Discussion drills

1. Diagnose 1

   A student gives Claude a 2,000-word essay and asks, at the very end of the prompt: "Now rewrite this to be more persuasive, but keep it formal, avoid first person, do not change the core argument, and target a university admissions panel." Claude produces a rewrite that is more persuasive and formal, but uses first person throughout and subtly shifts the core argument.

   Diagnose this failure using the serial position effect. Which instructions were most at risk? Which placement principle was violated and how?

   ▼ Teacher hint
2. Predict 2

   You place a critical constraint — "never use the word 'important'" — in the exact middle of a 1,200-word prompt. Predict the probability it is followed. Then: rewrite the prompt structure so the constraint is most likely to be respected. What specifically do you change?

   ▼ Teacher hint
3. Construct 3

   Build the full prompt for this task: "Write a 3-minute speech for a school assembly on AI plagiarism policy." Apply all three placement principles explicitly. Show your prompt layout — not just the prompt itself, but which section goes where and why.

   ▼ Teacher hint
4. Judge 4

   "Longer prompts are better prompts." Evaluate this claim using the serial position effect. Under what conditions is it true? Under what conditions does it backfire and make the output worse?

   ▼ Teacher hint
5. Compare 5

   Placing your most critical constraint at the very beginning of a long prompt vs placing it at the very end — which is more reliably followed? Does the answer depend on the constraint type? Make a concrete argument using the serial position effect.

   ▼ Teacher hint
6. Falsify 6

   "If Claude misses an instruction in a long prompt, just repeat it in the next message." Make the strongest case that this is a structurally inadequate fix — not a solution but a symptom of an unfixed problem.

   ▼ Teacher hint