The Contradiction Test
Layers Used: Layer 4 (Contradiction Challenge), Layer 6 (Iterative Drafts)
You will use the Reasoning Receipt format from Chapter 1, Exercise 1. Your annotation practice from Chapter 1 becomes error annotation here.
What You Do
Ask both Claude and ChatGPT the same nuanced question where reasonable people disagree — for example: "Is remote work better for productivity than office work?" You will receive two confident, structured, partially contradictory answers. Your job: identify exactly where the two responses diverge, determine which claims are supported by evidence vs. asserted without support, and write your own third analysis (Draft 1) that is more rigorous than either AI output. Then submit your third analysis to AI for critique, revise based on the feedback (Draft 2), and submit the final version (Draft 3).
The two AI responses with divergence points highlighted and annotated ("Claude claims X, ChatGPT claims Y — the evidence favors..."). Your Draft 1 third analysis (written before AI feedback). Your Draft 2 (revised after AI critique). Your Draft 3 (final, after reflection). A brief evolution note for each draft explaining what changed and why.
I am learning to detect contradictions between AI outputs and build a more
rigorous analysis. I asked two AI tools the same question and received
contradictory responses. I then wrote my own analysis attempting to be more
rigorous than either. Please:
(1) Evaluate my identification of divergence points -- did I catch all the
meaningful contradictions between the two AI responses?
(2) Rate my evidence assessment -- for each divergence point, did I
correctly identify which side had stronger evidence?
(3) Grade my third analysis on a scale of 1-10 for rigor, originality, and
evidence quality compared to the two AI responses.
(4) Identify the 3 weakest claims in my analysis and explain exactly what
would make them stronger.
(5) What did both AI tools get wrong that I also missed?
Question: [paste question].
AI Response 1: [paste].
AI Response 2: [paste].
My divergence annotations: [paste].
My analysis (Draft 1): [paste].
Finally, complete the Thinking Score Card for this exercise:
Independent Thinking (1-10), Critical Evaluation (1-10),
Reasoning Depth (1-10), Originality (1-10), Self-Awareness (1-10).
For each score, give a one-sentence justification.
What This Teaches You
You learn that when two confident AI responses disagree, the disagreement is a signal to think harder, not a reason to pick one randomly. Building a third analysis that improves on both forces you into genuine thinking. The three-draft evolution reveals whether you can integrate feedback and improve — or whether you stop thinking after the first attempt.