MEVIR 2: Trust Decisions Model

• MEVIR 2 is a formal, virtue-informed model that explains trust decisions by combining evidence processing, epistemic virtue theory, and moral analytics.
• It introduces a Trust Lattice that differentiates Truth Bearers from Truth Makers to systematically map claims to their justificatory evidence.
• The framework underpins decision support systems with algorithmic moral filtering and virtue checks to mitigate cognitive biases and enhance reflective trust judgments.

The MEVIR 2 Framework—"Virtue-Informed Moral-Epistemic Model of Trust Decisions"—is a formal, descriptive model explaining how human agents and collectives form trust judgments in complex, polarized, and contested informational environments. Extending the original MEVIR architecture, MEVIR 2 systematically integrates procedural evidence processing, epistemic virtue theory, and moral foundation analytics, providing a unified structure for modeling both individual and group-level trust decision processes, the emergence of "Truth Tribes," and the impact of metacognitive interventions. The framework offers both theoretical constructs—notably the Trust Lattice, Truth Bearer/Truth Maker distinction, and multi-factor moral profile—and concrete algorithms and tools for evidence elaboration, moral-content quantification, and cognitive bias mitigation (Schwabe, 2 Dec 2025, Schwabe, 20 Dec 2025).

1. Architecture and Components

MEVIR 2 posits that trust decisions emerge via continuous interaction among three analytical layers:

1. Procedural Evidence-Gathering and Reasoning: Agents recursively construct chains of evidence, terminating in "trust anchors"—accepted authorities, beliefs, resource-exhaustion, or pre-trusted statements. This produces a directed acyclic structure known as the Trust Lattice, rather than a scalar degree of belief.
2. Epistemic Virtue Theory: Following Zagzebski, belief formation occurs via two dynamically chosen paths: Path 1 ("Self-Reliance"), governed by conscientiousness and intellectual courage, versus Path 2 ("Deference") invoking humility, open-mindedness, and attentiveness in selecting and weighting authorities. Each agent maintains a virtue-weight and vice-weight vector $$(w_h, w_o, w_c, w_p, ..., v_{ar}, v_{cnf}, ...)\in\mathbb{R}^k$$.
3. Moral Foundations and Cooperative Domains: The framework unifies Extended Moral Foundations Theory (EMFT) (Care, Fairness, Liberty, Loyalty, Authority, Purity) with Morality-as-Cooperation (MAC) domains (kin selection, reciprocity, group, property, heroism, etc.), permitting moral-profile weighting $M: D\cup F\to [0,1]$ such that $\sum M = 1$. These weights function as priors, guiding the selectivity and salience of both evidence and authorities.

Interactions are bi-directional: fast, affect-laden moral intuitions influence procedural and authority-path selection, while cultivated virtues regulate the oscillation between intuitive and deliberative cognition.

2. Ontological Semantics: Truth Bearers, Truth Makers, and Unpacking

Central to MEVIR 2 is the ontological distinction between:

• Truth Bearers (TB): Symbolic entities to which truth values may be assigned (propositions, statements, data records).

$$\mathrm{TB} = \{\tau \mid \tau\ \text{is a declarative proposition, sentence, or datum}\}$$

• Truth Makers (TM): Ontic entities, events, or relations in the world that render a Truth Bearer true.

$$\mathrm{TM} = \{\mu \mid \mu\ \text{is an entity/event/relationship}\}$$

The correspondence relation $\vDash \subseteq \mathrm{TB} \times \mathrm{TM}$ encodes that $\tau$ is true in virtue of $\mu$. Agents perform "ontological unpacking," mapping high-level $\tau$ to minimal sets of $\mu$ that ground its truth:

$$U: \mathrm{TB} \to \mathcal{P}(\mathrm{TM}),\quad U(\tau) = \{\mu_1,\dots,\mu_n\}\ \text{such that}\ \{\mu_i\vDash\tau\}$$

Disagreements frequently arise from incompatible unpackings: different agents admit different ontological grounds as candidate truth-makers, based on both their moral priors and procedural habits.

3. Trust Lattice and Trust Tribes

Trust Lattice Construction

Each agent’s trust structure is formalized as a labeled meet-semi-lattice $\mathcal{L} = (C, \leq)$, where $C \subseteq \mathrm{TB} \cup \mathrm{TM}$ and $c_1\leq c_2$ if $c_1$ evidentially supports $c_2$. Each node $c$ carries a trust weight $w(c)\in [0,1]$:

$$w(c) = \sigma\left(\alpha v(c) + \beta m(c) + \gamma e(c)\right),\ \alpha + \beta + \gamma = 1$$

where $v(c)$ is the virtue alignment, $m(c)$ moral-profile resonance, $e(c)$ evidential strength, and $\sigma$ a logistic normalization.

