Counterfactual Plausibility Filter¶
Reasoning check — instantiates Counterfactual Proximity Signal Calibration
Admits a counterfactual as a valid near-miss only if the better-or-worse alternative was genuinely reachable given what was known at the time, screening out hindsight stories.
A Counterfactual Plausibility Filter is the gate that decides whether a proposed "it almost went differently" deserves any signal at all. It takes a named alternative outcome and interrogates one thing: was that alternative reachable under what was actually known, possible, and constrained at the moment of the event — not what became obvious afterward? If the alternative required information nobody had, an action nobody could take, or a fact quietly changed to make the story work, the filter rejects it or stamps it low-plausibility. Its defining idea is temporal-epistemic honesty: a near-miss is only "near" if the other outcome was live at the time. This is what keeps the whole apparatus from degrading into hindsight fiction.
Example¶
Two aircraft on converging paths get a resolution advisory from their collision-avoidance systems and separate cleanly; afterward a reviewer says the crews were this close to a mid-air collision — "one more turn and it would have been catastrophic." The Counterfactual Plausibility Filter tests that claim rather than accepting the drama. The proposed alternative — the paths actually converging to impact — is named and valued, then screened: given the aircraft's positions, closure rate, and the collision-avoidance system that was already commanding evasive climbs, was "they collide" a genuinely reachable state at that moment? If independent safety layers (the TCAS resolution advisory, still-intact separation minima) made impact effectively unreachable short of several simultaneous failures, the filter marks the "near-collision" low-plausibility — a frightening story, not a true near-miss, and no outsized signal is warranted. If instead those layers had already failed — the advisory ignored, separation gone, only luck left between the two hulls — the alternative was reachable, and the filter admits it as a genuine close call worth a strong signal. Same alarming framing; opposite verdicts, decided entirely by reachability-at-the-time.
How it works¶
The filter runs a named alternative through a reachability screen. It requires the counterfactual to be stated as a valued contrast — a specific better-or-worse state, not a vague "could have gone better" — and then checks it against the decision-time frame on three axes: was the needed information available, was the needed action possible, and does the story hold the fixed facts fixed (no silent editing of what could not be changed). Alternatives that pass are admitted with a plausibility weight; those that fail are rejected or flagged low-plausibility so downstream mechanisms discount them. The filter emits a gate verdict, not a magnitude — it says whether this counterfactual counts, leaving how close and how much to other mechanisms.
Tuning parameters¶
- Reachability strictness — how much benefit of the doubt an alternative gets. Strict settings reject all but clearly-live paths (protecting against hindsight but discarding some real lessons); loose settings admit more (richer learning, more fiction).
- Information-availability standard — whether "knowable in principle at the time" or "actually known to this actor" is the bar. The stricter actual-knowledge bar rejects more clever-in-hindsight stories.
- Plausibility resolution — binary admit/reject versus a graded confidence weight. Grading preserves marginal cases; binary is faster and blunter.
- Fixed-fact discipline — how aggressively the check forbids editing immutable facts to make the counterfactual work.
When it helps, and when it misleads¶
Its strength is that it is the one mechanism standing between calibrated learning and a swamp of self-serving "almosts." It directly counters hindsight bias — the tendency, once an outcome is known, to see the alternative as having been obvious and available all along.[1] Its own failure mode is miscalibrated strictness: set too tight, it rejects genuinely reachable close calls as "you couldn't have known," suppressing real lessons; set too loose, it waves through hindsight stories dressed up as near-misses. The guarding discipline is to fix the evidentiary frame explicitly to the decision moment — what was on the desk then — and to judge reachability against that frozen frame rather than against everything learned since.
How it implements the components¶
eligibility_and_plausibility_gate— it is the gate: the reachability screen that admits, weights, or rejects a counterfactual based on decision-time knowledge and constraints.valued_counterfactual_contrast— it forces the alternative to be named as a specific better-or-worse state before judging it, refusing to gate a vague "could have been different."
It does NOT implement salience_distortion_guard — policing whether a plausible near-miss then draws attention out of proportion to its stakes is Salience Overweighting Check's job; this filter only rules on reachability, not on how much airtime a reachable case later gets.
Related¶
- Instantiates: Counterfactual Proximity Signal Calibration — supplies the plausibility discipline the archetype names as its guard against hindsight storytelling.
- Near-twin, different signal: Counterfactual Plausibility Screen runs the same decision-time reachability, feasibility, and knowability gate, but the two feed different downstream signals. This filter feeds a near-miss / proximity-magnitude signal — how close a live alternative came; the screen feeds a regret-downgrade signal — quieting self-blame once the wished-for path turns out to have been unreachable.
- Sibling mechanisms: Salience Overweighting Check · Almost-Reward Annotation · Close-Call Review Protocol · Counterfactual Value-Delta Table · Near-Miss Distance Scorecard · Near-Miss Response Tier · Proximity Signal Backtest · Regret-Weighted Decision Log · Threshold Band Map
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Counterfactual Plausibility Filter operates as a bounded evaluation of existing evidence or work that produces a finding or disposition because it admits a counterfactual as a valid near-miss only if the better-or-worse alternative was genuinely reachable given what was known at the time, screening out hindsight stories.
Independent corroboration: The frozen evidence defines Counterfactual Plausibility Filter as 'Admits a counterfactual as a valid near-miss only if the better-or-worse alternative was genuinely reachable given what was known at the time, screening out hindsight stories', so its operative form is Assessment, Review & Assurance.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: History & Historiography
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Counterfactual historiography cohered plausibility constraints based on options, knowledge, and capacities genuinely present at the historical moment.
Related originating lineages:
- Philosophy — Possible-world and causal analysis supplies consistency and minimal-departure criteria for nearby counterfactuals.
- Psychology — Hindsight-bias research supplies the decision-time knowledge check that prevents retrospective reachability inflation.
Review resolution: Documented options, capacities, and information available at the historical moment are the core plausibility constraints, so historiography is primary. Philosophy and psychology materially supply minimal-world and hindsight-bias safeguards; the generalized gate is synthetic.
Attribution caveat: The gate combines historiographic minimal-rewrite discipline with philosophical consistency and psychological hindsight checks.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
- American Historical Association: Counterfactual History and the minimal rewrite
- RAND: Counterfactual analysis, plausibility, and closest-world tests
- Stanford Encyclopedia of Philosophy: Counterfactuals
References¶
[1] Hindsight bias (documented by Baruch Fischhoff, 1975) is the "knew-it-all-along" effect: once people learn an outcome, they overestimate how predictable and reachable the alternative was beforehand. It is the specific error a plausibility filter exists to catch. registry ↩