Behavioral Decision-Making and Cognitive Biases

Behavioral Decision-Making & Cognitive Biases

The descriptive account of judgment and decision-making: how people actually decide, not how they should. Use it to read why a customer, buyer, or stakeholder made a seemingly irrational choice, to ethically shape the decision environment, and to catch bias in your own forecasts and recommendations.

Installed as a Claude Code skill with four on-demand reference files (not duplicated here): references/biases-catalog.md (full heuristics-and-biases catalog with canonical experiments + TAM application notes), references/choice-architecture.md (defaults, EAST, MINDSPACE, sludge vs nudge vs boost, choice overload, ethics of influence), references/debiasing-and-application.md (debiasing procedures + worked operator scenarios), and references/replication-status.md (what survived vs what is contested/failed, with citable sources — read before citing any effect externally).

Descriptive vs normative: keep them separate. This skill is descriptive (the psychology of real decisions). For the normative side, computing the optimal action under constraints (linear programming, decision trees, EVPI, expected-utility maximization), use da-33-prescriptive-analytics. The gap between the two is the subject matter here: people deviate from the normative optimum in systematic, predictable ways.

The honesty rule (read first). Decision/social psychology went through a replication crisis. Several once-famous effects did not survive (power posing, social priming, and ego depletion are contested/failed). The core judgment-and-decision-making findings (anchoring, framing, the disposition effect, present bias, default effects) replicate well; several adjacent social-psych effects do not. Never present a debunked effect as established fact. See references/replication-status.md before citing any effect to a customer or in a written recommendation.

When to reach for this skill

• A customer/buyer made a choice that looks irrational -> name the bias, then address the real driver.
• You’re setting a price, an opening offer, or a contract renewal -> anchoring and framing.
• A renewal/expansion stalls on “we already invested in X” -> sunk-cost; on “let’s keep things as they are” -> status-quo/default bias.
• You’re designing a signup, plan-selection, or opt-in/opt-out flow -> choice architecture (and the sludge you should remove).
• You’re writing a forecast, capacity plan, or project timeline -> debias your own judgment (overconfidence, planning fallacy).
• Someone cites “power posing” / “priming” / “ego depletion” as fact -> check replication status first.

If the task is changing a customer’s behavior over time (adoption, habit, enablement), that’s behavior-change-psychology, not this skill.

Core concepts (the minimum working set)

1. Dual-process theory: System 1 / System 2

Two modes of cognition. System 1 is fast, automatic, associative, affect-laden, and effortless; it produces most snap judgments and most biases. System 2 is slow, deliberate, effortful, and lazy (it endorses System 1 unless prompted). Labels coined by Stanovich & West, popularized by Kahneman (Thinking, Fast and Slow, 2011). Biases are System 1 outputs that System 2 fails to catch.

• Caveat: treat “two systems” as a useful metaphor, not literal brain architecture. The strict two-box model is contested (better read as a continuum of automaticity). Don’t oversell it.
• Operator use: high-stakes decisions (renewals, escalations, architecture calls) deserve a deliberate System-2 step (a checklist or premortem) precisely because the default is a System-1 gut call.

2. Heuristics & biases (Tversky & Kahneman, 1974, Science)

Mental shortcuts that are “highly economical and usually effective” but produce “systematic and predictable errors.” The working set every operator should recognize: anchoring-and-adjustment (the first number dominates; the single most useful effect in negotiation), availability (judging probability by ease of recall; the loud outage feels likelier than the silent risk), representativeness (stereotype/similarity over base rates; the conjunction fallacy), confirmation bias (the engine behind most bad root-cause calls), hindsight bias (“knew it all along”; corrupts postmortems), overconfidence (90%-confident estimates are right far less than 90% of the time; the planning fallacy), status-quo / default bias, and sunk-cost fallacy (honoring unrecoverable past spend instead of deciding on the margin). Detail and operator scripts: references/biases-catalog.md.

