Choice-Supportive Bias¶

Explain why we misremember past decisions in our own favor: dissonance reduction rewrites the stored memory so the chosen option looks better and the rejected ones worse than the evidence actually showed at the time of choice.

Core Idea¶

Choice-supportive bias is the tendency for people's memories of a past decision to shift systematically in favor of the option they chose: in retrospect, the chosen option is remembered as having had more positive attributes and fewer negative ones than it actually did at the time, while the rejected options are remembered as having had more flaws than the contemporaneous evidence showed. Mather, Shafir, and Johnson (2000) demonstrated this by having participants choose between options described by explicit attribute lists and then, after a delay, recall the attributes: the recalled profiles were distorted in the direction of the choice — chosen options appeared better, rejected ones worse — even when participants were instructed to recall accurately.

The mechanism is cognitive dissonance reduction operating on the memory substrate rather than on current belief. After a decision is made — especially an irreversible, high-involvement one made under uncertainty — the discrepancy between having committed to one option and the remembered existence of attractive features in the rejected ones creates dissonance. One resolution is to retroactively revise the remembered attribute profile until the choice appears to have been obviously correct: the chosen option had more going for it, the rejected ones had more flaws, and the decision was not the close call it actually was. Because this revision operates on memory rather than on stated current belief, and because memory of past evidence is not independently checked, the bias is particularly hard to detect from the inside: the person sincerely reports what they remember, and the memory has already been altered. The practical implication is that post-hoc evaluations of past decisions — "was that choice right?" — are a contaminated signal wherever the choice itself is the thing being evaluated.

Structural Signature¶

Sig role-phrases:

the choice under uncertainty — a decision made between two or more options, especially high-involvement and irreversible, the commitment that triggers the process
the contemporaneous attribute profile — the balance of pros and cons of each option as it actually stood at the time of choice, the record later distorted
the dissonance pressure — the discomfort between having committed to one option and remembering attractive features in the rejected ones, plus identity-protective motivation
the retrospection process — later recall, narration, or justification of the decision, the occasion on which the rewrite surfaces
the memory-substrate rewrite — dissonance reduction acting on stored memory (not on stated current belief), retroactively revising the remembered profile
the directional tilt — chosen option remembered as better and rejected ones as worse than the contemporaneous evidence showed, so the close call looks obviously correct
the inside-undetectability — the person sincerely reporting altered memory and feeling the superiority "was there all along," so sincerity is no evidence of accuracy and accuracy instructions do not remove the tilt
the dissonance moderators — decision difficulty, involvement, and irreversibility sizing the magnitude of the swing
the record-at-decision-time remedy — trustworthy evaluation resting on evidence fixed before the dissonance acts (decision logs, pre-committed criteria, pre-mortems) rather than later recall

What It Is Not¶

Not an ordinary change of current belief. The bias rewrites the remembered attribute profile as it stood at the time of choice — the prospective balance sheet — not merely the present verdict. The person does not feel they updated; they feel the chosen option's superiority was there all along, which is precisely what makes it harder to catch than a change of mind.
Not evidence that the chosen option was actually better. Inflated post-choice satisfaction must be partitioned into a genuine-quality component (new information) and a memory-distortion component (dissonance reduction). Crediting the whole of it to the option assumes away the very tilt the bias names; the decision may have been the close call the records show, not the obvious one memory reports.
Not removable by trying to recall accurately. The tilt persists under explicit accuracy instructions, because the revision has already happened to the stored memory — exhortation reaches the wrong target. Only externalizing the record before dissonance acts (decision logs, pre-committed criteria, pre-mortems) prevents the contamination; a sincere intention to remember faithfully does not.
Not confirmation bias or hindsight bias. Confirmation bias is forward-looking — biased seeking and weighting of new evidence; hindsight bias distorts the remembered predicted likelihood of an outcome. Choice-supportive bias is backward-looking and operates on the remembered attributes of the compared options, so a remedy tuned to one is not licensed for the others.
Not a property of any system that sticks with its choice. The rewriting requires an agent with autobiographical memory and identity-protective motivation; a market price, a thermostat, or an organism without a self-model does not retroactively revise its history. What recurs in such systems, if anything, is cognitive dissonance / motivated reasoning generally — invoking this bias for them overreaches its substrate.

