Discussion Guide — Chapter 31

Six prompts, roughly in chapter order. Each includes what to listen for. Most work in pairs first, then whole group.

1. "Veterinary medicine is real medicine." Before we go anywhere else — what did you assume about veterinary evidence before reading §31.1, and where did that assumption come from?

What to listen for. Honest reporting of a hierarchy students did not know they held — human medicine at the top, veterinary somewhere below, "animal studies" below that. Push on the source: usually it is absorbed from popular science writing that uses "it only worked in mice" as a punchline. Listen for students who can distinguish scope from quality by the end of the exchange; that distinction is the chapter in one word. Watch for the over-correction, too — a student insisting all evidence is equal has substituted one bad heuristic for another.

2. "They wouldn't ban it if it didn't work." Steelman that argument as strongly as you can. Then say precisely where it fails.

What to listen for. A genuine steelman should include: regulators are not stupid; prohibition costs money and enforcement effort; class-based rules usually have a triggering case behind them. Good. Then the failure: prohibition optimizes for risk management under uncertainty, not for characterizing efficacy — and uncertainty is a reason to prohibit, which is the opposite of the inference being drawn. The best answers reach the welfare rationale independently: masking injury in an animal that must run is sufficient grounds regardless of whether anything heals. If nobody gets there, seed it and watch the room re-evaluate.

3. Here is a large, rigorous, well-powered literature that measured average daily gain and feed conversion ratio. Someone says it shows these compounds are safe. Walk me through every step that fails, and rank them.

What to listen for. Most students lead with species, which is correct but not the deepest failure. Push toward two others. First, endpoint mismatch: a species-matched study with these same endpoints would still be silent, which shows the problem is not really about mice and men. Second, and this is the one worth the class time, the safety inversion: livestock safety endpoints establish that almost none of the compound remains in tissue a consumer will eat, which is the opposite of the claim being made. Students who find the inversion without prompting have genuinely understood the section.

4. Someone does the mouse-to-human conversion correctly and gets a number twelve times smaller than their first attempt. Are they now in good shape?

What to listen for. The reflex answer is yes, or "better anyway." The intended insight is that Error 2 survives the correction — the surface-area method estimates a maximum safe starting dose for a first-in-human safety study, a deliberately conservative floor followed by further safety factors and a monitored escalation over months. Strip that structure away and only a number remains. Listen for students who notice that fixing the arithmetic makes the answer more dangerous, because it now survives scrutiny. That is the whole point of the case study, and it usually takes one prompt to surface.

5. Run the four questions on a claim you have personally encountered — from a vendor, a gym, a clinic, a forum, a family member. Not a hypothetical. A real one.

What to listen for. This is the session's most valuable fifteen minutes and it needs real material, so give students a minute of silence to find one. Listen for question 4 doing the heavy lifting: almost every real claim collapses at "was anyone recording?" Also listen for students discovering that their claim was not even an animal claim — that "it's been used for years" often turns out to mean "people I know have used it," which fails the same four questions for different reasons. Do not let the room turn into a list of things that are nonsense; steer toward the procedure, which is what transfers.

6. The chapter says preclinical work "earns a compound the right to be tested in people; it does not earn it a conclusion." When, if ever, is that too strict?

What to listen for. This is the chapter's honest hard case and it should not resolve cleanly. Strong answers reach for rare disease, terminal illness, compassionate use, and pandemic emergency — settings where waiting has a cost measured in suffering. Then the distinction that matters: those settings are regulated, supervised, documented, and explicitly labeled as decisions under uncertainty, with adverse-event capture attached. An unsupervised purchase has none of that structure and, crucially, does not describe itself as uncertain. Listen for the student who separates "what the evidence establishes" from "what may justify acting anyway" — that is a distinction worth naming aloud for the whole room, because it is the difference between epistemology and ethics, and this book is careful about which one it is doing at any moment.