Discussion Guide — Chapter 14

Prompt 1 — "The finding was real and the inference was wrong. Which is more dangerous for public understanding: a fabricated finding, or a real finding with an unsupported inference attached?"

Listen for: recognition that fabrications are self-correcting because they fail replication, while stretched real findings are durable precisely because the underlying result keeps checking out. Strong students will name the defensive asymmetry — every challenge to the anti-aging claim can be answered by pointing at Rudman, which is true. Redirect if: the room converges on "both are bad." Push for a mechanism of persistence.

Prompt 2 — "Whether the age-related decline in growth hormone is pathology or normal physiology is contested. Argue both sides, then say what would move you."

Listen for: actual use of both evidence bases — adult deficiency is a real syndrome with real morbidity, and reduced growth signaling extends lifespan in model organisms. Listen especially for: whether they notice that "what would move me" is hard to answer here, because the question may not be empirically decidable in the form it is usually asked. Redirect if: someone declares it settled. Ask which evidence they are discarding and why.

Prompt 3 — "Acromegaly establishes a direction of risk but not a magnitude. Is a direction without a magnitude actually useful for making a decision?"

Listen for: the recognition that direction plus organ systems tells you what to monitor, which is decision-relevant even without a number — this is the practical payoff and students often miss it. Also listen for the honest counterargument: an unquantified risk is easy to inflate or dismiss to taste, and both happen. Best outcome: the room arrives at "it tells you what to watch, not whether to act," which is exactly §14.9's position.

Prompt 4 — "Growth hormone has the best mechanistic story in this book and still gets ❌ for the anti-aging claim. Is 'never upgrade with mechanism' too strict?"

Listen for: engagement with the strongest counterargument — that mechanism legitimately guides priors, resource allocation, and which trials to run. The rule does not say mechanism is worthless; it says mechanism is not evidence of effect. Watch for: students who defend the rule by asserting it. Push them to the ~9-in-10 attrition figure and ask what that number means about how often good mechanisms are wrong. Advanced: ask whether the rule should apply symmetrically to harms — it does, and §14.7 shows the work.

Prompt 5 — "A person with no deficiency has used growth hormone for two years, feels better, and is not asking your permission. What do you say?"

Listen for: taking the reported experience seriously, naming which claim they are relying on without judgment, distinguishing what they can see from what matters, and offering concrete monitorables. Fail conditions to name explicitly: anything protocol-shaped, and any moralizing. The most valuable discussion usually comes from asking why moralizing is a failure rather than merely impolite — answer: it ends the conversation, and the person then gets their information from someone selling something.

Prompt 6 — "Who is responsible for what happened after 1990 — the authors, the journal, the press, the clinics, or the readers?"

Listen for: resistance to a single villain. The authors reported what they found. The journal published a competent study and later publicly noted its misuse. The clinics acted on incentives that were legal and lucrative. The readers did what readers do. Best outcome: the room concludes that the failure was systemic — no single actor behaved outrageously and the outcome was still bad — and then asks what structural change would have prevented it. Push toward: required scope statements in citation, or the fact that nobody had a commercial reason to run the trial that would have settled it. That second point is the more uncomfortable one and the more important.