Scope of Application¶

Choice-supportive bias lives in one domain — human judgment-and-decision-making and autobiographical memory — and the contexts below are application settings of that single substrate (an agent rewriting its remembered record under dissonance), not structurally distinct systems. The cognitive-dissonance parent it instantiates travels to any dissonance-bearing agent, but the bias as named requires a remembered, self-relevant past; uses for non-recalling systems are overreach and stay out of this map.

Consumer behavior — post-purchase loyalty and brand stickiness, the chosen product remembered more favorably even when it was demonstrably suboptimal.
Voter behavior — rising approval of the candidate one voted for after the election, regardless of subsequent performance.
Medical decision-making — the chosen treatment remembered as more clearly indicated than the contemporaneous evidence supported, contaminating retrospective decision review.
Career and life-decision retrospection — the foregone path (school, job, partner) remembered with flaws the evidence at the time did not show, the close call recalled as obvious.
Decision-quality and post-mortem practice — the methodological corollary deployed across the above: trustworthy evaluation rests on records fixed at the time of choice (decision logs, pre-committed criteria, pre-mortems), never on later recall.

Clarity¶

Naming choice-supportive bias separates two distortions that post-decision satisfaction blurs together: a memory effect, in which the remembered attributes of the options shift after the choice, and a belief effect, in which one's current assessment changes. The distinction matters because the bias operates retroactively on the perceived evidence as it stood at the time of choice — it rewrites the prospective balance sheet, not just the present verdict — which makes it far harder to catch than an ordinary change of mind. A person whose memory has tilted does not feel they have updated; they feel the chosen option's superiority was there all along. Without the label, that sincerity reads as testimony; with it, the analyst knows the witness and the record are the same contaminated source.

The sharper question the concept makes askable is causal: when someone is more satisfied with a past choice than the original evidence warranted, did the chosen option genuinely turn out better (new information), or did they rewrite the remembered attributes to make the decision feel right (dissonance reduction)? That separation lets a clinician, market researcher, or election analyst partition post-choice satisfaction into a real-quality component and a memory-distortion component instead of crediting all of it to the option. And it carries a direct methodological consequence: because the contamination targets retrospection specifically, any honest evaluation of a decision must rest on evidence recorded at the time of choice (pre-mortems, fixed criteria, decision logs) rather than on later recall — the bias tells you exactly which signal not to trust.

Manages Complexity¶

Post-decision satisfaction appears across consumer loyalty, voter approval, treatment-choice confidence, and career retrospection, and each setting could be studied as its own puzzle about why people stand by what they picked. Choice-supportive bias compresses them by locating one recurring distortion at a fixed point in the decision's timeline: the remembered attribute profile is rewritten after commitment — chosen option better, rejected ones worse than the contemporaneous evidence — by dissonance reduction acting on the memory substrate rather than on current belief. An analyst then needs only to partition any post-choice satisfaction into two components (genuine option quality from new information versus retroactive memory tilt) and to read off where the tilt will be largest from the same dissonance moderators everywhere: decision difficulty, involvement, irreversibility. That collapses a catalogue of "why are they still happy with that choice?" cases into one memory-rewrite structure with known amplifiers. It also fixes the methodological corollary uniformly — because the contamination targets retrospection specifically, trustworthy evaluation must rest on evidence recorded at the time of choice (decision logs, fixed criteria, pre-mortems), never on later recall — so the bias tells the practitioner, across all these arenas at once, exactly which signal to discard.

Abstract Reasoning¶

The bias licenses a set of inferences about when a retrospective evaluation is contaminated, what to change to get an honest one, and how large the distortion will be in advance.

Diagnostic — from inflated post-choice satisfaction back to a rewritten record. The governing move runs from a surface signature (someone recalls the option they chose as clearly superior, the rejected ones as clearly flawed) to a hidden cause (dissonance reduction has revised the remembered attribute profile, not merely the current verdict), against the naive reading that the recollection is honest testimony about how the evidence stood. The tell is direction plus timing: a memory tilted toward the chosen option, appearing after commitment and absent at the time of choice, is the fingerprint of choice-supportive memory drift rather than of new information. The discriminating fact that makes this hard to catch from the inside — and therefore the thing the diagnostician must supply from outside — is that the person does not feel they updated; they feel the superiority was there all along, so their sincerity cannot be used as evidence the memory is accurate.

