Prompt Engineering Workshop

Prompting is becoming a modern form of literacy. The precision of the instruction directly affects the quality of the response. Vague questions lead to superficial and less useful results.
Through this method, students change the way they think about AI. Instead of treating it as a source of ready-made answers, they work with it as a capable but literal assistant that needs clear instructions, context and constraints. They discover the importance of designing the prompt, defining the role, aim, response format and quality criteria. In this way, they learn to take responsibility for the outcome of their work, while developing their ability to formulate ideas and organise tasks.
The method can be implemented in several variants. In the basic variant, students create several versions of a prompt for the same task and compare the responses, analysing which elements of the prompt improved the quality of the result.
In the iterative variant, they work on a single prompt in successive stages, gradually refining it and observing how the response changes.
In the critical variant, they are given ready-made but imperfect prompts, and their task is to improve them and justify the changes they have made.
A project-based variant can also be used, in which students prepare a set of prompts for, a specific task, such as a report or analysis, thereby building a coherent system of instructions leading from the problem to the solution.

Course: Engineering Design
Topic: Designing a Research Plan
Students bring their own unstructured literature notes or initial project ideas. Their task is to create a prompt that acts as a process analyst: it reviews the materials and helps identify logical gaps in the planned study.

Warm-Up (10 min–15 min)
Naive approach. Everyone types: Develop a research plan on [TOPIC]. The AI produces a generic template that fits everything and nothing. Students quickly realise that, without their own input, the bot is of limited use.

Anatomy of a Prompt (10–15 min)
The lecturer presents the structure of a mega prompt:
Role: Act as an expert in quantitative research methodology.
Context: Students paste in their own theses and a list of collected sources.
Task: Identify three potential weaknesses in my research assumptions.
Constraints: Do not write the content for me; focus on structure and logic.

Workshop (20–30 min)

Students refine their prompts. They add instructions such as: Ask me guiding questions until you have a clear understanding of my research idea. The aim is not to produce a finished document, but to engage in a dynamic process of clarifying and refining the idea.

Stress Test (20–30 min)
Groups exchange their instructions. If Group A’s prompt helps to organise Group B’s chaotic notes in a meaningful way and identifies gaps in them, this indicates that the prompt has been well designed.

• When students already have a substantive outline but do not know how to refine it.
• When teaching students to write complex algorithms or medical/legal procedures.

• If students do not have their own materials, the AI may start inventing content instead of structuring it.

• AI writes so convincingly that students may stop checking facts, succumbing to the illusion of professionalism → every claim generated by the AI must be supported by a link to a primary source or academic publication.
• AI models often reproduce dominant Western patterns of thinking while overlooking other perspectives → students are required to add a paragraph presenting a viewpoint
  completely ignored by the AI (e.g. the perspective of another culture).

Easier → Provide students with a ready-made prompt template containing gaps to be filled in (e.g. Act as a [ROLE], analyse my [TEXT], and identify [ERRORS]).
More Challenging → Students must create a prompt that encourages the AI to reason step by step and ensures that the bot first critiques the assumptions before suggesting a structure.

• In the tool configuration, clearly specify that its purpose is to analyse and organise, not to create ready-made content.
• Encourage students to add instructions requiring step-by-step reasoning and careful checking of assumptions before any proposed solutions are offered.
• Introduce the principle that every AI response should lead to a more precise definition of the problem, rather than closing it down. The aim is better understanding, not quick completion.

Formative Assessment
Students present the history of changes to the prompt and explain why each modification was introduced, referring to the quality of the responses and the tool’s limitations. The key point is to show that students understand how individual elements affect the outcome of the work.

Summative Assessment
It involves submitting a tested prompt together with a brief justification of the prompt engineering techniques used, such as role prompting or worked examples, as well as an explanation of why the chosen solution is safe and does not lead to plagiarism.