3. Prospect theory (Kahneman & Tversky, 1979, Econometrica)

How people choose under risk, the descriptive replacement for expected-utility theory: reference dependence (outcomes judged as gains/losses from a reference point; whoever sets the reference frames the decision), loss aversion (losses loom larger than equivalent gains; classic estimate ~2x, but CONTESTED, do not state “2x” as universal law), diminishing sensitivity (concave for gains, convex for losses), probability weighting (small probabilities overweighted, hence lottery tickets AND insurance; the certainty effect), the fourfold pattern (risk-averse for likely gains and unlikely losses; risk-seeking for unlikely gains and likely losses), and framing effects (“90% uptime” vs “10% downtime” flip the choice).

Loss aversion is contested, not debunked. Gal & Rucker (2018), “The Loss of Loss Aversion,” argue it is far more context-dependent than the “universal 2x law” implies; gains can loom larger at small magnitudes (Harinck et al., 2007) and predicted pain of loss overstates the actual (Kermer et al., 2006). Real in many settings (especially higher-stakes, endowed goods) but not a context-free constant. Use it as a hypothesis to test for this customer, not a guaranteed lever.

4. Bounded rationality & satisficing (Simon) vs ecological rationality (Gigerenzer)

• Bounded rationality (Herbert Simon) — limited information/time/compute, so people satisfice: pick the first option clearing an aspiration threshold rather than optimizing. Buyers rarely run an exhaustive comparison; they stop at “good enough.”
• Ecological rationality / fast-and-frugal heuristics (Gerd Gigerenzer) — the counterpoint to heuristics-and-biases. Simple heuristics (take-the-best, recognition, 1/N) are adaptive and often more accurate than complex models under scarce/uncertain information (less-is-more). A heuristic is “rational” relative to its environment.
• Why both matter: one says shortcuts cause errors, the other says shortcuts are often the smart move. The truth is conditional: match the diagnosis to the environment before “fixing” a heuristic. These two programs are the respective foundations of nudging and boosting.

5. Mental accounting & present bias

• Mental accounting (Thaler) — money sorted into non-fungible mental “buckets” (budget categories, “house money,” renewal-vs-new-purchase), violating fungibility. A spend framed against the “innovation budget” lands differently than against “BAU/maintenance.”
• Present bias / hyperbolic discounting — near-term costs/rewards discounted far more steeply than distant ones (quasi-hyperbolic beta-delta; Laibson 1997). Produces time-inconsistency: “we’ll migrate next quarter” gets reversed when next quarter arrives. Upfront-cost / delayed-benefit work (migrations, upgrades, tech-debt paydown) is chronically under-chosen.

6. Choice architecture & nudges (Thaler & Sunstein, 2008) and boosts (Hertwig)

• Choice architecture — every presentation of options (order, defaults, count, framing) influences choice; there is no neutral presentation, so design it deliberately.
• Nudge — alters behavior predictably without forbidding options or changing incentives (libertarian paternalism). The most powerful nudge is the default (opt-out organ donation, 401(k) auto-enrollment, pre-checked tiers).
• EAST (UK Behavioural Insights Team) — make the desired action Easy, Attractive, Social, Timely. The most practical operator checklist.
• MINDSPACE — Messenger, Incentives, Norms, Defaults, Salience, Priming, Affect, Commitments, Ego. (The “Priming” element rests on social-priming research that largely failed to replicate; treat it as the weakest element.)
• Sludge — friction added against the person’s own interest (cancellation mazes, hidden opt-outs). Find and remove sludge in your own onboarding/renewal flows; don’t deploy it.
• Boosts (Hertwig & Grüne-Yanoff, 2017) — the contrast to nudges. Instead of steering the chooser, build their competence (teach a decision rule, give a fast-and-frugal tree, present risks as natural frequencies). Boosts preserve agency and persist after the intervention; prefer them for long-term, trust-based relationships, which is most TAM work.
• Ethics: nudge toward the chooser’s own interest, keep it transparent, never sludge. Full applied detail (incl. choice overload): references/choice-architecture.md.

