A Working "Theory" of Rationality

In this post, I want to describe how I think of rationality; what it means to rationally accept a proposition, what it means for evidence to support proposition, and when we should revise beliefs when confronted with incoming information. I'm mainly doing this as a means of juxtaposing the "ideal inquirer" against the "conspiracy theorist", for that blog series. There are a few resources I'll be consulting for this. Collectively, when weaved together, I think they offer a powerful understanding of clear thinking.

Fundamental Contextual Considerations

Before considering the validity of an argument, we need to consider the context in which that conclusion is generated. Rationality is inextricably linked to a context of discourse; we normally advance arguments within a dialogue with participants with varying degrees of commitments. The pragma-dialectical approach, developed by Frans van Eemeren and Rob Grootendorst, sees argumentation primarily as a critical discussion aimed at resolving differences of opinion in a rational way. Its foundation is in both pragmatics (language use in context) and dialectics (reasoned dialogue). Argumentation is understood as a regulated dialogue between parties who disagree. The goal is not just persuasion, but achieving rational resolution of disagreement. They propose an ideal model of critical discussion, structured in four stages: confrontation stage, opening stage, argumentation stage, and concluding stage.

In the confrontation stage, a difference of opinion emerges. Disagreement is made explicit, which sets the stage for rational engagement. In the opening stage, parties establish the common ground; shared premises, procedural rules, and commitments. They agree who has the burden of proof. This sets the stage for the "playing field" for fair debate. In the argumentation stage, the protagonist advances reasons to defend their standpoint. The antagonist raises critical doubts, asks for clarifications, or challenges their reasoning. Ideally, this proceeds cooperatively: each move aims at testing whether the standpoint can be rationally defended. Lastly, in the concluding stage, if the standpoint has been successfully defended, both parties should rationally agree it is justified. The outcome is ideally a closure of the disagreement, not endless quarrel.

Central to this is a set of ten rules for critical discussion. Breaking one of these rules constitutes a fallacy:

Freedom rule - No one should be prevented from advancing or challenging standpoints. (No silencing, intimidation, or straw-man distortions.)
Burden of proof rule - A party who advances a standpoint must defend it if challenged. (No shifting the burden onto the opponent.)
Standpoint rule - Attacks must address the standpoint actually put forward, not a distorted version.
Relevance rule - Defenses must respond to the criticisms actually raised.
Unexpressed premise rule - If an argument relies on implicit premises, these must be made explicit when challenged.
Starting point rule - Parties must not deny commitments already agreed upon in the opening stage.
Validity rule - Arguments must employ reasoning that is logically valid (deductively or otherwise appropriate).
Argument scheme rule - Use of specific schemes (e.g., analogy, authority, causal reasoning) must be legitimate and open to critical questioning.
Closure rule - If a standpoint is successfully defended, it should be accepted; if refuted, it should be withdrawn.
Usage rule - Participants must avoid unclear, ambiguous, or manipulative language.

On this view, fallacies are not simply bad arguments in an abstract logical sense, but strategic moves that hinder the resolution of disagreement by violating one or more of these rules. For example, a straw man violates the standpoint rule. Shifting the burden of proof violates the burden of proof rule. Thus, fallacies are derailments of critical discussion; they undermine rational resolution. A crucial insight from pragma-dialectics is that rationality is not reducible to formal validity. Logic alone is too narrow: a deductively valid inference can still be a fallacy if it unfairly evades commitments or silences an opponent. Dialectical rationality: rationality is tied to the rules of fair engagement between participants. Valid reasoning is necessary, but only as one element of a larger process. Pragmatic dimension: argumentation is language use in interaction. What counts as rational depends on how speech acts are performed, whether they fulfill roles in the dialogue, and whether they respect shared commitments. Collaborative ideal: rationality is about engaging in a cooperative, rule-governed process whose telos is resolving differences of opinion. So, unlike traditional logical positivist views (where rationality = application of valid inference rules), pragma-dialectics embeds rationality in a normative, pragmatic, dialectical framework.

