Week 5: Prompting Fundamentals

 

Welcome to Week 5 of the AI Literacy Course.

In Lesson 4, we learned that AI should be used for a suitable task, with safeguards that match the possible consequences.

Once we have decided that AI can play a useful role, we need to communicate the task clearly. We do this through a prompt.

A prompt is not a magic spell. It cannot guarantee a correct answer. But a clear prompt can reduce unnecessary guessing, improve relevance, and make the response easier to evaluate.

Learning goals

By the end of this lesson, learners should be able to:

• Explain what a prompt is

• Define a purpose before using AI

• Write one clear and specific task

• Provide necessary context without unnecessary personal information

• Request a useful response form

• Set simple boundaries for the task

• Evaluate whether the first response matches the original purpose

1. What is a prompt?

A prompt is the input you give an AI system.

It may include:

• a question;

• an instruction;

• text you want the AI to work with;

• an image or document;

• examples;

• information about your goal.

For example:

“Explain photosynthesis.”

This is a prompt. It contains a task, but the AI does not know who the explanation is for, what the learner already understands, or how detailed the response should be.

A clearer prompt might say:

“Explain photosynthesis to an upper-secondary learner in about 150 words. Use plain language and include one everyday example.”

The second prompt gives the AI more useful guidance.

2. Begin with a purpose

Before writing a prompt, ask:

What should I understand, practise, decide, or create after using AI?

A vague purpose often produces a vague or overly broad response.

For example:

“Help me study.”

The AI must guess what subject you are studying, what you already know, and what kind of help you need.

A clearer version would be:

“Help me practise explaining the difference between training and inference. Ask me one question at a time and wait for my answer.”

The learner’s purpose is now visible: practising an explanation rather than receiving a finished text.

A polished response is not useful if it solves the wrong problem.

3. Give one clear task

A good prompt tells the AI what action to perform.

Useful task verbs include:

• explain;

• compare;

• summarise;

• classify;

• suggest;

• translate;

• ask;

• give feedback;

• create practice questions.

Compare these prompts:

“Tell me about machine learning.”

“Compare supervised and unsupervised learning using one example of each.”

The second prompt gives the AI a more specific task.

Avoid placing too many different tasks in one prompt. For example:

“Explain climate change, create a presentation, find sources, write a speech, test me, and translate everything into Swedish.”

This request contains several tasks with different purposes. The result may be incomplete or poorly organised.

Begin with one main task. More complex work can be divided into smaller stages.

4. Add the context AI needs

Context is the relevant background information needed to perform the task.

Useful context may include:

• the intended reader or audience;

• the learner’s current level;

• the subject being studied;

• the purpose of the work;

• material that should be used;

• important local conditions.

For example:

“I am an adult learner studying AI literacy. I understand what machine learning is, but I still confuse training with inference.”

This information helps the AI adjust its explanation.

However, more context is not always better. Unnecessary details can distract from the task or create privacy risks.

Use this rule:

Give enough information for the task, but leave out information that is unnecessary, private, sensitive, or confidential.

Instead of sharing a real person’s details, you may be able to remove identifying information or use a fictional example.

5. Request a useful response

Tell the AI how you want the response to be presented when this matters.

You might request:

• a short explanation;

• five bullet points;

• a comparison table;

• two examples;

• step-by-step instructions;

• plain language;

• one question at a time;

• a response within a word limit.

For example:

“Compare machine learning and generative AI in a table. Include their purpose, typical output, and one everyday example.”

A useful response form can make the output easier to read and check.

But format does not guarantee quality. A well-designed table may still contain incorrect or incomplete information.

Avoid forcing a format that removes important detail. Some topics need explanation, uncertainty, or exceptions that do not fit neatly into a very short answer.

6. Set simple boundaries

A boundary tells the AI what it should or should not do during the task.

Useful boundaries might include:

• “Use only the text I provide.”

• “If the information is missing, say that it is missing.”

• “Separate facts from suggestions.”

• “Label any assumptions.”

• “Do not include personal information.”

• “If the task is unclear, ask me a question before answering.”

Boundaries can reduce unnecessary guessing and make the response easier to examine.

However, an instruction such as “Do not make mistakes” cannot guarantee accuracy. The AI may still misunderstand the task or produce unsupported information.

The user must continue to review the output.

7. AI cannot read your mind

AI works with the information available in the interaction, including the prompt, visible conversation, supplied documents, and any connected tools.

It does not automatically know:

• what you intended but did not write;

• which parts of a topic you already understand;

• what your teacher or workplace expects;

• which local rules apply;

• whether an unstated detail is important.

When information is missing, the system may make an assumption. That assumption can sound reasonable and still be wrong.

You can reduce this problem by writing:

“If you need important information that I have not provided, ask me before answering.”

This does not eliminate every mistake, but it makes hidden assumptions easier to notice.

8. Review the first response

The first response is a starting point, not necessarily a finished answer.

After receiving it, ask:

• Does it answer my actual question?

• Is it suitable for the intended reader?

• Did it follow the requested form?

• Did it make assumptions?

• Is anything important missing?

• Which claims need verification?

• Do I understand the response well enough to use it?

You might decide to accept part of the response, check a factual claim, change the task, or use a non-AI source.

A better prompt can improve relevance, but it cannot repair unreliable information, missing knowledge, or an AI tool that is unsuitable for the task.

Detailed methods for questioning and improving the first response will be introduced in Lesson 7.

9. From fundamentals to RUKF

A useful first prompt usually answers several questions:

• What is the purpose?