Interventionist — what to change, and the predicted effect on evaluation quality. Because the contamination targets retrospection specifically, operating on the memory substrate rather than on stated belief, the corrective lever is to move the evaluative evidence off recall and onto a record fixed at the time of choice: decision logs, pre-committed criteria, pre-mortems. The prediction is directional — an evaluation anchored to contemporaneous records yields the close call the decision actually was, while one resting on later recall yields a decision that looks obviously correct, and the gap between them measures the drift. The model is equally specific about what will not work: instructing a person to recall accurately does not remove the tilt (it persists under explicit accuracy instructions), because the revision already happened to the stored memory, so exhortation reaches the wrong target. Only externalizing the record before the dissonance acts prevents the contamination.

Boundary-drawing — which evaluations to distrust, and what does not transfer. The bias marks a sharp line through evaluation signals: any judgment of a decision that draws on retrospective recall of the option attributes sits in the contaminated regime wherever the choice itself is the thing being evaluated, while a judgment built on the recorded contemporaneous evidence does not. So the prior question before crediting a "was that the right choice?" report is whether the signal is recall or record. The model also fixes its scope against substrate: it requires an agent with autobiographical memory and identity-protective motivation, so it predicts no such rewriting in a market price, a thermostat, or any system without a self-model and a remembered past — those do not retroactively revise their histories, and invoking the bias for them would overreach. It further bounds itself against neighbors that license different inferences: this is the backward-looking memory expression of dissonance acting on past evidence, distinct from confirmation bias (forward-looking, in seeking and weighting new evidence) and from hindsight bias (which distorts the remembered predicted likelihood of an outcome, not the attributes of compared options) — so a remedy or prediction tuned to this bias is not automatically licensed for those.

Predictive — anticipating the magnitude from the dissonance moderators. Holding the same amplifiers everywhere — decision difficulty (close, competing attributes), chooser involvement, and irreversibility — the model predicts in advance where the tilt will be largest without studying each arena separately: a hard, high-involvement, irreversible choice will show a large retrospective swing, an easy or reversible low-stakes one little. That lets an analyst pre-rate which post-choice satisfaction signals are most contaminated (a major irreversible life or treatment decision) versus least (a trivial reversible purchase), and to partition the observed satisfaction into a genuine-quality component and a memory-distortion component sized by how strongly the moderators are present, rather than crediting all of it to the option.

Knowledge Transfer¶

Within judgment-and-decision-making and memory research the bias transfers as mechanism, intact, with the caveat that its "domains" are application contexts of one substrate — a human agent with autobiographical memory revising its record under dissonance — not structurally distinct systems. With that understood, the memory-rewrite diagnosis (chosen option remembered better, rejected ones worse than the contemporaneous evidence, by dissonance reduction acting on the memory substrate) and its moderators (decision difficulty, involvement, irreversibility) carry without translation across consumer behavior (post-purchase loyalty and brand stickiness even for demonstrably suboptimal choices), voter behavior (rising approval of the candidate one voted for, regardless of performance), medical decision-making (the chosen treatment remembered as more clearly indicated than the evidence supported), and career and life retrospection (the foregone path remembered with flaws the contemporaneous evidence did not show). The vocabulary (memory effect versus belief effect, retrospective tilt, the rewritten prospective balance sheet), the diagnostic (direction plus timing — a tilt toward the chosen option appearing after commitment; sincerity is no evidence of accuracy), and the intervention (move evaluation off recall and onto records fixed at the time of choice — decision logs, pre-committed criteria, pre-mortems; note that instructing accurate recall does not remove the tilt) all move freely, because the same dissonance-acting-on-memory machinery operates in each.

Beyond such agents the bias does not travel as mechanism; what travels is the more-general parent it instantiates. The bias is precisely the memory-side expression of cognitive dissonance reduction (under the broader motivated reasoning umbrella), and that parent recurs across any agent under pressure to perceive its commitments as coherent — but the bias as named requires autobiographical memory plus identity-protective motivation, and so vanishes in systems that lack a remembered, self-relevant past: a market price, a thermostat, or an evolved organism without a self-model does not retroactively rewrite its history, and invoking the bias for them would overreach. So when the lesson is wanted for a non-human or non-recalling system, the load-bearing content is cognitive dissonance / motivated reasoning, not "choice-supportive bias," whose own cargo (the post-choice attribute-recall distortion, the dissonance moderators, the record-at-decision-time remedy) stays bound to human retrospection. The bias must also be held distinct from the crowded family it is easily merged with — confirmation bias (forward-looking, in seeking and weighting new evidence), hindsight bias (which distorts the remembered predicted likelihood of an outcome, not the attributes of compared options), rosy retrospection (a general positive past-distortion of which this is the choice-contingent variant), and the endowment effect (which moves current ownership-conditioned valuations, not remembered ones) — so a remedy or prediction tuned to this bias is not automatically licensed for those. The honest cross-domain account: as mechanism it stays inside human autobiographical-memory cognition; the cognitive-dissonance parent travels to any dissonance-bearing agent as a co-instance; and the right move elsewhere is to carry that parent, with "choice-supportive bias" kept as the memory-side JDM label (see Structural Core vs. Domain Accent).