7. Debiasing

You cannot will a bias away, but structured procedures help: consider-the-opposite (best-evidenced general debiaser, strong against anchoring/overconfidence), premortem (Gary Klein) (imagine the project has failed and explain why; prospective hindsight, cheap and high-yield), reference-class forecasting (Flyvbjerg / Kahneman’s “outside view”) (estimate from the distribution of comparable past cases; the fix for the planning fallacy), and checklists (force System 2 through a disciplined pass; only work with consistent adherence). Procedures and worked scenarios: references/debiasing-and-application.md.

Operator quick-map (bias -> tell -> move)

This table is a hypothesis generator, not a verdict. Treat the observed “tell” (what the customer said or did) as data, not a confirmed diagnosis. A tell suggests a candidate effect; confirm it against this specific person/context before acting (several effects are context-dependent). If the tell is ambiguous, gather one more observation or ask a clarifying question first. Never state the bias label to the customer or imply they are irrational; the label is your internal hypothesis, the “move” is what you do.

Anti-patterns

• Calling a customer “irrational.” They’re predictably boundedly rational. Name the mechanism and design around it.
• Citing a debunked effect. Power posing, social priming, ego depletion are contested/failed; the “2x loss-aversion constant” is over-stated. Check references/replication-status.md first.
• Weaponizing nudges (sludge / dark patterns). Steering a customer against their own interest is self-defeating in a TAM relationship. Prefer boosts.
• Treating System 1/2 as literal neuroanatomy. It’s a model. Don’t overclaim.
• One-shot debiasing. Awareness alone barely moves biases; only structured procedures reliably help, and only with disciplined use.
• Over-applying loss aversion / “fixing” a heuristic that’s actually ecologically rational. Diagnose the environment first (Gigerenzer’s caution).

Cross-references

• behavior-change-psychology — adjacent and complementary. This skill = the descriptive psychology of a decision (biases, framing, choice architecture). That skill = changing behavior over time (motivation, Fogg B=MAP, stages-of-change, habit loops, adoption). “Design an onboarding nudge to drive adoption” -> that skill; “what default/framing shapes this purchase decision” -> this skill.
• da-33-prescriptive-analytics — the normative counterpart (optimal action under constraints: LP/MILP, decision trees, EVPI, utility theory). Compute the optimum there; understand why humans deviate from it here.
• executive-comms — persuasion and decision-driving communication (board memos, negotiation prep, influence). For the rhetoric/persuasion craft go there; for the underlying decision psychology stay here.
• postmortem-writing — applies hindsight-bias control in incident reviews.
• deep-research-methods — covers confirmation bias / echo chambers as research anti-patterns.

Sources

• Tversky, A. & Kahneman, D. (1974). “Judgment under Uncertainty: Heuristics and Biases.” Science 185(4157), 1124-1131.
• Kahneman, D. & Tversky, A. (1979). “Prospect Theory: An Analysis of Decision under Risk.” Econometrica 47(2), 263-291.
• Tversky, A. & Kahneman, D. (1991). “Loss Aversion in Riskless Choice: A Reference-Dependent Model.” QJE 106(4).
• Kahneman, D. (2011). Thinking, Fast and Slow.
• Simon, H. A. (1955/1956). Bounded rationality and satisficing.
• Gigerenzer, G. & ABC Research Group. Fast-and-frugal heuristics / ecological rationality.
• Thaler, R. & Sunstein, C. (2008/2021). Nudge (and Nudge: The Final Edition).
• Laibson, D. (1997). “Golden Eggs and Hyperbolic Discounting.” QJE.
• Hertwig, R. & Grüne-Yanoff, T. (2017). “Nudging and Boosting.” Perspectives on Psychological Science 12(6), 973-986.
• Dolan, P. et al. (2010). MINDSPACE; Behavioural Insights Team (2014). EAST.
• Gal, D. & Rucker, D. (2018). “The Loss of Loss Aversion.” Journal of Consumer Psychology.
• Replication: Open Science Collaboration (2015) Science; Many Labs 2; Ranehill et al. (2015) and Simmons & Simonsohn (2017) on power posing; Hagger et al. (2016) on ego depletion.