A short story describes a morally ambiguous conflict — war, betrayal, loyalty, revenge — involving several characters whose actions are defensible or indefensible depending on your perspective. You rank them individually. Then you negotiate a group consensus. Then the game changes: each team is assigned specific positions and point values, and suddenly you're arguing passionately for moral judgements you don't actually hold. The exercise makes the machinery of moral reasoning uncomfortably visible.

Setup

The exercise has three stages. Don't reveal Stage 3 until Stage 2 is complete.

Stage 1 — Individual ranking (5 mins).

Read the scenario aloud or distribute it. Each participant privately ranks every character as either honourable or dishonourable. No discussion. No justification needed. Just gut judgement.

The scenario should involve morally ambiguous characters whose actions can be read multiple ways. The classic version uses a territorial conflict between nations: a president who invades to feed his starving people, a king who retaliates, a prince with a legitimate claim who takes hostages, a friend who refuses to help, a brother who murders to restore justice. Every character has a defensible case and an indefensible one.

Stage 2 — Within-group consensus (10–15 mins).

Split into small teams (2–4 per team). Each team must reach unanimous consensus on every character: honourable or dishonourable. Track what happens — do they debate the meaning of "honourable"? Does the loudest voice win? Does anyone change their mind, and why?

Bring the teams back together. Write each team's rankings on the board. Discuss briefly: how was consensus reached? Rules, persuasion, or conformity?

Stage 3 — Cross-group negotiation (15–20 mins).

Distribute team scoring sheets. Each team receives assigned H/D positions for every character, plus point values. Teams have different assignments — one team scores highly if Adolf is honourable, another if he's dishonourable. Point values vary, so some positions matter more than others.

The instruction is simple: maximise your team score. Do not reveal your scoring sheet to other teams.

Give teams a few minutes to strategise, then mix groups so each new group has a representative from every team. The mixed groups negotiate — the goal is to agree on a final ranking, but every representative is now arguing for assigned positions, not personal beliefs, and their arguments are weighted by hidden stakes.

Reassemble and debrief.

Questions

What changed?

• How did your rankings in Stage 1 compare to what you argued in Stage 3?
• The characters didn't change. The story didn't change. What changed?
• Did you notice yourself generating moral arguments for positions you don't actually hold? How convincing were they?

How did consensus form?

• In Stage 2, how did your team resolve disagreements? Did you negotiate the meaning of "honourable" or just trade positions?
• In Stage 3, how was the negotiation different? What happened when someone pushed a position that contradicted their Stage 1 ranking?
• Who was most persuasive in each stage, and why?

What does this reveal about moral reasoning?

• If you can generate a convincing moral argument for a position you don't hold, what does that tell you about moral arguments you do hold?
• When your incentive changed (Stage 3 scoring), how quickly did your reasoning follow? Was there a lag, or did it snap into line?
• Where in your own experience have you seen people argue moral positions that are actually driven by incentives?

Guidance

Stage 1 reveals the prediction engine's moral pattern-matching. Participants produce gut rankings fast — honourable or dishonourable — before they can articulate why. The characters are designed to be ambiguous enough that different participants' prediction engines match different patterns. The divergence in Stage 1 rankings is the raw material: it shows that moral intuition is pattern-matching against personal experience, not access to objective truth.

Guidance

Stage 2 forces the lazy controller to generate justifications for gut rankings. Watch for post-hoc rationalisation: people rarely say "I just feel he's dishonourable." They produce reasons — but the reasons arrived after the judgement. Stage 3 makes this starkly visible. Participants generate equally fluent moral arguments for assigned positions they don't hold, demonstrating that the lazy controller's rationalisation machinery is indifferent to the direction it's pointed — it will justify whatever position it's given.

Guidance

Stage 3 is the affordance shift made visceral. When the scoring sheets arrive, the environment changes. The same characters, the same story, the same participants — but now the affordance landscape is different. Arguing that Adolf is honourable has become high-utility (6 points) even if you ranked him dishonourable five minutes ago. The input-output machine doesn't care about moral consistency; it evaluates the current affordance. Watch how quickly behaviour follows the new incentive structure. That speed is the mechanism working.

Guidance

The cross-group negotiation in Stage 3 puts representatives under competing social pressures. Each representative carries a mandate from their team (maximise our score) into a mixed group where everyone has different mandates. Persuasion, coalition-building, and strategic concession all become visible. The social mapping mechanism is doing the work: who has influence? Whose arguments carry weight? Does the representative with the highest-stakes position argue hardest, or does the most socially skilled person dominate regardless of stakes?

Guidance

Stage 2 is the clean case: a genuine group forming a genuine consensus. Some groups use rules ("majority wins"), some use debate, some defer to the loudest voice. The mechanism is conformity under social pressure — even with only four people, the pull toward agreement is strong enough that individuals abandon positions they held moments ago. Compare this to Stage 3, where the group dynamic shifts: now everyone is performing a position rather than holding one. The conformity mechanism still runs — people still converge — but what they're converging on has been shaped by hidden incentives rather than shared belief.