Extending this concept, Douglas Walton introduces the idea of "Dialogue Types". Walton distinguishes different contexts of dialogue, each with its own goals and norms. This extension allows for a wider variety of dialogical purposes, giving us further clarity into whether a particular move in a dialogue is fallacious. This relates to his broader work on argumentation schemes, which I've belabored upon elsewhere, in that rationality consists of applying the correct schemes within the correct dialogical context. Walton’s typology of dialogues extends the pragma-dialectic insight by showing that rationality is context-sensitive: what counts as a reasonable argumentative move depends not only on the ten rules of critical discussion, but also on the purpose of the dialogue itself. Therefore, like the pragma-dialectic tradition, rationality cannot be decoupled from the context of discussion. Together, these approaches yield a rich picture of rationality. Rational argument is not reducible to formal validity; rather, it is: collaborative (parties must respect rules that allow disagreement to be tested fairly), pragmatic (arguments are moves in a structured conversation with speech-act conditions), and contextual (what counts as a good or bad move depends on the kind of dialogue being conducted). Pragma-dialectics gives us the baseline norms of rational discussion, while Walton’s dialogue typology shows how these norms are contextually specified depending on the type of argumentative activity. Rationality is thereby reconceived as a dynamic, dialogical practice, where validity of inference is only one component within a larger rule-governed, context-sensitive process.

Walton proposes that dialogues are structured interactions with a defining purpose (goal), initial situation, participant roles, and normative rules. Rationality in each type means acting in accordance with these contextual norms. Below are the most common types of dialogue he has identified:

Persuasion Dialogue (critical discussion): The goal is to resolve a difference of opinion through reasoned argument. One party doubts or denies a claim, the other advances it. Some of the norms or rules governing this type include: each side stating their commitments clearly, burden of proof is shouldered by the one advancing a standpoint, attacks must be actual standpoints (not straw men), and the use of arguments can be tested against critical questions.
Inquiry Dialogue: The goal here is to establish the truth of a claim. It is inherently knowledge seeking. Participants share ignorance but want to find out "what is the case". Evidence must be gathered and weighted impartially, hypotheses must be tested rigorously, unsupported claims are not acceptable, and arguments must be relevant to the inquiry question. Some fallacies include cherry-picking data, suppressing evidence, or appealing to irrelevant considerations.
Deliberation Dialogue: The goal is to decide the most prudent course of action. Participants face a practical problem and must choose among alternatives. Some of the rules governing this are: consequences of each option must be considered, trade-offs and values at stake must be acknowledged, preferences must be justified by appeal to shared goals or principles, and irrelevant personal attacks or rhetorical diversions are irrational. Some fallacies include oversimplifying options, ignoring consequences, or false dilemmas.
Negotiation Dialogue: the goal is to reach mutually acceptable settlement of conflicting interests. Participants have opposed preferences but must reach a compromise. Each side must state their positions and concessions openly. Offers and counteroffers must be genuine and feasible. Agreements must not rest on deception or coercion. compromises should balance interests fairly. Some fallacies include withholding crucial information or bad-faith bargaining.
Information Seeking Dialogue: Transfer knowledge from one party (who knows) to another (who doesn’t). One participants lacks needed information, the other possesses it. The questioner must ask relevant, clear questions. The answerer must give accurate, complete, and honest answers. Ambiguity or evasion counts as irrational. some fallacies include misleading answers, refusal to provide relevant information, or some other sort of evasive behavior.
Pedagogical Dialogue: The goal is to facilitate learning and understanding, often in an educational or mentoring context. One party (teacher) has expertise; the other (learner) seeks structured guidance. The teacher should provide explanations suited to the learners level, questions by the learner should be encouraged/answered constructively, misunderstandings should be clarified patiently, not exploited, and progress should be cumulative by building knowledge step by step. Some fallacies include discouraging questions or dogmatic assertion without explanation.
Eristic Dialogue (Quarrel): The goal is to vent emotions, score points, or “win” in confrontation. This is often a conflict of attitudes and is personal or hostile. While not oriented to rational resolution, eristic dialogue still has norms: fairness in turn-taking, relevance of attacks to the conflict, avoidance of outright deception. Emotional expression is expected, but not at the cost of sheer incoherence.

The Limitations of Argument

Graham Oppy’s “What Derivations Cannot Do” is a provocative critique of a dominant argumentative strategy in philosophy of religion. Derivations concluding “P" or “not P”, unless they amount to reductios, are philosophically worthless in disputes. Oppy contends that much of contemporary philosophy of religion misplaces its focus on formal derivations that conclude in hotly contested claims, like God’s existence. These derivations, he argues, contribute nothing unless they can show an internal inconsistency in a worldview (i.e., function as reductio ad absurdum arguments).

Oppy begins with an idealized dispute between PRO vs CON, who disagree about CLAIM. If PRO offers a derivation from premises that CON already accepts, then the argument has force; it may expose contradiction or force belief revision. But if PRO uses premises that CON rejects, the derivation is useless in dispute; it merely reiterates what CON already disbelieves. He concludes that only reductio arguments, those showing a contradiction within CON’s belief system, are valuable. This is because, from a pure argumentative stance, anyone is at liberty to reject any proposition CLAIM depends on, which fundamentally undermines the force of an argument. Oppy is essentially showing us that there needs to be an alternative method for resolving disputes that depend on unshared intuitions. Instead of viewing disputes as isolated arguments, they should be framed as comparisons of best theories. To do this, you should construct a theory including CLAIM, and a theory including the negation of CLAIM. Then check if either theory is self-defeating (internal incoherence will be exposed via reductio). If they both survive, then assess which theory is superior on theoretical virtues; simplicity, explanatory power, fruitfulness etc. Derivations (arguments from within a system), only matter if they expose fatal internal inconsistencies.

This might seem disconnected from "Conspiracy Theories" at first, but I think it's a useful starting point because many conspiracies do not operate in isolation; they are part of a broader system, of which this particular conspiracy claim is a derivation from. Oppy contents that the derivations that don't function as reductios merely serve to boost confidence in your worldview, make beliefs more resilient, and provide argumentative "insurance". They maybe useful for understanding structure or relationships among beliefs—but this is epistemically insignificant for justifying the worldview itself. Oppy's critique is primarily directed at the current state of philosophical argumentation, but I think this has applications in public discourse. He contends that arguments focusing on "P" vs "not P" are entirely misleading, largely because they rely on the interlocutors intuitions about premise plausibility rather than proper theory comparison. Public debates devolve into “stockpile of arguments” competitions rather than meaningful comparisons of worldviews. For any worldview, infinite derivations of its conclusions are possible. So exhibiting such derivations is epistemically idle; they do not increase justification. More effort should go into identifying reductios or proper comparative evaluation, not refutation. I'll show examples of this later. Often, derivations from conspiracy hypotheses can be refuted, up to a point where the conspiracy theorist relies on intuitions of the plausibility of some proposition (usually a skeptical hypothesis), that more or less leads to a dead end. The proper solution is to show the performance of the entire view relative to a non-conspiratorial view.

Oppy suggests that rational belief is not about accepting conclusions from stand-alone arguments. Instead, it’s about embedding claims within entire worldviews or theoretical frameworks and assessing those worldviews for internal consistency and theoretical virtues like simplicity, explanatory power, predictive success, fit with data, and coherence. This is particularly relevant in the context of conspiracies, because the conspiracy theorists can simply fall back on a stockpile of arguments, if a single argument was debunked. From Oppy's view, rationality means choosing the best overall theory or worldview, not just believing the conclusion of an argument. Oppy accepts that reasonable people can disagree rationally. He acknowledges deep disagreement and theory-ladenness of evidence—what counts as evidence or a plausible premise depends on your background beliefs (more on this later). Rationality is contextual and perspectival; tied to what counts as the best worldview given the evaluator’s epistemic starting point. It's just that derivations are not rationally persuasive unless it uses shared premises or reveals a contradiction. So this means that rational persuasion is showing that one's worldview fares better on theoretical virtues. Oppy resists the idea of a universal “rationality filter” that can decide which beliefs are justified independently of any background commitments. Rationality is internal to worldviews and judged by comparative evaluation, not external, absolute criteria. This minimizes the role of intuitions and acknowledges that rational disagreement is often persistent and unavoidable.

The Role and Proper Use of Evidence

So far, I think it's clear that rationality isn't exhausted by argumentation. I think this is more or less obvious, but how about an evidential perspective on rationality? The standard evidential account of rationality comes from Bayesian Epistemology. This is a formal framework that models rational belief and belief updating using probability calculus. It gives us a process for updating our degrees of belief in a proposition or hypothesis after observing evidence. Mathematically, it is:

\[ P(H \mid E) = \frac{P(E \mid H) \cdot P(H)}{P(E)} \]

where:

H is a hypothesis,
E is evidence,
\( P(H \mid E) \) is the posterior probability of H given E.

It emphasizes prior probabilities, likelihoods, and conditionalization on new evidence. The idea is that belief in a hypothesis should be responsive to evidence but that it also depends on background knowledge. This framework enforces consistency in probabilistic assignments, requiring them to be consistent with Kolmogorov's axioms. This is similar to Oppy; he focuses on internal coherence, that a worldview should not defeat itself and should maintain structural consistency. It's dissimilar in that Oppy rejects the need for such quantification. He appeals to qualitative judgments about simplicity, explanatory power, etc., without reducing these to probability values.

