Rule Out the Worst First¶

Why It Matters¶

Naive decision-making optimizes for the most probable explanation. This works when errors are symmetric, but fails catastrophically when low-probability scenarios have asymmetric consequences. A patient with crushing chest pain most likely has acid reflux or anxiety - but if it's actually a heart attack and you miss it, the patient dies. This tool corrects the systematic error of probability-matching when consequences aren't proportional to probabilities.

The Key Move¶

For any situation requiring a decision, explicitly ask: "What's the worst thing this could be?" Then apply a test - however quick and crude - to exclude it. Only after ruling out catastrophic scenarios do you pursue the most likely explanation. The threshold isn't certainty; it's adequate confidence that catastrophe is sufficiently unlikely to justify moving on.

Classic Application¶

Chest pain evaluation. Emergency physicians don't start by treating the most likely cause (gastroesophageal reflux, which accounts for 30-40% of cases). They start with an EKG and troponin to rule out acute coronary syndrome (5-10% of cases), because missing a heart attack is fatal. Only after excluding cardiac causes do they pursue the probable diagnosis.

Surprising Application¶

Software deployment decisions. When releasing code, the most likely scenario is "everything works fine" - that's why you're releasing. But deploying worst-first means checking: "What would break the entire system?" Before deploying new authentication code, you verify the fallback mechanism works, even though the primary code will probably succeed. The surprising insight: in both medicine and software, the correct sequence is counterintuitive - address the unlikely disasters before the likely successes.

Failure Modes¶

Over-application creates paralysis - you can always imagine worse scenarios. The tool requires calibrated judgment about what counts as "worst" and "sufficiently excluded." Misapplication occurs when consequences are actually symmetric (most daily decisions), where probability-matching is correct. It also fails when tests for worst-case scenarios are themselves costly or risky - sometimes the rule-out process causes more harm than the thing you're ruling out.

Go Deeper¶

Croskerry, P. (2003). "The Importance of Cognitive Errors in Diagnosis and Strategies to Minimize Them." Academic Medicine, 78(8), 775-780; Weingart, S.D. & Wyer, P.C. (2003). "The Emergency Medicine Meta-Analysis and Review Course." Emergency Medicine Clinics of North America.