas…
   ⇾“Summarise this text in five key ideas.”

4. Indicate the format you need

List, table, outline, numbered steps, short or long text.
   ⇾“Present it in a table with two columns.”

5. Specify the tone or style

Academic, informative, formal, simple…
   ⇾“Explain it using clear, accessible language.”

6. Set the level of detail

Brief, in depth, with examples, without technical jargon…
   ⇾“Include two examples related to the educational field.”

7. Indicate what should be avoided

This helps reduce errors and improves accuracy.
   ⇾“Avoid technical jargon and do not repeat ideas.”

8. Provide prior information if necessary

Include relevant texts, data, or descriptions.
   ⇾“Use only this text as the basis for the summary.”

9. Ask for verification or nuance when there is uncertainty

This is particularly important in research.
   ⇾“Indicate any data that cannot be verified accurately.”

10. Critically review the response

AI does not replace academic judgement: cross-check, validate, and correct.
   ⇾“Review the response in light of these additional criteria.”