So Bayes theorem tells us how to rationally update new beliefs when exposed to evidence. I think this is a good starting point for thinking about rationality when it comes to handling evidence, but is still not comprehensive for a theory of rationality. In Bayesianism, priors (initial degrees of belief) are crucial, and controversial. Rational disagreements can often be traced to differing priors. Differing priors often reflect different worldview commitments; and can often be unconstrainted or arbitrary or can import bias into what is presented as a neutral framework. In Bayesianism, rational disagreement is tricky. If two people update on the same evidence, they should converge—given common priors (per Bayes’ Convergence Theorem). Oppy explicitly defends the idea that rational agents can permanently disagree, because their worldview structures differ at a deep level and no shared "prior" exists to resolve the gap. One might simply assume this could be resolved by reference to predictive success of the model, but often two models can describe observations equally well.

This is relevant to the topic of conspiracy because often conspiracy theorists will point to data in support of their hypothesis, and updated their beliefs in accordance with valid probabilistic updating. So as we can see from Oppy's analysis and the Bayesian framework, formal validity is insufficient; a conspiracy theory can be formally valid so this is not enough to rule one out as absurd. This brings me to my next point, something I find to be crucially underappreciated when it comes to discussions about rationality. Any theory of rationality should probably include something about responsiveness to evidence. Bayesian seem to do this, but it begs the question as to what counts as evidence in the first place. This seems crucial, that evidence is theory laden, and needs argumentation to justify it as evidence before we can assign a probability to the evidence. We need to demonstrate its relevance. For example, claiming to a climate change denier that there’s a consensus would likely not count as evidence in favor of climate change but rather its negation since they might have a higher order skeptical principle that interprets that as evidence that there’s corruption. What counts as evidence is not theory-neutral, and any account of rationality must grapple with the problem of evidence-responsiveness under deep disagreement.

Bayesianism presumes you already know what counts as evidence (E) and how it probabilistically relates to a hypothesis (P(E|H)). But what counts as evidence, and what it means, is theory-laden—informed by prior commitments, background beliefs, and sometimes, ideology. This is the "evidence relevance problem": how do we know that E is evidence for H, and not for ~H? A datum like “scientific consensus” is not neutral—it gets interpreted within a worldview. To a climate change denier, the consensus might be evidence of corruption or elite control, not of climate change itself. So what counts as evidence is up for debate, and that debate happens at the level of meta-theory or worldview justification. Evidence only has evidential force within a theoretical framework that renders it intelligible and relevant. So a more practical model would require first order and second order rationality. First order would be Bayesian style credence updating within a worldview. The second order rationality would be a comparative assessment because on how they interpret data, handle anomalies, accommodate disagreement, or explain their own epistemic foundations. The second order layer is where argumentation plays a crucial role. You can't say “Here’s my evidence” until you've argued for why it counts as evidence. You can’t assign P(E|H) without making a case for why E is even relevant to H—and that case depends on your background theory. Evidence-responsiveness is essential to rationality—but this must include the capacity to argue for what counts as evidence and why. Rationality involves not just adjusting beliefs in light of evidence, but also being willing to critically examine what “evidence” even means—and being able to engage in argumentation about that. That’s why Bayesianism alone is insufficient, and why Oppy’s worldview-comparison model, while informal, addresses something more fundamental: the interpretive rationality that must precede belief revision.

The Bayesian approach also doesn’t seem to consider the fact that evidence is ambiguous, and often requires significant interpretation. Evidence is rarely raw or straightforward. Bayesian models assume clean, discrete pieces of evidence with clear probabilistic relationships to the hypothesis. But in practice, evidence is often very ambiguous, incomplete, and highly interpretive. This leads me to another aspect of rationality, the handling, structuring, and collecting of evidence and how this relates to the robustness of a hypothesis. Henry Wigmore (a legal scholar, early 20th century) developed a method for analyzing evidential reasoning in legal trials. His “Wigmore charts” are non-quantitative, more like a map of inferences than a statistical model, that lay out how the evidence supports intermediate conclusions which in turn support ultimate claims (like guilt or innocence). In other words, his charts are structural and relational, showing how propositions relate within a broader network of inference, rather than individual arguments. They show how evidence directly supports a link in a broader structure of argument, allowing the legal analyst to identify weak points in these links by evaluating the facts or evidence on which they depend.

How do we recognize something as evidence?
How do we relate evidence to hypotheses or claims?
How do we assess incomplete or ambiguous evidence?