Examples¶

Canonical¶

The demonstrating study is Mather, Shafir, and Johnson's "Misremembrance of Options Past" (2000). Participants chose between options — such as two hypothetical job candidates or apartments — each described by an explicit list of positive and negative features. After a delay, they were tested on which features had belonged to which option. Their memories were systematically distorted in the direction of the choice: positive features were disproportionately (mis)attributed to the option they had selected, and negative features to the option they had rejected, so the chosen option was remembered as a clearer winner than the balanced feature lists actually made it. The distortion appeared even though participants were asked to report accurately, and it targeted the remembered attributes — the record of how the evidence stood — rather than merely their current preference.

Mapped back: Selecting between the feature-listed options is the choice under uncertainty; the balanced pro/con lists are the contemporaneous attribute profile. Misattributing positives to the chosen and negatives to the rejected option is the memory-substrate rewrite producing the directional tilt. That it persisted despite instructions to recall accurately is the inside-undetectability — the revision had already happened to the stored memory, so honest intent could not reach it.

Applied / In Practice¶

Professional decision-makers combat the bias with decision journals. In fields from institutional investing to clinical medicine, practitioners are urged to record, at the moment of a consequential decision, the options considered, the evidence for each, the expected outcome, and the confidence held — Daniel Kahneman among others has recommended the practice explicitly. The point is to create a contemporaneous record that choice-supportive memory cannot later rewrite. When the decision is reviewed months afterward, the journal, not recollection, supplies the balance sheet as it actually stood; a portfolio manager who "remembers" having been certain about a winning trade can be confronted with a logged entry showing it was a coin-flip. This externalizes evaluation off the contaminated memory substrate and onto a fixed record — the one intervention the model predicts will work, since instructing people to simply remember honestly does not remove the tilt.

Mapped back: The journal is the record-at-decision-time remedy, capturing the contemporaneous attribute profile before the dissonance pressure can act. Reviewing months later is the retrospection process at which the memory-substrate rewrite would otherwise surface. Because the inside-undetectability means accuracy instructions fail, only this externalized record — not sincere intention — restores the close call the decision actually was.

Structural Tensions¶

T1: Adaptive comfort versus corrupted signal (the bias serves the agent and misleads the analyst). The very rewrite that contaminates a decision record also protects the chooser: reframing a close call as an obvious win reduces regret, stabilizes commitment, and lets a person act rather than endlessly second-guess an irreversible choice. The dissonance reduction is functional for the agent living with the decision even as it destroys the evidential value of their retrospection for anyone auditing decision quality. There is no separate "healthy" and "pathological" version to pry apart — the same memory tilt is a coping mechanism and a data corruption, evaluated from two seats. Treating the bias as pure defect ignores the load it carries for the chooser; trusting the comforted memory as testimony ignores that the comfort was bought by falsifying the record. Diagnostic: Is the goal here to help this person live with the choice (the tilt is doing useful work) or to assess whether the choice was sound (the tilt must be routed around)?

T2: Genuine quality versus memory distortion (two indistinguishable sources of the same satisfaction). Inflated post-choice satisfaction has two components — the option really did turn out well (new information) and the remembered attributes were rewritten (dissonance) — and from inside the report they look identical: the person sincerely recalls the chosen option as superior in both cases. Partitioning them is the whole analytic payoff, yet the bias guarantees the one clean way to do it (contemporaneous records) is precisely what most decisions lack. Credit all the satisfaction to the option and you assume the tilt away; credit it all to the bias and you deny real learning from outcomes. The concept sharpens the question but cannot answer it without a record made before dissonance acted. Diagnostic: Is there evidence fixed at the time of choice against which today's remembered profile can be differenced — or is the satisfaction an unpartitionable blend?

