Epistemological Theories

• Epistemology studies rational cognition and how we can know various claims.

• Investigations occur on several levels:

  • Lowest level: Theories of particular kinds of knowledge (perceptual knowledge, induction, knowledge of other minds, mathematical knowledge, etc.).

  • Intermediate level: Topics that apply to most specific kinds of knowledge, such as theories of reasoning (deductive and defeasible).

  • