Prompting: conversing with AI
Using generative AI is not about obtaining the perfect answer on the first attempt. Effective interaction is built step by step, refining the prompt according to the responses obtained. This is the foundation of expert use.
AI responds to what is written, not to what is intended.
The tool does not interpret hidden intentions. Iteration allows us to make explicit what was missing.
⇾The first attempt is exploratory
It helps verify what the AI understood and what adjustments the instruction needs.
⇾Each adjustment reduces ambiguity
Adding details about context, format, or purpose improves the accuracy of the result.
⇾Effective prompting is a dialogue
Quality emerges from the cycle: ask → review → refine → ask again.
⇾Iterating is part of the method, not a mistake
Refining a prompt is a key digital competence that improves with practice.
⇾Iteration develops academic judgement
It helps identify which details are essential, which parts cause errors, and how to guide the request more effectively.
Understanding how AI “reasons” improves iteration
Generative AI does not understand the world as a human does. Knowing how it works allows prompts to be adjusted more effectively.
⇾AI predicts probabilities, not meanings
It generates the most probable word according to learned patterns; it does not “understand” the concept.
⇾AI does not know what is not written
If the prompt does not specify level, tone, format, or purpose, the tool will decide for itself… and it will not always choose correctly.
⇾AI may invent details to fill gaps
This occurs when the instruction is ambiguous or incomplete. For this reason, it is essential to critically review every response.
Prompt model
This example of a long prompt shows how to integrate context, objective, format, and constraints into a single complete instruction.
(Mechatronics)
- I need a clear and accessible explanation of the concept of mechatronics for first-year Engineering students.
- Write the text in an informative tone and without unnecessary technical jargon, so that it is understandable for students with basic scientific knowledge.
- Include:
- a brief definition of the term,
- its main areas or components,
- and two practical examples of real applications of mechatronic systems.
- Present the response as a three-part numbered outline.
- Avoid repeating ideas or introducing advanced content (such as complex mathematical modelling or real-time control).
- If a concept cannot be explained precisely without technical terminology, state this explicitly and provide a simplified version.
- This prompt integrates the objective, context, tone, format, constraints, and required level.
Risks of a poor prompt
An imprecise or ambiguous prompt may produce:
- superficial or unhelpful responses,
- errors or fabricated information,
- incorrect interpretations of context,
- convincing but false results.
- A good prompt improves the outcome, but only critical review ensures that the response is valid in an academic context.