--- title: "Moral Psychology" description: "The descriptive science of how people judge right, wrong, and fair — and why" --- # Moral Psychology The descriptive science of how people judge right, wrong, and **fair** — and why those judgments so often refuse to move when you hand someone a better argument. This skill explains **what is happening in the moral machinery** so you can diagnose a values clash, predict whether "this isn't fair" is really about the outcome or the process, and notice when a team is talking itself into conduct it would normally reject. It is **descriptive, not prescriptive.** It tells you how moral cognition works; it never tells you what is actually right (see Theme 7). And it is the *moral* lens specifically — for the neighboring lenses, see **When to use something else**. ## When this skill fires Reach for it when the question is about **rightness, values, or fairness**, not tactics: - "Two stakeholders agree on every fact and *still* won't converge — why?" - "The customer keeps saying it's *unfair*, but the number is objectively market-rate." - "I've made the rational case three times and the champion won't budge." - "They're treating a victimless change as a moral betrayal. Where's the victim?" - "This decision is mathematically right but feels viscerally wrong to everyone." - "Is our team rationalizing something we shouldn't be doing?" ## The map: four questions decide which lens 1. **Which value is in play?** → Moral Foundations / Morality-as-Cooperation (T1) 2. **Why won't reasoning move them?** → Social Intuitionist Model (T2); dual-process (T3) 3. **Who is the harmed party in their head?** → Theory of Dyadic Morality (T4) 4. **Is this fair — and are we rationalizing?** → Organizational Justice (T6); moral disengagement (T5) Theme 7 (is–ought) bounds what you may *ethically do* with any of the above. --- ## 1. Moral Foundations Theory — *which* value governs the call Moral judgments run on a few intuitive foundations: **Care/harm, Fairness/cheating, Loyalty/betrayal, Authority/subversion, Sanctity(purity)/ degradation**, plus a proposed **Liberty/oppression**. They cluster into **individualizing** (care + fairness — protect the person) vs **binding** (loyalty + authority + sanctity — protect the group). Best-replicated result: a **political asymmetry** — progressives lean on care + fairness; conservatives weight all roughly equally; libertarians weight liberty highest. Measured by the MFQ → MFQ-2. **Contested — say so.** Factor analysis frequently recovers only **two** clean factors, not five, and the rival **Morality-as-Cooperation** (Curry 2019) derives **seven** cooperative domains from game theory and claims better grounding. Treat MFT as a **working vocabulary**, not settled fact. > **Operator application.** A values-based stakeholder conflict is usually a fight > over **which foundation applies** (a "fairness" framing vs a "we honor our > commitments / loyalty" framing), not over the facts. **Name the foundation in > play** instead of adding more evidence — evidence aimed at the wrong foundation > slides off. ## 2. Social Intuitionist Model — intuition first, reasoning is the press secretary Haidt 2001: moral judgment is **fast intuition first**; conscious reasoning is mostly **post-hoc justification** — the "press secretary, not the judge." But the model is *social*: reasoned persuasion and others' stances **can** reshape a person's intuition (links 3/5/6), so reasoning has **limited but real** force — mostly through people and over time, not by winning the argument in the moment. Evidence: **moral dumbfounding** ("it's wrong, I just can't say why" — the Julie & Mark vignette). **Caveat:** the original dumbfounding study was an unpublished, **n≈30** demonstration; later work supports the phenomenon but debates its rate. > **Operator application.** When **purely rational arguments bounce off**, it's > because the intuition came first and reason is defending it. **Reframe the > intuition** (engage a different gut read) or use **social/relational channels** > (a trusted peer moves the intuition more than your logic does) rather than > arguing harder. ## 3. Dual-Process Moral Judgment — the emotional vs cognitive split Greene: an **automatic/emotional** route pushes toward **deontological** judgments ("it's just wrong to push him"), a **controlled/cognitive** route toward **utilitarian** ones ("5 > 1"). **Trolley/switch** (impersonal) → most permit; **footbridge** (personal, hands-on) → most refuse, despite identical math. Cognitive load **selectively slows utilitarian** judgments; VMPFC-lesion patients are **more utilitarian**. **Contested — attach the caveats.** Kahane 2015: sacrificial "utilitarian" answers track **reduced harm-aversion (Instrumental Harm)**, *not* genuine **Impartial Beneficence** (Oxford Utilitarianism Scale) — so "more utilitarian on trolley" is not "more moral." The **CNI model** (Gawronski 2017) shows the dichotomy conflates Consequences-sensitivity, Norms-sensitivity, and a general **Inaction** bias. Trolley vignettes also draw heavy **ecological-validity** criticism. > **Operator application.** When a decision is mathematically sound but feels > viscerally wrong (a cold-formula layoff, a technically-fair clawback), expect > **deontological recoil** that no spreadsheet beats. Reduce the > *personalness/directness* of the harm, or wrap it in fair process (Theme 6) — > don't just re-present the arithmetic. ## 4. Theory of Dyadic Morality — find the victim in their head Schein & Gray 2018: moral cognition is organized by a template — an **intentional Agent harming a vulnerable Patient (A→P)**. Two phenomena: **moral typecasting** (agents and patients seen as opposites — villains feel less pain, victims can do no wrong) and **dyadic completion** (for a "victimless" wrong, people **invent a victim** — society, the relationship, future users, their own integrity). **Contested vs MFT** on whether *all* morality reduces to harm — unresolved. > **Operator application.** When a stakeholder frames a **victimless** change as a > moral breach, they've already cast **someone as harmed** (their team, their > users, the relationship) and **you as the intentional agent**. Surface **who > they perceive as harmed** before debating intent — "no one was hurt" just gets > the template re-completed. ## 5. Moral Licensing & Moral Disengagement — how people excuse bad acts **Moral licensing** (Monin & Miller 2001): a past good act licenses a later questionable one. **It is small and fragile — state this plainly:** meta-analytic **d≈0.31** with strong publication-bias signals → bias-corrected to **d≈0.18 or near null**, and a large registered replication found ≈zero. **Do not build a plan around inducing it.** **Moral disengagement** (Bandura 1999) is the actionable one — **eight mechanisms** for switching off self-sanction, used as a **detection checklist**: ① moral justification ② euphemistic labeling ③ advantageous comparison ④ displacement of responsibility ⑤ diffusion of responsibility ⑥ distortion/disregard of consequences ⑦ dehumanization ⑧ attribution of blame. > **Operator application.** When a team slides toward conduct it would normally > reject (misleading a customer, burying a known defect), listen for these eight > verbal tells. **Naming the mechanism** out loud ("that's euphemistic labeling — > say what we're actually doing") re-engages the self-sanction the language was > suppressing. ## 6. Organizational Justice — fairness is process + treatment, not just outcome The highest-impact theme. Colquitt 2001's **four dimensions**: **distributive** (the outcome — equity theory; equity/equality/need norms), **procedural** (the process — Leventhal's six rules: **consistency, bias-suppression, accuracy, correctability, representativeness, ethicality** + **voice**), **interpersonal** (dignity and respect), and **informational** (honest, adequate explanations). **The fair-process effect** is the single most useful finding: a **fair process raises acceptance of even an unfavorable outcome.** People accept a price increase, a denied feature, or a lost escalation they dislike **if** the process was consistent, gave them **voice**, treated them with **dignity**, and was **honestly explained** — and resent a *good* outcome delivered arbitrarily. > **Operator application.** "This isn't fair" is rarely only about the dollar > amount (distributive). It's usually **process** (no warning, no voice, > inconsistent with last time) and **treatment** (curt, no real explanation). You > often can't change the outcome — but you can almost always improve **process and > explanation**: advance notice, voice, consistent rule, candid reasons, an appeal > path. That is the lever that turns "unfair" into "I don't love it, but I get it" > — in pricing changes, escalations, negotiations, and missed-SLA conversations. ## 7. Is–Ought / Naturalistic Fallacy — the limit on ethical nudging Everything above is **descriptive**. **Hume**: you cannot derive an *ought* from an *is*. **Moore** (*Principia Ethica*, open-question argument): "good" is not definable by any natural property. Describing how moral cognition works **never prescribes** what is right — disgust driving purity judgments doesn't make them correct, and fairness perception being hackable via process doesn't make it ethical to hack. > **Operator application.** This bounds the **ethics of nudging**: use moral > psychology to **understand and communicate honestly**, not to manufacture > consent or steer people past their own judgment. When someone argues "it's > natural, therefore it's fine," flag the is→ought gap. --- ## When to use something else | You actually need… | Use | | --- | --- | | Persuasion / attitude-change mechanism — ELM/HSM, cognitive dissonance, reactance, attribution, social-influence & norm theory | `persuasion-and-influence-psychology` | | Cognitive biases, prospect theory, choice architecture, nudges *as decision theory*, System 1/2, debiasing | `behavioral-decision-making` | | Interpersonal **trust**, rapport, psychological safety between people | `trust-and-psychological-safety` | | **AI** ethics, model fairness, algorithmic bias, privacy, responsible-AI governance | `da-11-ethics-and-privacy` | | Emotion **mechanics** — disgust/affect generation, emotion regulation, EI | `emotion-and-affect-psychology` | | Driving product **adoption** via motivation/habit models (SDT, Fogg, stages-of-change) | `behavior-change-psychology` | | Clinical / forensic / therapeutic moral-development assessment | a licensed professional | This skill owns the **moral judgment / fairness-perception** layer: which value is in play, why reason won't move a moral stance, who is cast as harmed, whether a process is *perceived* as fair, and whether a team is rationalizing. ## Deeper detail & full citations Full theory depth (the six SIM links, the CNI parameters, all eight disengagement mechanisms, Leventhal's rules in full, MFQ→MFQ-2, the MAC seven domains) and the complete source list (Graham 2013; Haidt 2001; Greene 2001/2008; Koenigs 2007; Kahane 2015; Gawronski 2017; Schein & Gray 2018; Monin & Miller 2001; Blanken 2015; Bandura 1999; Colquitt 2001; Hume; Moore; moralfoundations.org) live in `references/moral-psychology-detail.md` — read it before a deep or contested answer. ## A reusable diagnostic prompt Paste this when a values/fairness clash is stuck: ``` Act as a moral-psychology analyst. For the situation below, work through: 1. FOUNDATION — which moral foundation(s) is each party invoking (care, fairness, loyalty, authority, sanctity, liberty)? Is the conflict a foundation mismatch rather than a factual disagreement? 2. INTUITION vs REASON — is a fast moral intuition driving this, with reasoning as post-hoc justification? Will more logic move it, or do I need to reframe the intuition or route through a trusted peer? 3. DYADIC TEMPLATE — who is cast as the harmed Patient and who as the intentional Agent? If it looks "victimless," who have they completed the dyad with? 4. FAIRNESS — separate distributive (outcome) from procedural (consistency, voice, accuracy, correctability), interpersonal (dignity), and informational (honest explanation). Which is actually driving "this isn't fair"? What fair-process move raises acceptance without changing the outcome? 5. DISENGAGEMENT CHECK — are we (my side) using any of Bandura's 8 mechanisms (moral justification, euphemistic labeling, advantageous comparison, displacement/diffusion of responsibility, distortion of consequences, dehumanization, attribution of blame) to rationalize our position? 6. ETHICAL LIMIT — is any proposed move an is→ought overreach — exploiting the machinery to override informed consent rather than to communicate honestly? Flag replication caveats (MFT factor structure, dumbfounding origin, licensing fragility, Kahane/CNI critique of "utilitarian" answers) where they apply. Situation: ```