Base-Rate Alternative Prompt¶
Reference-class prompt — instantiates Alternative Hypothesis Generation
Asks how often the leading explanation actually holds in cases like this — dragging the boring, common alternative that the reference class favors onto the table before a vivid but rare story is accepted.
The Base-Rate Alternative Prompt generates one particular, easily-overlooked rival: the explanation that is common in cases like this one. Its distinguishing move among its siblings is that it does not brainstorm freely — it reaches for a reference class and asks how frequently the leading explanation is actually the right one there, then names the base-rate-favored alternative explicitly. "Common things are common" is its whole creed. Where an open generator asks what else is possible, this prompt asks what is usually the case, importing prior frequency as the check on a story that feels compelling because it is vivid or fits a stereotype rather than because it is likely.
Example¶
An allocator reviews a fund that has beaten its benchmark three years running and concludes the manager has genuine skill. Before that conclusion sets, the prompt runs: in the reference class of all comparable funds, how many post a three-year winning streak by chance alone? With thousands of funds in the class, a fair number will string together three good years on luck — so "lucky, and likely to regress to the mean" is the base-rate-favored alternative, and it is not the exciting one. The prompt doesn't deny that the manager might be skilled; it forces the mundane, high-prior explanation onto the table and sets the bar the skill story now has to clear, which is far higher than "three good years" on its own.
How it works¶
- State the focal claim — the explanation currently in the lead.
- Choose the reference class — the set of prior cases genuinely like this one, which is the load-bearing and contestable step.
- Pull the base rate — how often, in that class, the leading explanation actually holds versus the common alternative.
- Name the base-rate-favored rival explicitly and put it on the table as a competitor, with its prior attached.
The heuristic it counters is representativeness: judging a case by how much it resembles a vivid explanation while ignoring how rare that explanation is in the population it is drawn from.
Tuning parameters¶
- Reference-class choice — how the comparison set is drawn. Too broad and the base rate is irrelevant to this case; too narrow and it becomes unstable and easy to gerrymander (the reference-class problem).
- Base-rate source — hard frequency data versus a rough estimate. A guessed base rate beats none, but must not masquerade as measured.
- Prior strength — how much weight the base rate carries against case-specific evidence. Over-weighting it ignores genuine signal in the individual case.
- Trigger — applied to vivid, surprising, or stereotype-fitting claims versus every claim.
- Update discipline — whether the prompt merely names the alternative or feeds an explicit prior into a Bayesian update.
When it helps, and when it misleads¶
Its strength is that it directly counters base-rate neglect — the well-documented tendency to let a representative, vivid story crowd out the common one whose prior probability is far higher.[n1] When the leading explanation is the interesting one, this prompt reliably surfaces the boring rival that is statistically more likely, which is exactly the alternative the reasoner was primed to skip.
Its failure modes both live in the reference class. The class is a choice, and a wrong or self-serving one yields a base rate that misleads rather than corrects; a base rate may also be genuinely unknown, or the case truly novel enough that no clean class applies. The mirror misuse is over-applying the base rate to dismiss real case-specific evidence — "most alerts are false positives," used to wave away the one alert that is a genuine breach. The discipline that guards against this is justifying the reference class out loud and combining the base rate with the case evidence rather than letting either the prior or the particulars dominate alone.
How it implements the components¶
focal_claim_or_leading_hypothesis— the leading explanation is the claim whose prior probability the prompt interrogates.alternative_hypothesis_set— it contributes one targeted, high-prior rival to the field: the base-rate-favored explanation.base_rate_or_reference_class_context— supplying the reference class and its frequencies is the prompt's defining contribution.
It does not map which observations would separate the rivals (discriminating_evidence_map) — that is Discriminating Test Matrix — nor rule any candidate in or out as test results arrive (rival_retention_or_elimination_rule), which is Differential Diagnosis List's job.
Related¶
- Instantiates: Alternative Hypothesis Generation — it supplies the prior-probability check that keeps a vivid story from outrunning its base rate.
- Sibling mechanisms: Differential Diagnosis List · Why Else Could This Be True? Prompt · Discriminating Test Matrix · Counter-Narrative Probe · Red-Team Rival Explanation Review
Editorial Notes¶
Form Classification¶
Form family: Interface, Display & Cue
Rationale: Asks how often the leading explanation actually holds in cases like this — dragging the boring, common alternative that the reference class favors onto the table before a vivid but rare story is accepted, making its operative form a user-facing perceptual surface, prompt, label, or affordance that shapes attention or action.
Independent corroboration: The frozen evidence defines Base-Rate Alternative Prompt as 'Asks how often the leading explanation actually holds in cases like this — dragging the boring, common alternative that the reference class favors onto the table before a vivid but rare story is accepted', so its operative form is Interface, Display & Cue.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Psychology
Origin pattern: Single lineage
Present-day reach: Multi-domain
Rationale: Kahneman and Tversky's judgment research identified base-rate neglect and the corrective value of an explicit reference class.
Related originating lineages:
- Behavioral Economics — Behavioral decision research applies the bias correction to practical choice.
- Medicine & Healthcare — Differential diagnosis uses prevalence to rank common alternatives.
- Statistics & Experimental Design — Bayesian reasoning formalizes the prior prevalence the prompt restores.
Review resolution: Psychology is the agreed primary through base-rate-neglect research. Bayesian statistics, behavioral decision practice, and prevalence-ranked differential diagnosis materially extend the established corrective prompt, which is not newly synthesized here.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
The prompt is only as good as its reference class, and the class is where the arguing should happen. A base rate quoted with a confident number but drawn from the wrong population is more dangerous than no base rate, because it lends a spurious statistical authority to a comparison that doesn't hold. Make the choice of class explicit and contestable, not buried.
[n1] Base-rate neglect is the finding, from Kahneman and Tversky's work on judgment under uncertainty, that people systematically under-weight prior probabilities in favor of case-specific resemblance (representativeness) — as in the taxicab and lawyer-versus-engineer problems, where vivid individuating detail crowds out the far more informative population frequency. This prompt is the deliberate corrective. ↩