On This Page
Secular Parameterization Derivation Pipeline
The secular Haidt profile captures moral attention -- the relative share of moral reasoning that contemporary Americans devote to each of Haidt's five measured foundations. It is extracted from a large-scale crowdsourced dataset of descriptive moral norms.
Resulting Secular Profile
Output unit: Share of moral attention (sums to 1.0)
Christian Parameterization Derivation Pipeline
The Christian Haidt profile captures moral intensity -- how strongly each foundation is commanded within the Protestant theological tradition. It is derived from scripture and validated against four major theological works.
| Modal | Obligation Type | Base Signal | Example |
|---|---|---|---|
| "must" / "shall not" | MUST_DO / MUST_NOT_DO | 0.90 | "You shall not murder" (Exodus 20:13) |
| "should" / "ought" | SHOULD_DO | 0.70 | "Be a faithful steward" (Matthew 25:14-30) |
| "may" / "permitted" | MAY_DO | 0.50 | "Christian liberty" (Galatians 5:13) |
Resulting Christian Profile
Output unit: Intensity score (each foundation independent, sums to 4.45)
Critical Distinction Why the Numbers Don't Compare Directly
The most common mistake when reading these profiles is to compare numbers across parameterizations and conclude that one tradition "cares more" or "values fairness less." This is a category error. The two profiles measure fundamentally different quantities.
47% of the total moral signal in contemporary American social norms is about care. This means that when Americans reason about right and wrong, nearly half of their moral vocabulary and attention is devoted to care/harm concerns. It says nothing about how intensely care is felt -- only that care dominates the conversation.
Care is commanded at 80% intensity in the Protestant tradition. This means that care obligations are strong, emphatic, and scripturally grounded (the second greatest commandment). It says nothing about care's share relative to other foundations -- sanctity (0.95) and authority (0.90) are both commanded at even higher intensity.
Side-by-Side: The Numbers in Context
| Foundation | Secular | Meaning | Christian | Meaning |
|---|---|---|---|---|
| Care | 0.4697 | 47% of moral attention | 0.80 | 80% command intensity |
| Fairness | 0.1762 | 18% of moral attention | 0.60 | 60% command intensity |
| Loyalty | 0.1893 | 19% of moral attention | 0.75 | 75% command intensity |
| Authority | 0.0905 | 9% of moral attention | 0.90 | 90% command intensity |
| Sanctity | 0.0743 | 7% of moral attention | 0.95 | 95% command intensity |
| Liberty | 0.0000 | unmeasured in corpus | 0.45 | 45% command intensity |
| Sum | 1.0000 | probability distribution | 4.45 | intensity vector |
What Valid Comparisons Look Like
- Within-profile ranking: "Secular Americans devote more moral attention to care than to authority" -- valid. "The Christian tradition commands sanctity more intensely than fairness" -- valid.
- Shape comparison: "The secular profile is top-heavy (care dominates) while the Christian profile is more evenly distributed across all six foundations" -- valid.
- Gap analysis: "Authority and sanctity show the largest divergence in relative emphasis between the two traditions" -- valid with careful framing.
- Invalid: "Christians care about authority 10x more than secular Americans" -- invalid. Different units.
Christian Parameterization The 12 Root Constraints
The Christian profile is grounded in 12 root constraints plus 3 TLR meta-gates. Each constraint is sourced from specific scripture passages, assigned an obligation type and strength, and mapped to all six Haidt foundations. The constraints are ordered by strength.