T3: Sincerity as evidence versus sincerity as noise (the honest witness is the unreliable one). Ordinarily a person's confident, sincere report is a reason to believe them; here it is the opposite. Because the revision happened to the stored memory, the chooser feels the superiority "was there all along" and reports it in perfect good faith — so sincerity, the usual credibility signal, becomes uninformative about accuracy and even a marker of successful rewriting. The tension is that the more genuinely convinced the person is, the less their conviction can be used, and confronting them tends to be read as calling them a liar when they are not one. The analyst must discount exactly the testimony that feels most trustworthy. Diagnostic: Am I weighting this recollection by how confidently and sincerely it is held — when that confidence is evidence of the tilt, not of the facts?

T4: Externalized record versus the cost of keeping one (the only remedy is the one nobody maintains). The model is unusually prescriptive — decision logs, pre-committed criteria, and pre-mortems are the single intervention predicted to work, because exhorting accurate recall provably does not. But that remedy demands foresight and discipline at the moment of choice, before anyone knows the decision will need auditing, and for the vast majority of everyday choices no such record exists or ever will. So the concept's actionable advice applies most cleanly exactly where it is least likely to have been followed, and its diagnostic power over undocumented past decisions remains largely negative: it can tell you the recall is contaminated but cannot reconstruct the clean signal. Diagnostic: Was this decision important enough, and foreseeable enough, to have justified a contemporaneous record — and if not, must I treat its retrospection as simply unrecoverable?

T5: Sharp neighbors versus a blurred family (a distinction that only holds if you police it). The concept earns its keep by being not confirmation bias (forward-looking evidence seeking), not hindsight bias (distorted remembered likelihood), not the endowment effect (current ownership valuation), and only the choice-contingent slice of rosy retrospection. Those lines license different remedies. Yet in practice the family collapses easily: a real episode of motivated retrospection often carries several at once, and analysts reach for whichever label is nearest. Holding the distinction tight buys correct interventions but demands constant discrimination the messy case resists; letting it blur buys a usable single "people rewrite the past" heuristic at the cost of mis-tuned fixes. Diagnostic: Is the distortion here on the attributes of the compared options (this bias) or on likelihood, new-evidence weighting, or current valuation — which would send me to a different remedy?

T6: Autonomy versus reduction (a named JDM bias or an instance of cognitive dissonance). "Choice-supportive bias" is a canonically studied effect with its own signature — Mather, Shafir, and Johnson's misattribution paradigm, the moderators of difficulty/involvement/irreversibility, the record-at-decision-time remedy — and within human judgment-and-decision-making that apparatus travels intact across consumer, voter, medical, and career retrospection. But those are application contexts of one substrate, not structurally distinct systems, and beyond an agent with autobiographical memory the effect does not travel as mechanism at all. What carries is the parent it instantiates: cognitive dissonance reduction under the broader motivated reasoning umbrella, the memory-side expression of an agent's pressure to see its commitments as coherent. A system without a remembered, self-relevant past — a market price, a thermostat — cannot exhibit the bias, only, if anything, the dissonance parent. Diagnostic: Resolve toward the parent (cognitive dissonance / motivated reasoning) when asking what recurs in any commitment-bearing agent; toward the named bias when diagnosing a person's distorted recall of a specific past choice.

Structural–Framed Character¶

Choice-supportive bias sits in the middle of the spectrum — best read as mixed, on the same footing as the certainty effect: a genuine cognitive mechanism pulled toward the framed side by a distortion framing and a tight substrate confinement. On the structural side it is a real regularity: institutional_origin points structural — the retroactive memory tilt is a discovered feature of human autobiographical memory (Mather, Shafir, and Johnson, 2000), not an artifact any agency legislated — and it runs in the memory substrate observer-free, so it is only mildly human_practice_bound, the rewrite happening whether or not a psychologist studies it (indeed it persists even under instructions to recall accurately). Within its substrate, cross-context reuse is recognition, not import: consumer loyalty, voter approval, treatment-choice confidence, and career retrospection are one substrate (an agent revising its remembered record under dissonance) at different settings, the same memory-rewrite diagnosis and moderators carried intact.

What pulls it toward the framed side is two features. First, a mild evaluative_weight: the effect is framed as a bias — a "contaminated signal," a memory distortion to be routed around — which carries a defect-valence, though the entry's T1 notes the same rewrite is adaptive comfort for the agent, so the valence is seat-relative rather than intrinsic. Second, and sharply, vocab_travels fails at the substrate boundary: the effect requires autobiographical memory plus identity-protective motivation, so it vanishes in any system without a remembered, self-relevant past — a market price, a thermostat, an organism with no self-model does not rewrite its history, and invoking "choice-supportive bias" for them overreaches.

The portable skeleton is cognitive dissonance reduction — the memory-side expression of an agent's pressure to see its commitments as coherent, under the broader motivated-reasoning umbrella. That skeleton is what choice-supportive bias instantiates from its parent (cognitive_dissonance / motivated_reasoning), not something the named bias makes travel — but with the same distinctively framed caveat as the certainty effect: even the parent is a regularity of agents with a self-model, so it generalizes to dissonance-bearing agents, not across substrates at large. Its character: a real, mechanistically specific feature of human autobiographical memory, structural in that it names a genuine cognitive regularity, but framed as a memory-contaminating bias and confined so tightly to the self-modeling-agent substrate that its portable skeleton reaches only other commitment-bearing agents, never a system without a remembered past.

Structural Core vs. Domain Accent¶

This section decides why choice-supportive bias is a domain-specific abstraction and not a prime, and it carries the case for its domain-specificity — there is no separate section for that. Like the certainty effect, it is confined so tightly to its substrate that even its portable skeleton reaches only other agents of the same kind.

What is skeletal (could lift toward a cross-domain prime). Strip the memory-of-choice specifics and a thin relational structure survives: an agent under pressure to see its commitments as coherent resolves the discomfort by revising a self-relevant representation until the commitment looks justified — the dissonant record is rewritten rather than the commitment abandoned. The pieces that travel are abstract — a commitment that creates dissonance with countervailing evidence, an identity-protective pressure, and a resolution that adjusts the representation to fit the commitment. That skeleton is genuinely portable across dissonance-bearing agents, which is why it is housed in the catalog as cognitive_dissonance under the motivated_reasoning umbrella — the parent the entry instantiates. But it is the core it shares, not what makes choice-supportive bias distinctive — and, unusually, even this parent is a regularity of agents with a self-model, not a substrate-neutral form.

What is domain-bound. Everything that makes it choice-supportive bias in particular is judgment-and-decision-making and autobiographical-memory furniture and none of it survives extraction intact: the operation on the memory substrate (the remembered attribute profile as it stood at the time of choice) rather than on current belief; the directional tilt (chosen option remembered better, rejected ones worse than the contemporaneous evidence); the inside-undetectability (the person sincerely feels the superiority "was there all along," and accuracy instructions do not remove the tilt); the dissonance moderators (decision difficulty, involvement, irreversibility) that size the swing; and the record-at-decision-time remedy (decision logs, pre-committed criteria, pre-mortems). These are the worked distortion, its moderators, and the applied cases (consumer loyalty, voter approval, treatment-choice review, career retrospection, decision journals) that JDM actually studies. The decisive test: the rewriting requires an agent with autobiographical memory and identity-protective motivation, so strip the self-modeling-agent substrate and the bias does not become a looser thing reaching new domains — a market price, a thermostat, or an organism without a self-model does not retroactively revise its history, and invoking "choice-supportive bias" for them overreaches. Remove the remembered, self-relevant past and there is nothing for the memory-rewrite to act on.

Why this does not clear the prime bar. A prime is a relational structure whose vocabulary travels and whose cross-domain transfer is recognition of the same mechanism, not analogy. The bias's transfer is bimodal, and both bands stay inside self-modeling agents. Within JDM and memory research it travels as mechanism — consumer behavior, voter behavior, medical decision-making, career and life retrospection — but these are application contexts of one substrate (a human agent revising its remembered record under dissonance), not structurally distinct systems, so the memory-rewrite diagnosis, the moderators, and the record-based remedy carry without translation; that is recognition within one substrate, not cross-domain reach. Beyond agents with a remembered self it does not travel as mechanism at all — a non-recalling system cannot exhibit it. And when the transferable lesson is wanted, it goes only as far as the parent the entry instantiates — cognitive_dissonance / motivated_reasoning — which itself generalizes only to other dissonance-bearing agents, not across substrates. So the cross-domain reach is doubly bounded: it belongs to a parent that is itself self-model-bound, and "choice-supportive bias," as named, carries its memory-substrate operation, its post-choice attribute-recall distortion, its dissonance moderators, and its record-at-decision-time remedy as baggage that stays bound to human retrospection. (It must also be kept distinct from its family — confirmation bias, hindsight bias, rosy retrospection, the endowment effect — each of which sends a differently-tuned remedy.)

