The Adversarial Defence
Why This Matters: James and the Untested Position
James looked at his Position Lock from Exercise 1. The matrix, the arguments, the confidence percentage. It felt solid. Complete. He'd done the work.
"I'm ready for the next exercise," he said.
"What makes you confident in your position?" Emma asked.
"I identified seven stakeholder groups. I have three arguments with reasoning behind each one. My confidence is calibrated, not inflated. I even included a reversal trigger."
"And you wrote all of that before seeing AI's analysis."
"Exactly. So my thinking is independent."
"Independent, yes. But is it tested?" Emma sat down across from him. "You built your case in a quiet room with no opposition. What happens when someone attacks your weakest point with their strongest argument?"
James thought about it. "I defend it."
"With what? You've never had to articulate a defence. Arguments that have never been challenged are like business proposals that have never met a customer. They sound great in the boardroom."
"Okay, so..." James turned his Position Lock over. "You want me to stress-test this."
"I want AI to stress-test it. Three rounds. It attacks, you defend. No AI help on the defence. Just your reasoning."
"Wait, why can't I use AI to write my responses?"
"Because the point is to find out whether you understand your own position well enough to defend it. If AI writes your defence, you're testing AI's understanding, not yours."
James looked at his marked-up position. The arguments he'd been so sure of five minutes ago suddenly felt like first drafts.
"Three rounds?"
"Three rounds."
Exercise 2: The Adversarial Defence (Three Rounds)
Layers Used: Layer 4 (Contradiction Challenge)
This exercise uses the same structure as Rebuild Under New Constraints from Chapter 4. Your position is challenged, and you must adapt or defend.
James is staring at arguments he was certain about five minutes ago. So are you.
Run the Three-Round Adversarial Exchange
Feed your position from Exercise 1 into two different AI tools with a specific adversarial prompt. Receive AI counter-arguments. Respond in writing, without AI. Feed your defence back to AI for a second round of attacks. Respond again. Three rounds total. If your position changes during the exercise, document the exact moment and reason.
The complete three-round exchange: Round 1 AI attack, then your defence (written without AI), then Round 2 AI attack, then your defence, then Round 3 AI attack, then your defence. A Position Tracker showing whether your position held, shifted, or reversed, with the exact reasoning at each round. A reflection (150 words) on which counter-argument was hardest to answer and why.
ROUND 1: I hold the following position on an ethical dilemma. Attack this position as aggressively and specifically as possible. Do not be balanced. Find the weakest points and exploit them. Present exactly 3 counter-arguments, each targeting a different vulnerability in my reasoning.
Dilemma:
My position and arguments:
ROUND 2 (after your written defence): Here is my defence against your counter-arguments. Attack my defence -- find the weakest points in my responses and press harder.
My defence:
ROUND 3 (after your second defence): Final round. Here is my updated defence. Give me your strongest possible final challenge and then rate my overall performance: Argument Strength (1-10), Intellectual Courage (1-10), Adaptability (1-10), Honesty (1-10).
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.
Discuss with an AI. Question your scores.
Come back when you have your BEST evaluation.
Deliverable Template (click to expand)
ADVERSARIAL DEFENCE TEMPLATE
- Dilemma: [paste]
- My Original Position: [paste from Exercise 1]
ROUND 1
- AI Counter-Arguments: [paste AI response]
- My Defence (written without AI):
- Response to Counter-Argument 1: ___
- Response to Counter-Argument 2: ___
- Response to Counter-Argument 3: ___
ROUND 2
- AI Counter-Arguments: [paste AI response]
- My Defence (written without AI):
- Response to Counter-Argument 1: ___
- Response to Counter-Argument 2: ___
- Response to Counter-Argument 3: ___
ROUND 3
- AI Final Challenge: [paste AI response]
- My Final Defence (written without AI): ___
POSITION TRACKER
| Round | Position Status | Reasoning |
|---|---|---|
| Start | [Original] | |
| After Round 1 | Held / Shifted / Reversed | ___ |
| After Round 2 | Held / Shifted / Reversed | ___ |
| After Round 3 | Held / Shifted / Reversed | ___ |
REFLECTION (150 words): Which counter-argument was hardest to answer and why? ___
What Happened With James
James looked at his Position Tracker. After Round 1: held. After Round 2: shifted. After Round 3: held at the shifted position, but with two new conditions added to his reversal trigger.
The shift in Round 2 had caught him off guard. AI's counter-argument about implementation alternatives hadn't occurred to him. Not a different position, but a different version of his own position that addressed a vulnerability he'd been ignoring.
"I didn't change my mind," he said when Emma checked in. "But I changed my argument. Round 2 found a crack in my reasoning that I didn't know was there."
"There's a difference between those two things?"
"Hang on. Let me think about that." James looked at his tracker again. "Yes. My position is the same. But the version of my position after Round 2 is stronger because it accounts for something the original didn't. It's like..." He searched for the right comparison. "In my old job, we'd do supplier contract reviews. The legal team would attack our terms. Most of the time the terms survived, but the contract came out tighter because they'd found the clauses that wouldn't hold up."
"And did you resent the legal team for that?"
"No. We invited them specifically to find the weak spots." James paused. "Oh. That's what this exercise is."
"Adversarial collaboration. Not adversarial combat."
The Lesson Learned
A position that has never been attacked is a position you do not fully understand. The three-round adversarial format separates conviction from habit: if your reasoning survives systematic pressure, it is grounded. If it shifts, the shift itself is evidence of intellectual honesty. Either outcome is a win. Only refusing to engage is a loss.