The Idea
The quality of AI output is almost entirely determined by the quality of the prompt. This is not a minor point — it’s the central practical skill for anyone who works with AI tools, and it’s rarely taught systematically.
This challenge makes that relationship visceral. Participants try five prompts for the same task, score each output, and develop a concrete sense of what makes prompts work.
Choose Your Challenge Task
Pick one task for the session. All prompts will be variations on this task.
Good challenge tasks:
- Write an email declining a job offer politely
- Explain how photosynthesis works
- Help me prepare for a conversation with my parent about getting a later curfew
- Write an introduction for a presentation about climate change to a skeptical audience
- Help me understand what went wrong in this argument I had with a friend (they’ll add details)
Write the chosen task at the top of your scoring sheet.
The Five Prompts
Write five versions of a prompt for the same task, ranging from vague to highly specific. Here’s a worked example for “write a college essay introduction”:
Prompt 1 (Vague): “Write a college essay.”
Prompt 2 (Slightly better): “Write an introduction for a college application essay.”
Prompt 3 (With context): “Write an introduction for a college application essay. I’m applying to study environmental science. I grew up in a drought-affected agricultural region and want to write about how water scarcity shaped my career interests.”
Prompt 4 (With specifics and constraints): “Write an opening paragraph (100–150 words) for a college application essay. The essay should start with a specific scene or memory, not a generic statement. I’m a first-generation college student applying to environmental science programs. I grew up on a farm in California’s Central Valley during years of drought and want to connect that to my interest in water policy. The tone should be personal and reflective, not formal.”
Prompt 5 (Full specification + role): “You are helping a talented high school student write a college application essay. Write a 100–150 word opening paragraph that: (1) starts with a specific sensory memory from childhood on a farm during the California drought, (2) smoothly transitions to the student’s interest in environmental science, (3) avoids clichés like ‘ever since I was young’ or ‘I have always been passionate about,’ (4) uses concrete, specific language rather than abstractions, (5) ends with a sentence that makes the reader want to know more. The student is applying to UC Berkeley’s environmental science program.”
Run the Test
For each prompt:
- Enter the exact prompt as written (no editing mid-session)
- Record the output in full
- Score it immediately before moving to the next prompt
Scoring rubric (1–5 on each dimension):
- Relevance: Did it actually address the task?
- Quality: Is the output genuinely good?
- Specificity: Did it incorporate the specific details you gave it?
- Would I use this?: Honest gut check — would this actually be useful?
- Total: /20
Analyze the Results
After all five prompts:
The pattern: In almost every session, scores improve dramatically from Prompt 1 to Prompt 4, then sometimes level off or decrease slightly at Prompt 5 (over-specification can constrain the AI counterproductively).
Discussion questions:
- “What’s the biggest difference between the output from Prompt 1 and Prompt 3?”
- “What elements of Prompt 4 made the biggest difference?”
- “Was there a point where adding more constraints made the output worse?”
- “What would you add or remove from Prompt 5 to improve it further?”
Round 2: Beat Your Best
Now, try to write a better prompt than Prompt 4. Use everything you’ve learned about what made the higher-scoring prompts work.
Run it. Score it. Did you beat your previous best?
What Good Prompts Have in Common
Draw out from the session:
- Context: Who is this for? What situation are they in?
- Specificity: What concrete details matter?
- Constraints: Length, tone, format, what to avoid
- Role: Sometimes giving the AI a role (“you are a college admissions expert”) improves output
- Examples: Showing what you want works better than describing it
- Output format: Be explicit about what the response should look like
The Bigger Lesson
Prompt engineering is, fundamentally, the skill of communicating clearly and precisely. The AI can only do what you tell it to do — exactly like a very literal computer program. If your instructions are vague, the output will be generic. If your instructions are specific, you get something more useful.
This is why the skill transfers beyond AI. Clear communication, precise specification of requirements, the ability to give feedback that actually helps — these matter in every professional context. Prompt engineering is a training ground for a more general skill.
The uncomfortable flip side: If the AI only does what you tell it, and you can make it do almost anything with the right prompt, who is responsible for what it produces? When someone uses an AI to write something harmful or false, and the AI complies — who bears responsibility?
This question doesn’t have an easy answer. It’s worth sitting with.