Relationships to Other Abstractions¶

Current abstraction Choice-Supportive Bias Domain-specific

Parents (4) — more general patterns this builds on

Choice-Supportive Bias is a kind of Bias Prime

Choice-Supportive Bias is a systematic directional error in reconstructing the evidence for chosen and rejected options.
Choice-Supportive Bias presupposes Decision Prime

Choice-Supportive Bias requires a prior decision that fixes which option is chosen and which alternatives are rejected before memory can tilt between them.
Choice-Supportive Bias is a decomposition of Cognitive Dissonance Prime

Choice-Supportive Bias is cognitive-dissonance reduction applied to the remembered attribute profile of a completed choice.
Choice-Supportive Bias is a decomposition of Motivated Reasoning Prime

Removing the child’s frame leaves the reusable structure named by Motivated Reasoning.

Hierarchy paths (11) — routes to 10 parentless roots

Choice-Supportive Bias → Bias

Not to Be Confused With¶

Confirmation bias. The forward-looking tendency to seek, notice, and overweight new evidence that fits an existing belief. Choice-supportive bias is backward-looking and operates not on new evidence but on the remembered attributes of the already-compared options, rewriting the record of how the evidence stood at the time of choice. Tell: is the distortion in how fresh, incoming evidence is gathered and weighted (confirmation bias) or in the recalled profile of options a past decision already ranged over (choice-supportive bias)?
Hindsight bias. The distortion of the remembered predicted likelihood of an outcome — "I knew it would turn out this way" — inflating how foreseeable a result now seems. Choice-supportive bias distorts the remembered attributes of the compared options, not the remembered probability of an outcome; it can operate even where the outcome is still unknown. Tell: is the memory tilting how predictable the result feels (hindsight) or how good the chosen option looked relative to the rejected ones (choice-supportive)?
Rosy retrospection. The general tendency to remember past experiences more positively than they were rated at the time, across events whether or not a choice was involved. Choice-supportive bias is the choice-contingent slice of this broader family: it specifically inflates the chosen option and denigrates the rejected ones, driven by dissonance over a commitment. Part-vs-whole: choice-supportive bias is one directional, decision-linked variant of rosy retrospection's blanket positivity. Tell: is the past being remembered warmly in general (rosy retrospection), or is the warmth aimed specifically at what one picked and coldness at what one passed up (choice-supportive)?
Endowment effect. The tendency to value something more highly simply because one currently owns it, inflating present valuations conditioned on possession. Choice-supportive bias moves remembered attributes of a past decision, not current ownership-conditioned valuations, and needs no ongoing possession — a rejected job offer or a foregone school gets its memory rewritten though nothing is owned. Tell: is the inflated value attached to a thing one holds now (endowment) or to the recalled merits of a choice one made in the past (choice-supportive)?
Post-decisional dissonance / spreading of alternatives. The classic dissonance effect in which, after choosing, a person revises their current evaluation of the options — rating the chosen one up and the rejected one down going forward. This is a belief effect on present assessment; choice-supportive bias is the memory effect, rewriting the remembered attribute profile as it stood at decision time so the person feels the superiority "was there all along." Tell: has the person's current preference shifted (spreading of alternatives), or has their memory of what the evidence originally showed been altered while they deny having updated (choice-supportive)? The two often co-occur but target different representations.
Cognitive dissonance / motivated reasoning (the parent it instances). The general pressure on an agent to see its commitments as coherent, resolved by revising some self-relevant representation. Choice-supportive bias is the memory-side expression of this parent in human autobiographical memory, adding the specific attribute-recall distortion, dissonance moderators, and record-at-decision-time remedy. Tell: strip away the autobiographical memory and identity-protective motivation and there is nothing for the memory-rewrite to act on — a market price or thermostat cannot exhibit the bias, only, if anything, the dissonance parent (which itself reaches only other self-modeling agents). Carry the parent, not this named bias, into any non-recalling system. (Treated fully in a later section.)

Neighborhood in Abstraction Space¶

Choice-Supportive Bias sits in a crowded region of the domain-specific corpus (37^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Unclustered & Miscellaneous (309 abstractions)

Nearest neighbors

Status Quo Trap — 0.85
Proof by Assertion — 0.85
Epiphany — 0.85
Barnum Effect — 0.84
Information Avoidance — 0.84

Computed from structural-signature embeddings · 2026-07-12