Wigmore's framework (and later expansions by Twining, Anderson, Schum, etc.) highlights that:

Evidence Requires Interpretation: Raw facts must be translated into claims, and their relevance must be argued. This is prior to any assignment of probability. Interpretation is epistemically upstream of belief updating.
Evidential Support Is Often Indirect and Multi-layered: A fingerprint doesn’t prove guilt. It supports presence at a scene, which supports opportunity, which supports capacity to commit the crime. This is guided by defeasible generalizations; the causal links between evidence and a proposition. Fingerprints are only evidence because they connect to the hypothesis through this chain of reasoning.
The Collection Process Matters: Rational belief depends not just on what evidence you have, but how you came to have it; Was it filtered? Selected with bias? Acquired from trustworthy sources? Bayesianism treats evidence as given. Wigmorean analysis forces us to consider procedural epistemology—how evidence enters our reasoning space.

Therefore a thicker account of evidential rationality would include:

Component	Description
Evidential relevance	How do you argue that a datum is relevant to a claim? (Needs argumentation, not just statistics.)
Evidential structure	How do bits of evidence combine? Do they corroborate, conflict, or imply one another?
Inferential chains	How do you trace indirect inferences from data to hypotheses?
Procedural access	How was the evidence gathered? Is it reliable, complete, selective?
Interpretive frames	What theory or background beliefs mediate your understanding of the evidence?

A full theory of rationality cannot start with "Here is the evidence". It must account for how evidence is identified, interpreted, structured, and integrated. And it must recognize that this process is argumentative, contextual, and often messy. So far, a theory of rationality should include internal consistency, theory comparison, proper evidential updating, and discussion about how something is considered evidence in the first place. However, I do not think this is sufficient.

Requirements of the Inquirer

I’ve always been interested in understanding the role of intellectual or epistemic virtues with theories of rationality. It seems like any theory of rationality should incorporate these to some extent, for example reflective awareness, intellectual humility, or discernment seem crucial for evaluating whether a conclusion should be accepted; if we discover the person proposing a conclusion violated these virtues in the process of coming to their conclusion. This seems under emphasized in a pure Bayesian or evidentialist frame, but something indirectly noticed by Twining. For example, proper handling and filtering of evidence often presumes a certain intellectual standard with regards to data collection. Or with regards to prior setting in a Bayesian perspective, priors can often be manipulated and this seems to be the result of violating an epistemic virtue and instead being an epistemic vice.

For a reminder of basic intellectual standards applied by the critical thinker, refer to the image below (Student Evaluation Using an Intellectual Standards Rubric for Critical Thinking).

If the process by which someone formed a conclusion involved epistemic vices (dogmatism, wishful thinking, motivated reasoning, negligence in data collection), this undermines the rational credibility of their conclusion—even if, formally, they can present “evidence” for it. This is because epistemic rationality is not just about what evidence you have but also about how you came to have it. For example, in a Bayesian context, you can set priors however you like; the formal updating rule is “rational” no matter how absurd the starting point. But if those priors were formed through epistemic vice (e.g., prejudice, overconfidence), your belief system is formally coherent but substantively irrational. Alternatively, you can build an evidential chart from poorly sourced, cherry-picked data, and the chart will look decent despite having a corrupted evidential base. Virtue epistemology—championed by thinkers like Linda Zagzebski, Ernest Sosa, and Jason Baehr—integrates epistemic character into the analysis of rationality:

Reliabilist virtue epistemology: Intellectual virtues are stable faculties (like good vision, memory, reasoning) that reliably produce true beliefs.
Responsibilist virtue epistemology: Intellectual virtues are character traits like intellectual humility, open-mindedness, fairness, courage, and attentiveness. Rational belief = belief formed by a person exercising these virtues.

In either view, how you handle evidence matters; not just what the evidence is. This interacts with Oppy, Bayesianism, and the legal analysis of evidence.

Oppy’s worldview-comparison model indirectly presupposes intellectual virtues:
- Intellectual honesty in assessing your own worldview’s weaknesses.
- Open-mindedness in considering alternative theories.
- Fairness in weighing theoretical virtues like simplicity and explanatory scope.
- Without these virtues, the comparative method degenerates into motivated defense of one’s prior worldview.
Bayesianism assumes:
- Priors are honestly set.
- Conditional probabilities reflect genuine assessment of evidence.
- In practice, priors and likelihoods can be manipulated to fit desired conclusions.
- Epistemic virtues are the only safeguard here—formal rules can’t stop a biased agent from making themselves “Bayes-coherent” but substantively irrational.
Proper evidence charting assumes:
- Intellectual carefulness in collecting data.
- Integrity in not suppressing unfavorable evidence.
- Patience in mapping indirect chains of reasoning.
- Without virtues, the chart becomes a polished presentation of cherry-picked or distorted evidence.

