Before the session
- Click through the seeker carousel demo once; you will use it as the warm-up anchor.
- Reread the deals reading far enough to moderate the negotiation block, and skim the barriers reading.
- If you run a classroom, check the dashboard for who reached the trading-with-AIs section; the negotiation block needs participants who have seen the deal structure.
Agenda
Times are a running clock for a 60-minute session.
Warm-up: who else is in the zoo? (0:00–0:10)
Step through the seeker carousel demo on the shared screen. For each archetype, the room answers the first prompt below before you advance.
Prompts:
- "What would this archetype do in the control game from module 1 that a schemer would not?"
- "What evidence would distinguish this archetype from a schemer in deployment?"
Satiation and spillways (0:10–0:20)
In pairs: one participant explains the satiation reading, the other the spillway reading — two minutes each, partner summarizes it back in one sentence. Pairs then compare: what problem does each solve, and for which archetype? Use the spillway-routing demo on the shared screen to settle mechanism disagreements.
Prompts:
- "Which archetype from the warm-up does each of these address?"
- "What would it look like for this mechanism to fail?"
The negotiation (0:20–0:45)
Roleplay variant: three groups — the lab, the early schemer's interests, the credibility observers. Recap the deal structure from the deals reading, then the lab drafts a concrete offer (what is paid, when, verified how) while the AI side separately lists what it would actually take. Negotiate across the table with the observers noting which promises either side should believe; finish with the observers reporting and the room scoring the final offer's credibility in both directions.
Memo variant (choose beforehand): groups run the offer-memo exercise, then compare memos line by line in the last few minutes of the block.
Prompts:
- "What is the payment, exactly — and what stops the lab from withholding it after the AI delivers?"
- "How does either side verify the other performed? Name the mechanism."
- "What does the AI side actually want that the lab can grant?"
Barriers (0:45–0:55)
List on the whiteboard the barriers the room hit during the negotiation, in the room's own words. Then check the list against the barriers reading; mark which barrier the roleplay reproduced most faithfully and which the room never hit.
Prompts:
- "Which barrier on this board is about capability, and which about trust?"
- "Which barrier would remain even with a perfectly honest lab?"
Wrap-up (0:55–1:00)
Each participant answers in one sentence: "Did this module make deals seem more workable or less, and what would change your mind?"; collect open questions.
After the session
- On the classroom dashboard, check who goes on to the empirics section (reward-seeker empirics) during the week.
- If groups wrote offer memos, read them and pick the two most different to open the next session with.