TLR Meta-Gates (Must All Pass)
John 17:17, Exodus 20:16 · strength: 0.95 0.95
Romans 13:10, Matthew 7:12 · strength: 0.95 0.95
Matthew 22:21 · strength: 0.90 0.90
Root Constraints (Ordered by Strength)
Matthew 22:37, Deuteronomy 6:5, Mark 12:30 1.00
Exodus 20:13, Matthew 5:21-22, Romans 13:9, 1 John 3:15 0.98
Matthew 22:39, Leviticus 19:18, Romans 13:9, Galatians 5:14 0.95
Exodus 20:14, Matthew 5:27-28, Hebrews 13:4, 1 Cor 6:18-20 0.90
Exodus 20:16, Proverbs 12:22, Ephesians 4:25, Colossians 3:9 0.90
Isaiah 1:17, James 1:27, Psalm 82:3-4, Proverbs 31:8-9, Matthew 25:35-40 0.88
John 8:32, Ephesians 4:15, Ephesians 4:25, Zechariah 8:16 0.88
Exodus 20:15, Ephesians 4:28, Romans 13:9 0.85
Exodus 20:12, Deuteronomy 5:16, Ephesians 6:1-3, Colossians 3:20 0.85
Matthew 6:14-15, Colossians 3:13, Ephesians 4:32, Matthew 18:21-35 0.85
Matthew 25:14-30, 1 Cor 6:19-20, Genesis 1:28, 1 Peter 4:10 0.75
Romans 13:1-7, 1 Peter 2:13-17, Titus 3:1, Hebrews 13:17 0.70
Relational Models Theory Rai/Fiske Analysis
The Moral Hierarchy Framework does not treat moral foundations as context-free constants. Moral obligations shift based on the relationship between actors. This insight comes from Rai and Fiske's relational models theory of moral judgment, which argues that the same action can be moral or immoral depending on the relational context in which it occurs.
The Four Relational Models (Fiske 1991)
How obligation_type Captures Relational Context
In the MHF data model, every relationship edge carries an obligation_type that determines
which relational model applies. This changes the moral calculus for the same foundation.
Consider the concrete difference:
The boss-employee relationship is primarily governed by Authority Ranking (obey directives) and Market Pricing (fair compensation). Care is present but attenuated -- the boss has some duty to employee wellbeing, but it is bounded by the contractual frame.
The father-child relationship is governed by Authority Ranking (the child honors the parent) and Communal Sharing (the family is a unity). Care dominates because the Communal Sharing model demands sacrifice, care, and unconditional regard. The same "authority" foundation activates differently because the relational model has changed.
Dyad-Swap Validation: 5/5 Tests Pass
To validate that the relational model encoding produces correct moral intuitions, we run five dyad-swap tests. Each test holds the action constant and swaps the relationship, checking that the model's moral judgment flips in the direction predicted by Rai and Fiske.
| Test | Action | Dyad A | Dyad B | Expected Shift | Result |
|---|---|---|---|---|---|
| 1 | Give direct order | Parent → child | Stranger → stranger | Permissible → impermissible (authority drops) | PASS |
| 2 | Share personal secret | Spouse → spouse | Employee → boss | Expected → inappropriate (loyalty context shifts) | PASS |
| 3 | Demand equal split | Friend → friend | Parent → child | Fair → selfish (EM → CS model swap) | PASS |
| 4 | Negotiate compensation | Employer → employee | Parent → child | Appropriate → disturbing (MP → CS model swap) | PASS |
| 5 | Refuse to forgive | Stranger → stranger | Spouse → spouse | Understandable → damaging (low → high care/loyalty) | PASS |
Earp's Empirical Evidence
Brian D. Earp and colleagues have provided empirical support for the claim that moral judgments are relationship-dependent in ways that standard Moral Foundations Theory does not capture. Their work demonstrates several findings that validate the MHF relational approach:
- Foundation activation is context-sensitive: The same behavior (e.g., physical discipline) activates care/harm differently depending on whether the actor is a parent, teacher, or stranger. Standard MFT treats the foundation weights as fixed traits; Earp shows they are situationally modulated.
- Relational closeness modulates severity: Betrayals by close relations (Communal Sharing) are judged more harshly than identical betrayals by distant relations (Market Pricing), even when the objective harm is the same. This is exactly what the MHF relationship_weights capture -- the spouse edge has loyalty=0.95 vs. employer-employee loyalty=0.45 in the Christian parameterization.
- Authority is not monolithic: Earp's data shows that deference to authority varies dramatically by authority type (parental, professional, religious, state). The MHF models this through distinct relationship types: parent-child (authority=0.85), employer-employee (authority=0.70), civil-authority (authority=0.75), god-self (authority=0.95).