• What should AI do?

• What background information does it need?

• Who is the response for?

• How should the response be presented?

• What boundaries are needed?

In Lesson 6, we will organise these ideas using the RUKF framework:

• Role

• Task

• Context

• Format

For now, concentrate on choosing a suitable purpose and writing one clear request.

Classroom activity: From vague to clear

Choose a topic that you are currently studying.

Step 1: Write a vague prompt

Example:

“Help me understand climate change.”

Step 2: Identify what is missing

Consider:

• What is the purpose?

• What exact task should AI perform?

• What context does it need?

• What kind of response would be useful?

• What boundary should be included?

Step 3: Rewrite the prompt

Example:

“Explain the difference between climate and weather to an adult learner. Use plain language, approximately 150 words, and one example from Sweden. If an important detail is uncertain, identify the uncertainty.”

Step 4: Compare the responses

Enter both prompts into the same AI tool and compare the results.

Record:

• both prompts;

• both responses;

• which response better matched your purpose;

• what the clearer prompt improved;

• what still required checking.

Step 5: Write a short reflection

In three or four sentences, explain which response was more useful and why.

Do not judge only by length or appearance. Consider relevance, clarity, assumptions, and whether the response supports your learning.

Reflection questions

1. What is a prompt?

2. Why should purpose come before prompting?

3. Why is one clear task often better than several tasks at once?

4. What context would help AI answer a study question?

5. What information should you leave out?

6. How can a requested response form make checking easier?

7. What boundary can reduce unnecessary guessing?

8. Why is the first response not necessarily the final answer?

9. How can a well-written prompt still produce an incorrect response?

Key vocabulary

Prompt:
The question, instruction, material, or other input given to an AI system.

Purpose:
The reason for using AI and the result the user wants to achieve.

Task:
The action the AI is asked to perform, such as explaining, comparing, summarising, or asking questions.

Context:
Relevant background information that helps the AI respond appropriately.

Response form:
The requested structure or presentation of the output, such as a table, short explanation, or list.

Boundary:
An instruction that limits the task or states what the AI should include, avoid, or identify.

Assumption:
Something treated as true even though it has not been confirmed.

Evaluation:
Examining whether the response matches the purpose and meets the requirements of the task.

Summary

In Week 5, we learned that prompting is a practical communication skill.

A useful first prompt begins with a clear purpose. It gives AI one main task, provides the context needed, requests a suitable response, and sets reasonable boundaries.

Clear prompting can reduce unnecessary guessing and make an answer easier to evaluate. It cannot guarantee that the answer is true, complete, or suitable for every use.

The goal is not to discover a perfect prompt.

The goal is to communicate the task clearly while keeping judgement, checking, and responsibility with the person using AI.

Lesson 5 Interactive Quiz: Prompting Fundamentals

Choose one answer for each question. Then select Check my answers. You will receive feedback for every question. This practice quiz does not collect names or scores.

1. What is a prompt?

A. Only a question that can be answered with yes or no. B. The final text produced by an AI system. C. A question, instruction, or other input given to an AI system. D. A guarantee that the AI response will be correct.

2. What should you decide before writing a prompt?

A. What you want to understand, practise, decide, or create. B. How long the AI should spend producing the answer. C. Which technical words will make the prompt sound advanced. D. How to make the prompt as long as possible.

3. Which prompt gives the clearest first task?

A. “Tell me everything about energy.” B. “Write something useful about electricity.” C. “Make energy easy and interesting.” D. “Explain renewable energy to an adult beginner in 150 words and include one example from Sweden.”

4. A learner asks AI to explain a topic, find evidence, write an essay, translate it, and create a quiz in one prompt. What is the best improvement?

A. Add more tasks so the AI understands the whole project. B. Divide the request into smaller tasks and begin with one clear step. C. Remove the topic and let the AI choose what to discuss. D. Ask for a longer answer without changing the request.

5. Which information is the most useful context for a study prompt?

A. The learner’s full name, address, and date of birth. B. Every detail about the learner’s family and workplace. C. The learner’s study level and the part of the topic that is difficult. D. Unrelated information that makes the prompt appear more detailed.

6. A learner wants to compare two AI concepts. Which response form would probably make checking easiest?

A. A table with definitions, differences, and one example of each concept. B. A long paragraph with no headings or examples. C. A creative story that does not identify the concepts clearly. D. A list of technical terms without explanations.

7. Which boundary is most helpful when important information may be missing?

A. “Make a reasonable guess and present it as certain.” B. “Always give an answer, even if the request is unclear.” C. “Use any missing details that normally appear in similar cases.” D. “If important information is missing, ask me before answering.”

8. Why can a vague prompt produce an unsuitable answer?

A. AI refuses to answer every short prompt. B. AI may guess the user’s purpose or make assumptions about missing context. C. AI can answer only when every prompt contains a table. D. Short prompts always produce shorter but more accurate answers.

9. What should a learner do after receiving the first AI response?

A. Submit it immediately if it is written clearly. B. Accept it if it followed the requested word limit. C. Compare it with the purpose and check assumptions, missing content, and important claims. D. Replace the original purpose with whatever the AI decided to discuss.

10. Which statement about prompt quality is correct?

A. A clearer prompt can improve relevance, but the response may still require checking. B. A detailed prompt guarantees that every factual claim is correct. C. A good prompt can give an AI system knowledge it does not possess. D. A well-formatted response no longer requires human judgement.

Reflection: Rewrite one vague prompt by adding a clear purpose, one task, necessary context, a useful response form, and one boundary.