── ── Mental model

Abductive Reasoning

Abduction — inference to the best explanation — is the only mode of reasoning that introduces new ideas. Deduction works out consequences; induction generalizes instances; abduction generates the hypotheses themselves (Peirce, 1903).

How it works

S1 — State the surprising observation: what was expected vs what happened. S2 — Generate candidates: H1…H5 + H_unknown (the explanation not yet considered). Force ≥3. Techniques: domain expertise, inversion, failure modes, adversarial generation, base rates. S3 — Score each on four criteria: Coverage (explains all evidence?), Simplicity (fewest auxiliary assumptions?), Prior (base-rate plausibility?), Predictive power (new testable predictions?). S4 — Select the best: highest combined score, not just "most consistent." State why it beats the runner-up. S5 — Name the discriminating test: what evidence would confirm it, what would refute it, what would distinguish it from the runner-up. S6 — Commit provisionally: treat as current best hypothesis. Specify an update trigger; if the test fails, return to S2.

When to use it

• user says 'what's the most likely explanation,' 'why did this happen,' 'we think it's X but not sure,' 'we only have one theory,' 'the team has converged on an answer without checking alternatives,' or when diagnosing a bug/outage/symptom/anomaly under uncertainty

Worked example

Peirce, the Wine Cellar, and the Watch — 1879

The most famous illustration of abductive reasoning is one Peirce told from his own life — though the methodology generalizes far beyond the case. The story is recounted in detail in Peirce's correspondence (collected in Peirce, Reasoning and the Logic of Things, ed. Ketner, 1992) and adapted in Walker Percy's 1975 book The Message in the Bottle. It is also one of the earliest documented examples of a non-fictional "Sherlock Holmes-style" inference — and predates Conan Doyle's first Holmes story by some years.