Extended Moral Foundations Theory

• EMFT is a formalized model of human moral intuitions that integrates evolutionary, psychological, and linguistic evidence to explain variations in trust and polarization.
• The theory maps moral foundations to evolutionary cooperation games, systematically linking moral rhetoric with decision-making processes.
• EMFT underpins decision-support systems and interventions by visualizing individual moral weight vectors to address epistemic polarization.

Extended Moral Foundations Theory (EMFT) is an empirically informed, formalized model of human moral intuitions and rhetorical patterns that systematizes and extends the original Moral Foundations Theory (MFT). Developed as a core pillar in the MEVIR and MEVIR 2 frameworks, EMFT provides a multidimensional account of how pre-rational, evolutionarily shaped “moral taste receptors” inform individual differences in trust judgments, epistemic authority selection, and susceptibility to polarization. It integrates evolutionary, psychological, and linguistic evidence and specifies both descriptive and normative functions, enabling applications in computational modeling, decision-support, and analysis of online misinformation (Schwabe, 2 Dec 2025, Schwabe, 20 Dec 2025, Mutlu et al., 2020).

1. Theoretical Foundations

EMFT builds directly upon Haidt’s Moral Foundations Theory, which posits that human moral judgments arise from a limited set of affect-laden, domain-specific systems—such as Care/Harm and Fairness/Cheating—that are fast, pre-rational, and evolutionarily conserved. EMFT extends MFT in two key respects:

• Foundation Expansion and Refinement: EMFT formalizes six core foundations, splitting Fairness into two sub-foundations—Fairness-as-Equity (emphasizing equality of outcome) and Fairness-as-Proportionality (reward-by-merit), and adding Liberty/Oppression to the list endorsed in recent MFT revisions.
• Cooperation Game Mapping: Each foundation is systematically mapped to an evolutionary “cooperation game” (the MAC model), operationalizing why distinct intuitions cluster together and how they become salient in social reasoning (Schwabe, 20 Dec 2025).

Each agent possesses a vector of moral “weights” $$w = (w_{\text{care}}, w_{\text{f-equity}}, w_{\text{f-prop}}, w_{\text{liberty}}, w_{\text{loyalty}}, w_{\text{authority}}, w_{\text{purity}})$$ that governs the salience of each foundational intuition in filtering experience and evidence (Schwabe, 2 Dec 2025, Schwabe, 20 Dec 2025).

2. Formal Ontology and Computational Models

EMFT operationalizes moral judgment via several formal and semi-formal constructs:

• Detection Function: When encountering a claim $c$, $\operatorname{detectFoundations}(c) \subset \{$Care, Equity, Proportionality, Liberty, Loyalty, Authority, Purity$\}$ identifies which foundations are activated.
• Moral Resonance Functions: Each foundation $k$ scores a claim $c$ per $M_k(c) = w_k \cdot s_k(c)$, where $s_k(c)$ denotes the degree to which $c$ invokes the relevant social pattern (e.g., harm, fairness, contamination) (Schwabe, 20 Dec 2025).
• Trust Decision Integration: The propensity to accept a claim is modeled as $$P_{\text{trust}}(c) \propto \operatorname{Policy}(c) \cdot \exp{\left[\sum_k w_k \phi_k(c)\right]}$$, integrating procedural, virtue, and moral resonance filters.

The ontology distinguishes:

• Truth Bearers: Symbolic representations (e.g., propositions, data entries).
• Truth Makers: Real-world elements that, if instantiated, render a proposition true (e.g., statistical data vs. phenomenological sovereignty). EMFT determines which Truth Makers an agent deems acceptable, acting as an epistemic filter (Schwabe, 2 Dec 2025).

3. Extended Moral Foundations: Taxonomy and Characteristics

The seven EMFT dimensions and their operational motivations are as follows:

Foundation	Motivation	Sample Moral Heuristic
Care / Harm	Kin-selection, protecting vulnerable	Value empathy, avoid harm
Fairness-as-Equity	Conflict-avoidance via equal share	Demand equal outcomes
Fairness-as-Proportionality	Prevent free-riding, match merit	Reward in proportion to contribution
Liberty / Oppression	Resist illegitimate domination	Defend autonomy, oppose coercion
Loyalty / Betrayal	Foster group coordination	Prioritize ingroup interests
Authority / Subversion	Maintain stable hierarchies	Defer to legitimate authority
Purity / Sanctity	Pathogen/contaminant avoidance	Avoid contamination/disgust

The dual-pole structure—virtue (positive) and vice (negative)—is operationalized in lexicons such as the Extended Moral Foundations Dictionary (EMFD), which expands on the original Moral Foundations Dictionary by doubling the coverage with modern, synonym-rich terminologies validated against human annotation (Mutlu et al., 2020).