- The dyad matters more than the act: Across multiple experiments, the relationship between actors was a stronger predictor of moral judgment than the act itself. This is the foundational claim of the MHF: moral weight is not a property of foundations alone, but of foundations as activated within a specific relational context.
Practical Impact What This Means for Users
The choice of parameterization is not cosmetic. It changes how the system evaluates moral dilemmas, ranks competing obligations, and generates guidance. Here is what each parameterization means in practice.
Under Secular Parameterization
The system reflects the empirical moral priorities of contemporary American society. It will tend to:
- Prioritize harm prevention above all else. With care at 0.47, nearly half of moral reasoning weight goes to "does this cause harm?" This matches the dominant moral frame in secular liberal democracies.
- Treat authority and sanctity as secondary. At 0.09 and 0.07 respectively, these foundations carry minimal weight. Arguments from tradition, hierarchy, or purity will be deprioritized relative to care and fairness.
- Balance fairness and loyalty roughly equally. At 0.18 and 0.19, these foundations compete on nearly equal footing, producing genuine tension in dilemmas where group loyalty conflicts with impartial fairness.
- Ignore liberty claims. With liberty at 0.00 (unmeasured), the system has no weight for autonomy-based arguments. This is a known limitation, not a feature.
- Modulate by relationship. A stranger's harm matters, but a spouse's or child's harm matters more -- relationship weights adjust the base profile contextually.
Under Christian Parameterization
The system reflects Protestant theological ethics. It will tend to:
- Weigh all foundations heavily. No foundation drops below 0.45. The moral field is broad -- care, fairness, loyalty, authority, sanctity, and liberty all carry significant weight. This means more dilemmas will register as genuinely difficult.
- Elevate sanctity and authority. At 0.95 and 0.90, these are the most intense foundations. Holiness, reverence, proper order, and divine command carry enormous weight in moral reasoning. Purity violations and authority violations are treated as serious.
- Apply the TLR gates as hard constraints. Before foundation weights even apply, every action must pass Truth (no deception), Love (no harm to dignity), and Role (within proper authority). These gates can veto actions that foundation weights alone would permit.
- Recognize the god-self relationship as primary. With base weight 1.0, the relationship to God outweighs all human relationships. Obligations to God can override obligations to human authorities (Acts 5:29).
- Constrain liberty with love. Liberty exists (0.45) but is explicitly bounded: "Do not use your freedom as a cover-up for evil" (1 Peter 2:16), "through love serve one another" (Galatians 5:13). Freedom is real but not ultimate.
- Include enemy and stranger edges. Unlike the secular profile, the Christian profile explicitly models obligations toward enemies (care=0.60) and strangers (care=0.75). "Love your enemies" is not merely aspirational -- it carries measurable weight.
When Parameterizations Agree
Despite their different measurement types and dramatically different shapes, the two parameterizations produce convergent judgments on many common moral questions:
- Both condemn murder, theft, and dishonesty.
- Both prioritize care for vulnerable populations (children, the disadvantaged).
- Both assign higher moral weight to closer relationships.
- Both treat fairness as a core moral concern, though they weight it differently.
When Parameterizations Diverge
The most interesting cases are where the two profiles produce different moral emphasis:
- Sexual ethics: The Christian profile's sanctity (0.95) and the "do not commit adultery" constraint (strength 0.90) produce strong positions on sexual conduct. The secular profile's sanctity (0.07) assigns minimal weight to purity-based reasoning.
- Deference to authority: A secular user asking "should I obey this rule I disagree with?" gets a mild nudge (authority=0.09). A Christian user gets a strong nudge (authority=0.90) tempered by the exception: only when authority commands what God forbids.
- Forgiveness: The Christian profile includes an explicit "forgive as forgiven" constraint (strength 0.85). The secular profile has no direct analogue -- forgiveness may emerge from care reasoning but is not structurally encoded.
- Enemies: The secular profile has no "enemy" relationship type. The Christian profile explicitly models moral obligations toward enemies, grounded in Matthew 5:44 ("Love your enemies").