A virtue-integrated theory of rationality could look like this:

Evidence Responsiveness – Update beliefs in light of new information.
Worldview Coherence – Maintain internal consistency.
Evidential Structuring – Organize and evaluate inferential relationships among data (Twining/Wigmore).
Epistemic Virtue Exercise – Ensure the process of evidence gathering, interpretation, and updating is governed by virtues like:
- Intellectual humility – Awareness of your cognitive limits.
- Reflective awareness – Monitoring your own reasoning processes.
- Intellectual courage – Willingness to follow evidence against your preferences.
- Open-mindedness – Serious engagement with alternative views.
- Conscientiousness – Thoroughness in inquiry.
Vice Avoidance – Guard against dogmatism, closed-mindedness, cherry-picking, motivated reasoning.

With virtues, rationality becomes process-sensitive; it accounts for the fact that how you collect, interpret, and integrate evidence determines whether your beliefs deserve epistemic credit. In short, epistemic virtues are the quality control system for rationality. Bayesian rules, evidentialist standards, and Wigmorean structuring tell you how to handle information—but epistemic virtues determine whether you handle it well.

Environmental Considerations

So far, this framework says nothing about the environment in which we come to form beliefs. This is a question of social epistemology. For understanding something like the formation and propagation of conspiratorial style beliefs, we must acknowledge our embeddedness within a broader information ecosphere that's frequently hostile to cognitively vulnerable agents like ourselves. This brings me to Hostile Epistemology, proposed by C. Thi Nguyen. Nguyen starts by noting the current landscape of misinformation, conspiracy theories, and denialist movements. One common response in philosophy has been vice epistemology (Quassim Cassam, et al.), which explains these problems in terms of individual epistemic vices (gullibility, closed-mindedness, prejudice, etc.). Nguyen is skeptical of relying too heavily on vice epistemology: it tends to individualize the blame, focusing on personal defects or character flaws, and therefore proposes solutions that amount to fixing the agent. Hostile epistemology instead studies the ways environments exploit our unavoidable cognitive vulnerabilities. Humans are cognitively limited; we must user heuristics, take shortcuts, and reason under time pressure. We must trust others, because we cannot master all domains ourselves. These are not optional features, they are essential to finite beings like ourselves. Because they’re unavoidable, they create structural vulnerabilities. Environments can be designed (or can evolve) to exploit those vulnerabilities. So in other words, the framework I presented above, as a standard of rationality, might not be entirely feasible in practice. We must attend to these environmental factors to understand this much broader sense of rationality. These can all exploit built-in vulnerabilities in our cognition and social dependence. Nguyen broadens “environmental features” to include:

Other people (propagandists, trolls)
Social structures (echo chambers, polarization)
Cultural practices (gamification of attention, information bubbles)
Institutions (media, political organizations, tech platforms)
Technologies (algorithms, recommendation engines, metrics)

He highlights a few examples:

Clarity as seduction: simple, vivid, “clickable” clarity that overrides nuanced thought. (More on this in a bit)
Trust as vulnerability: since we must trust others (experts, institutions), that trust can be co-opted, gamed, or spoofed (fake experts, manufactured consensus).
Transparency traps: full “radical transparency” often backfires—forcing experts to oversimplify for lay audiences, or creating overwhelming “data dumps” that are easily manipulated.
Echo chambers: not mere information bubbles, but structures of distrust that actively inoculate members against outside information.

Nguyen resists framing all epistemic failures as either vices (blameworthy flaws) or defects (like blindness). Many exploitations happen not because of defects, but because we are doing the best we can with what we have. Our finite strategies (heuristics, trust, shortcuts) are reasonable adaptations, but they remain exploitable. This response suggests shifting from individual blame to structural awareness; instead of only correcting peoples character, look at the design of our epistemic environment. Cultivate intellectual playfulness as a counter to epistemic traps (openness to exploring ideas for fun prevents rigid self-reinforcement). Use error metabolism—systems that detect and correct mistakes quickly. Develop institutional and cultural safeguards against manipulation, not just individual virtues. So how does this relate to "conspiracy thinking"? Well, often people who propagate conspiracy theories are embedded in a very hostile epistemic environment. Understanding that environment can help understand why certain types of conspiracies proliferate; they are often appealing to a very specific set of cognitive vulnerabilities.

Nguyen has another paper called "The Seductions of Clarity" which can help explain the rigidity of certain beliefs. Given our cognitive limitations and epistemic environment, we often forgo nuance for simplicity. This heuristic is highly essential to function in a modern information environment. Clarity can be problematic. Nguyen argues that clarity in thought and communication, though often treated as an intellectual virtue, can also be seductive and misleading. Clarity promises simplicity, accessibility, and immediate grasp. But some subjects—especially political, social, and moral ones—are inherently messy and complex. Forcing clarity can flatten or distort that complexity. In this way, clarity can function like a drug: it feels good, it satisfies, but it can leave us epistemically worse off. Nguyen identifies two modes in which clarity operates:

Epistemic virtue mode: Clarity illuminates, disambiguates, helps us structure reasoning and avoid confusion.
Epistemic vice mode: Clarity seduces us into oversimplification, masks uncertainty, and encourages us to accept elegant but shallow accounts.

For example, conspiracy theories and ideologies often thrive because they provide a clear, coherent story that feels more satisfying than the messy, incomplete truth. This is the "seductive" aspect of clarity; it can be highly misleading. It gives us a false sense of understanding; people might feel like they "get it", even if crucial nuances are erased. Clarity also provides a sense of emotional comfort; it reduces anxiety about complexity and uncertainty, making it attractive in turbulent times. Politicians, propagandists, or bad-faith actors use clarity as a weapon—“bumper sticker” simplicity outcompetes nuanced analysis. Social media and news environments reward content that is sharp, vivid, and clear over that which is cautious or qualified (structural incentives).

“The Seductions of Clarity” fits directly into Nguyen’s broader hostile epistemology framework. Clarity becomes an epistemic vulnerability; because humans crave cognitive ease, clarity can be exploited. Hostile environments weaponize clarity; algorithms, propaganda, and institutions amplify simple, emotionally satisfying messages, even when false. We naturally trust clear presentations as indicators of truth or expertise—but in hostile contexts, this heuristic is systematically turned against us. So, clarity is not just a tool of individual reasoning—it becomes a structural feature of hostile epistemic environments, one that bad actors can exploit to hijack attention, trust, and belief. This is why I don't think the framework I've established above is sufficient for a theory of rationality without incorporating these concepts from social epistemology.

Rationality can’t be thought of as just following formal rules (like Bayesian updating) or cultivating virtues in isolation. It must be understood as situated in environments that can be hostile or supportive. So a theory of rationality needs to include virtue level traits and structural design principles (robust institutions, trustworthy systems, defenses against exploitation). It requires understanding the epistemic environment in which you operate to understand which of your cognitive vulnerabilities might be exposed and manipulated. A rational agent must develop virtues of suspicion toward seductive clarity, balancing the desire for simplicity with sensitivity to genuine complexity. Institutions and environments must be designed to inoculate against clarity traps, promoting epistemic resilience rather than rewarding oversimplification.

A unified framework for rationality

Let’s weld everything together into one working model that assumes a hostile epistemic environment, harnesses virtues, structures evidence before quantifying it, uses Bayes locally where appropriate, and finishes with worldview-level comparison in Oppy’s sense. I’ll anchor the moving parts that come from Nguyen’s two papers with inline cites.

0) Threat model: start hostile by default

Nguyen’s core warning: the feeling of clarity often works as a thought-terminating heuristic. It’s the (pleasant) signal many of us use to decide “I’ve investigated enough,” and that signal can be faked and weaponized—by conspiracy narratives and by quantified institutional metrics—so we stop too soon.

Two recurring exploit patterns:

Echo-chambers / conspiracy frames provide sweeping, “everything-fits” explanations that feel like understanding and make the world suddenly navigable; this engineered ease then locks in belief and short-circuits further inquiry.
Quantification / gamification (KPIs, scores, likes) delivers “hyper-clear” value signals (points, ranks) whose cognitive appeal outstrips their epistemic value, again triggering premature closure.

Design implication: rational methods must treat clarity as potentially adversarial, not automatically virtuous. It must include an understanding of the epistemic environment from which an agent operates. Treat unusually smooth “obvious” stories and crisp dashboards as risk signals (possible clarity-hacking), not automatic indicators of truth.

1) Virtue guardrails (process pre-conditions)

Formal rules can be gamed by bad priors/likelihoods or by clarity-hacking. So we bake epistemic virtues into the procedure:

Intellectual humility & fallibilism: mark all major claims as revisable; schedule periodic re-opens of “obvious” conclusions.
Conscientiousness & diligence: log what was searched, what was not found, and why items were excluded (anti-cherry-pick), actively seek disconfirmation.
Open-mindedness & courage: require adversarial self-tests (steel-man rivals; specify disconfirming conditions up front).
Reflective awareness: add a clarity-trigger: whenever a conclusion feels too easy, too sweeping, too delicious, raise effort, don’t lower it. Nguyen explicitly recommends developing counter-heuristics that make us suspicious of “too sweet” clarity and escalate inquiry when ideas go down “a little too smoothly.”