Trust Tribes

A core innovation is the formalization of "Truth Tribes" (TTs): maximal sets of agents whose procedural algorithms, virtue vectors, and moral profiles are mutually within $\epsilon$ of each other. Each TT constructs a unique, internally consistent Trust Lattice $L_T = (C_T, \prec_T)$, with shared anchors $A_T$, virtue-weights $V_T$, and moral profiles $M_T$.

The framework proves that mutually unintelligible epistemic worlds typically form when anchor sets are disjoint or when moral-profile divergence exceeds a threshold $\delta$, i.e., no partial isomorphism exists between Trust Lattices that preserves anchors and salience labels. This formalizes the empirical observation of polarization and incommensurability between, for example, "Anti-Vax" and "Pro-Vax" or "Climate Skeptic" and "Climate Advocate" TTs (Schwabe, 20 Dec 2025).

4. Algorithmic Process and Decision Workflow

The MEVIR 2 algorithmic flow for evidence evaluation comprises:

1. Initial Moral Filtering: Using NLP-based content analysis, the agent’s moral profile is instantiated.
2. Recursive Elaboration: Builds the agent-specific Trust Lattice via evidence/proxy search, terminating when a trust anchor is encountered.
3. Virtue Check: At each node, virtue alignment is computed. Sub-threshold values trigger Path 2 (deference); otherwise, Path 1 continues.
4. Authority Selection: For deference, candidate authorities are selected respecting the agent’s moral profile and further vetted for virtue alignment.
5. Trust Weight Calculation: Trust values propagate up the lattice by fixed-point iteration until global confidence in the top-level claim converges.
6. Decision: Acceptance or rejection of a claim is determined by thresholding the final trust weight.

This process operationalizes the metacognitive cycle and points of failure (e.g., overreliance on self-assessment or over-narrow deference).

5. Case Studies and Empirical Illustration

Vaccination Mandate Dispute

• Profile A: Emphasizes Liberty/Purity, uses anchors such as personal anecdotes, alternative medical blogs, privileges risk aversion, and is triggered by disgust and sovereignty concerns.
• Profile B: Emphasizes Care/Fairness/Reciprocity, relies on scientific institutions and aggregate data, is triggered by empathy and duty to others.

Each group’s Trust Lattice relies on incommensurable anchors, systematically excluding the other’s evidence.

Climate Policy Debate

• Profile C: Emphasis on Liberty, Loyalty, Property; Trust Anchors in economic modeling, national interest, industry advocacy.
• Profile D: Emphasis on Care, Fairness, Authority, Kinship; Trust Anchors in scientific consensus, intergenerational responsibility.

These configurations produce distinct, mutually reinforcing trust lattices, explaining persistent policy deadlock and echo chambers.

6. Decision Support System and Metacognitive Tools

MEVIR 2 enables the design of Decision Support Systems (DSS) that visualize the Trust Lattice, surface the user's moral and virtue-profile activations, and recommend interventions:

• Moral frame detection tags each claim with activated MAC/EMFT domains.
• Real-time lattice construction and anchor mapping highlight doctrinal gaps and opportunities for adversarial deference.
• Virtue enhancement modules nudge excessive self-reliance toward expert consultation, and curtail confirmation-driven, myopic deference.
• Utility functions balance belief confidence against cognitive resource costs:

$$U(\tau) = E_{\mathrm{True}}[v(\tau)] - \lambda C(\mathcal{L})$$

The DSS can prune sublattices with negative marginal utility and promote exposure to counter-anchored lattices, fostering reflective reframing and epistemic humility (Schwabe, 2 Dec 2025).

7. Limitations and Future Extensions

MEVIR 2’s limitations arise from its subcomponents:

• EMFT/MAC foundations’ dimensionality and causal status remain empirically debated.
• The situationist critique challenges the stability of epistemic virtues.
• Static versus dynamic modeling of moral profiles may mischaracterize real-time shifts.

Proposed research directions include longitudinal tracking of moral and virtue profiles, empirical measurement of trust lattice evolution, and further integration of value theories (e.g., Schwartz) and computational argumentation logics (i.e., refining the Trust Lattice into a residuated lattice for entailment and attack relations).

Planned architectural extensions include adaptive learning of virtue–moral weights, DSS collaboration across Truth Tribes, and empirical validation of intervention efficacy in reducing epistemic polarization (Schwabe, 2 Dec 2025, Schwabe, 20 Dec 2025).