4. Integration in the MEVIR and MEVIR 2 Frameworks

Within the MEVIR family of models, EMFT constitutes one of three pillars alongside procedural reasoning and virtue epistemology:

• Procedural Elaboration: EMFT supplies the initial “trust anchors,” determining which authorities, evidence types, or beliefs are admitted and prioritized in the recursive trust lattice. The procedural model elaborates subsequent claims and evidence chains subject to these moral priors (Schwabe, 2 Dec 2025).
• Virtue-Governed Override: The virtue component mediates when rapid (System 1) moral intuitions should be reconsidered in light of intellectual virtues (e.g., humility, diligence). EMFT configures the default thresholds at which agents rely on intuition versus effortful reasoning, as described in the dual-process framework (Schwabe, 2 Dec 2025, Schwabe, 20 Dec 2025).

These processes explain the emergence of “Truth Tribes”—social clusters with internally coherent trust lattices but mutually unintelligible evidentiary standards, leading to epistemic polarization (Schwabe, 20 Dec 2025).

5. Empirical Applications and Measurement

EMFT underlies both computational and qualitative studies of moral rhetoric and polarization:

• Textual Operationalization: The EMFD enables identification of the moral dimension and valence (virtue/vice) of texts at the granular level. For instance, each tweet can be represented as a 10-dimensional vector spanning all virtue and vice poles, normalized per token and per tweet length (Mutlu et al., 2020).
• Moral Loading and Polarization: Aggregating these vectors over time or by topic, researchers can quantify the dominance of specific foundations and the degree of daily or event-based polarization. For example, Net Harm/Care loading or co-fluctuation entropy between foundations can be tracked longitudinally (Mutlu et al., 2020).

Table: Inventory sizes used in EMFD (Mutlu et al., 2020):

Dimension	Virtue Stems	Vice Stems
Care/Harm	95	85
Fairness/Reciprocity	69	57
In-group/Loyalty	99	72
Authority/Respect	160	101
Purity/Sanctity	97	161

The majority of tweets in large corpora activate at least one moral pole, and virtue-based rhetoric appears more frequently than vice, though Harm/Care is generally underused yet more polarized (Mutlu et al., 2020).

6. Case Studies: Vaccine Mandates and Climate Policy

EMFT enables systematic analysis of highly polarized domains:

• Vaccine Mandate Debate: “Sovereignty-Purity” Truth Tribes (high Liberty/Purity weights) regard mandates as bodily assault, privileging anecdotal and purity-based Truth Makers; “Community Health” Truth Tribes (high Care, Fairness-equity, Authority) privilege institutional evidence of public benefit. Divergent trust lattices become impervious to counter-foundational evidence (Schwabe, 2 Dec 2025, Schwabe, 20 Dec 2025).
• Climate Policy: “Economic-Liberty” Truth Tribes (high Liberty, Proportionality, Loyalty) oppose regulation on grounds of economic harm and unfair burden distribution, while “Global-Responsibility” Truth Tribes (high Care, Equity, Authority) emphasize justice and collective welfare, selecting scientific authorities and discounting economic loss projections. These choices are structurally predictable by EMFT weight assignments (Schwabe, 20 Dec 2025).

7. Normative Implications, Intervention, and Limitations

EMFT functions descriptively (explaining the roots of trust and polarization) and normatively (suggesting strategies for improving deliberation and communication):

• Decision-Support Systems: EMFT-guided tools can infer activated moral frames, visualize user-specific w-vectors, prompt users when high-arousal intuitions may block inquiry, and recommend counter-tribal authorities for consideration. Such systems aim to make the moral architecture of reasoning explicit for metacognitive reflection—without prescribing beliefs (Schwabe, 20 Dec 2025).
• Moral Reframing: Effective communication requires targeting the moral foundations salient to one’s audience. “Inter-tribal translation tables” and “Inter-Tribal Phronesis” training are recommended for practitioners and moderators seeking depolarization.
• Limitations: EMFT is not a utility function for moral value, and intra-foundation disagreements (e.g., what counts as “fairness”) persist. The model’s effectiveness in longitudinal change remains to be empirically tested (Schwabe, 2 Dec 2025).

EMFT thus provides a multidimensional, computationally tractable vocabulary for understanding why individuals and communities adopt divergent epistemic stances—and a toolkit for diagnosing, visualizing, and potentially bridging deep-seated moral divides.