Regret-Weighted Decision Log¶
Documentation artifact — instantiates Counterfactual Proximity Signal Calibration
A running ledger of decisions, each tagged with a regret weight that counts only when the better alternative was genuinely available at the time.
A Regret-Weighted Decision Log is a standing record of decisions, kept over time, where each entry carries three things: what was actually decided and how it turned out, whether a better path was genuinely available at the moment of choosing, and — only if it was — how much better that forgone path would have been. Its defining feature is the availability gate on the regret column: an entry earns a regret weight only when the better alternative was reachable with what the decider knew and could do then. Decisions that turned out badly but where no better option was actually available are logged with zero regret, not punished for bad luck. It is a documentation artifact built to make decision quality reviewable across many choices without contaminating it with hindsight.
Example¶
An NBA front office keeps a Regret-Weighted Decision Log of its draft picks. Three years after a draft, they revisit the entry for the 24th pick. The factual anchor is unchanged: they selected Player X, who became a serviceable rotation player. The log then asks the gated question — at pick 24, was a clearly better player actually available and identifiable with the information the team had then? If the future All-Star who went 25th had a glaring, knowable red flag that every board including theirs had flagged, the "we should have taken him" story fails the availability gate: it required hindsight, and the entry gets a low regret weight. If instead their own pre-draft board had that player ranked above X and they deviated for a since-discredited reason, the alternative was available, and the entry earns a high regret weight scaled by the value gap between an All-Star and a rotation player. Over a decade of picks, the regret-weighted column shows where the process — not the luck — actually cost them.
How it works¶
Each entry is built in three passes. First the factual anchor: record the decision, the outcome, and the decision-time context, before any alternative is scored. Second the availability gate: judge whether the better alternative was genuinely reachable given what was known and possible then; if not, the regret weight is zero regardless of how the outcome looks in retrospect. Third, for entries that pass the gate, the value-delta frame: quantify how much better the forgone path would have been, in a stated value frame, and use that magnitude as the regret weight. The log accumulates these across many decisions so patterns — persistent, gated regret in a certain kind of call — become visible. It records and weights; it does not itself prescribe the fix.
Tuning parameters¶
- Availability bar — how demanding the "genuinely reachable" test is. A high bar credits regret only to clearly-live alternatives (protecting against hindsight); a low bar logs more regret but admits more retrospective invention.
- Value frame for regret — the units in which "how much better" is scored. Sets what the regret column actually measures and lets entries be compared.
- Review cadence — how long after the decision the regret weight is assessed. Longer waits give truer outcomes but weaker memory of the decision-time frame.
- Aggregation window — whether patterns are read per-decider, per-category, or across the whole book, which changes what the accumulated regret can reveal.
When it helps, and when it misleads¶
Its strength is separating process from luck across a portfolio of decisions: by gating regret on genuine availability, it credits bad decisions that could have gone better and spares good decisions that simply didn't — the exact discrimination raw outcome tallies destroy. It is a structural counter to outcome bias, the tendency to judge a decision's quality by how it happened to turn out.[1] Its failure mode is that the availability gate is itself judged in hindsight and can be quietly relaxed to manufacture (or to excuse) regret, and a tidy regret number can lend false precision to a soft counterfactual. The guarding discipline is to fix the decision-time evidence frame explicitly per entry, judge availability against that frozen frame, and — where reachability is genuinely contested — send the call to a Counterfactual Plausibility Filter rather than settling it in the log.
How it implements the components¶
factual_outcome_anchor— every entry records the actual decision, outcome, and context first and keeps that record unaltered.eligibility_and_plausibility_gate— the regret weight is gated on whether the better alternative was genuinely available at decision time; failing the gate zeroes it.value_delta_frame— for gated-in entries, the regret weight is the stated-frame magnitude of how much better the forgone path would have been.
It does NOT implement valued_counterfactual_contrast as its organizing artifact — enumerating and naming the field of plausible alternatives is Counterfactual Value-Delta Table's job; this log records one decision per row and gates a single regret weight, rather than cataloguing the alternative set.
Related¶
- Instantiates: Counterfactual Proximity Signal Calibration — supplies the persistent, availability-gated record that keeps regret signals honest across many decisions.
- Consumes: Counterfactual Plausibility Filter adjudicates contested "was it genuinely available?" calls.
- Sibling mechanisms: Counterfactual Plausibility Filter · Counterfactual Value-Delta Table · Almost-Reward Annotation · Close-Call Review Protocol · Near-Miss Distance Scorecard · Near-Miss Response Tier · Proximity Signal Backtest · Salience Overweighting Check · Threshold Band Map
Editorial Notes¶
Form Classification¶
Form family: Record, Log & Register
Rationale: Regret-Weighted Decision Log operates as a persistent ledger, log, register, or case record that preserves history and traceability because it a running ledger of decisions, each tagged with a regret weight that counts only when the better alternative was genuinely available at the time.
Independent corroboration: The frozen evidence defines Regret-Weighted Decision Log as 'A running ledger of decisions, each tagged with a regret weight that counts only when the better alternative was genuinely available at the time', so its operative form is Record, Log & Register.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Behavioral Economics
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Weighting only feasible counterfactuals follows regret-based behavioral decision theory; operations research formalizes opportunity comparison, psychology supplies experienced regret, and management supplies the running ledger.
Related originating lineages:
- Operations Research — Decision-analysis logs provide structured option and availability evidence.
- Organizational & Management Science — Decision logs materially supply the longitudinal governance and review artifact.
- Psychology — The psychology lineage materially shaped Regret Weighted Decision Log through memory, judgment, emotion, and behavior.
Review resolution: The blind reviewers disagreed on primary lineage. Light authoritative research resolves the defining form in favor of behavioral_economics: Weighting only feasible counterfactuals follows regret-based behavioral decision theory; operations research formalizes opportunity comparison, psychology supplies experienced regret, and management supplies the running ledger. The rejected primary is retained only when it materially shaped the mechanism, and present-day breadth is recorded separately as domain_reach=multi_domain.
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:
References¶
[1] Outcome bias (Baron & Hershey, 1988) is the tendency to judge the quality of a decision by the quality of its outcome, even when the outcome hinged on information unavailable at the time. Gating regret on decision-time availability is the log's structural defense against it. registry ↩