What Is a Prompt?

on

 


1. Start with the Basics: What Is a Prompt?

Answer:
A prompt is simply a question, instruction, or request you give to an AI (or a person) to get a useful response. Think of it like asking a librarian for help: the clearer your question, the better the answer.

Evidence:

  • AI systems (like me) rely on the quality of the input to produce quality output. This is often referred to as "garbage in, garbage out" in computer science (a principle documented in early computing literature, e.g., IBM, 1960s).
  • Studies on human-AI interaction show that users with no technical background struggle most with formulating clear requests (e.g., Lank et al., 2017).

Limitations:

  • This analogy (librarian) may not capture the iterative nature of AI interactions, where follow-up questions are often needed.

2. Core Principles for Writing Effective Prompts

Teach these four simple rules (with examples):

Rule

Explanation

Bad Example

Good Example

Be Specific

Vague questions get vague answers. Add details.

"Tell me about dogs."

"What are the best dog breeds for families with small children?"

Give Context

Help the AI understand why you’re asking.

"Write a letter."

"Write a polite email to my landlord asking to fix a leaky faucet."

Break It Down

If your request is complex, split it into smaller steps.

"Plan my wedding."

"Suggest a 3-course menu for a 50-person vegetarian wedding."

Ask for a Format

Specify how you want the answer (list, paragraph, table, etc.).

"Explain photosynthesis."

"Explain photosynthesis in 3 bullet points for a 10-year-old."

Evidence:

Limitations:

  • These rules assume the AI has the capability to understand and act on the details provided. Not all AI systems are equally skilled at interpreting nuance.

3. Step-by-Step Teaching Method

A. Start with Familiar Scenarios
Use everyday examples (e.g., ordering food, asking for directions) to show how details improve outcomes.

  • Example: "If you ask a chef, ‘Make me food,’ you might get anything. But if you say, ‘Make me a vegetarian pasta with no garlic,’ you’ll get exactly what you want."

B. Practice with Templates
Provide fill-in-the-blank prompts:

  • "I want to learn about [topic] for [purpose]. Give me [format] with [specific details]."
    • Filled example: "I want to learn about yoga for stress relief. Give me a 5-step beginner routine with no equipment needed."

C. Compare Bad vs. Good Prompts
Show side-by-side examples and ask learners to guess which will work better. Then, test both in a real AI (like me!) to see the difference.

D. Iterate and Refine
Teach that the first prompt is rarely perfect. Encourage:

  1. Try a prompt.
  2. Look at the answer.
  3. Ask: "Is this what I wanted? If not, what’s missing?"
  4. Adjust and try again.

Evidence:

  • Iterative prompting is a core technique in human-centered AI design (Nielsen Norman Group).
  • Classroom studies show that scaffolding (breaking tasks into steps) improves learning outcomes (Vygotsky, 1978).

Limitations:

  • Iteration requires patience, which may frustrate beginners. Emphasize that this is normal.

4. Common Mistakes to Avoid

Mistake

Why It’s a Problem

Fix

Too short

The AI guesses what you want.

Add 1–2 more details.

Too long/windy

The AI gets confused by irrelevant info.

Remove unnecessary words.

Assuming knowledge

The AI doesn’t know your background.

Explain your level (e.g., "for a beginner").

No clear goal

The AI doesn’t know what "good" looks like.

Say what you’ll do with the answer.

Evidence:

5. Hands-On Exercises

Exercise 1: "The Shopping List"

  • Task: Write a prompt to get a grocery list for a week of healthy meals for a family of 4.
  • Goal: Learners realize they need to specify dietary restrictions, budget, or meal preferences.

Exercise 2: "The Travel Plan"

  • Task: Write a prompt to plan a 3-day trip to Paris.
  • Goal: Learners add details like budget, interests (museums vs. food), or travel style (relaxed vs. fast-paced).

Exercise 3: "The Email"

  • Task: Write a prompt to draft an email to a client apologizing for a delay.
  • Goal: Learners include tone (professional, friendly), key points (reason for delay, new deadline), and length.

Evidence:

6. Addressing Uncertainty

  • What if the AI gets it wrong?
    • Teach learners to rephrase: "That’s not quite right. Let me try again: [revised prompt]."
    • Example: If the AI suggests a meat dish for a vegetarian, the user might say: "I said vegetarian—try again with no meat, eggs, or dairy."
  • What if the learner doesn’t know the details?
    • Encourage them to start broad, then narrow down. Example:
      • First prompt: "Tell me about cars."
      • Follow-up: "Now, focus on electric cars under £30,000 with good safety ratings."

Evidence:

7. Key Takeaways for Learners

  1. More details = better answers.
  2. Context is your friend. (Who are you? Why are you asking?)
  3. Format matters. (List? Paragraph? Step-by-step?)
  4. Practice makes perfect. The more you try, the better you’ll get.

8. Further Reading (Optional)

Final Note

Avoid jargon like "tokens," "temperature," or "fine-tuning." Stick to plain language and real-world analogies. The goal is to make learners feel capable, not overwhelmed.

Comments

Post a Comment