2) Evidence first: Wigmore/Twining structuring before numbers. Because “what counts as evidence” is theory-laden, we argue relevance before quantifying:

Build a Wigmore-style map: ultimate claim → intermediate propositions → evidences; mark support/attack links, undercutters, and credibility/collection dependencies (provenance, chain-of-custody).
Map relevance (Wigmore chart): hypothesis → intermediates → items of evidence; mark supports/attackers/undercutters and credibility dependencies. This argues something counts as evidence before you quantify it (solving the “what even is E?” problem you raised).
Provenance & handling: who collected it, with what incentives, what’s missing, and how selection/filters were applied.
Ambiguity accounting: record multiple plausible interpretations and the background assumptions each needs.

This front-loads the handling and collection stages highlighted earlier; so that Bayesian updating isn’t smuggling in the very disputes at stake.

3) Local quantification: Bayes inside vetted sub-links

Use Bayesian updates within well-specified sub-inferences (e.g., “test reliability given contamination vs. no contamination”), not as a worldview adjudicator.

Priors discipline: exhibit priors with their justifying background model; set sunset reviews for high-impact priors (they expire unless renewed), stress test your priors.
Likelihood hygiene: for every likelihood, list at least one hostile manipulation pathway (metric gaming, selection bias, interface nudges) and how it was controlled. (Nguyen’s hostile lens: environments exploit our shortcuts and trust.)

4) Trust architecture in hostile contexts

We can’t avoid dependence on others; that’s the exploit surface. Engineer trust with defensive design:

Independence triangulation: prefer convergence across institutions with genuinely different incentives (defeats single-ecosystem capture).
Pre-emption detection: flag worldviews that preemptively dismiss all outside sources as untrustworthy—classic echo-chamber structure that converts counter-evidence into confirmation.
Beware “communicative facility” illusions: bureaucratic/metric systems give credibility advantages to those who speak in their simplified terms; they appear clearer and more expert because the institution is optimized to take up that language. That’s a form of epistemic injustice.
Bounded transparency: demand auditable artifacts (methods, data access, prereg) rather than PR-style simplifications that merely feel clear.

5) Oppy’s level: theory-vs-theory comparison (derivations only as reductio)

With local pieces in place, compare best rival worldviews (T vs ¬T):

Try to force internal incoherence (where derivations matter, as reductios).
If both survive, compare theoretical virtues (simplicity, explanatory scope, fit with total evidence).
Don’t expect one “killer argument” to dismiss all disputes; expect incremental belief-revision pressure across the full package.

This reframes public debate away from “stockpiles of arguments” and toward package comparison, which also surfaces where “clarity hacks” are doing illegitimate work.

6) Error metabolism: institutionalize learning

Nguyen, drawing on Wimsatt, emphasizes that limited agents need systems that detect and metabolize error (not ideals that assume unlimited cognition). Build: red-team drills, public post-mortems, track “time-to-notice / time-to-repair.”
And expect adversaries to stack cognitive pleasures, epiphany, group belonging, moral outrage, and gamified progress, to keep people in the trap; design responses with that psychology in view.
Playfulness sprints: brief, good-faith runs where you try on rival models (Nguyen on playfulness vs traps).

7) Anti-clarity operations (practical moves)

Clarity tripwire (always on): If an explanation suddenly makes everything easy to categorize or explain, raise the investigation priority; don’t close it. (This flips clarity’s “thought-terminator” into a “keep digging” alarm.)
Pre-emption audit: Does the view contain a built-in story that downgrades all outsiders? If yes, treat future counter-evidence as diagnostic (the system may be converting refutation into confirmation).
Metric check: When a single number (score/rank) is doing too much work, force a plural-metrics view or revert to qualitative assessments to recover the lost nuance.
Fluency friction: Add a “legibility tax” to overly smooth presentations—require explicit alternatives/uncertainties. (Nguyen’s discussion of cognitive ease/fluency is the mechanism; clarity can be manufactured.)

8) Rule Governed Dialogue Constraints Aimed at Resolution

Rationality cannot be decoupled from it's pragmatic and social elements. Like rules governing a game, normative considerations govern our interactions within the context of a dialogue; which have bearings on the conclusions we deem acceptable.

Identify Communicative Goals: What is the purpose of this dialogue? What are we trying to do?
Establish Rules of Engagement: Is the dialogue structured such that it is truth enabling?
Rules as Constraints: Without rules, communication can become derailed and ineffective for rational resolution.

Search This Blog

CriticalThinkingAcademy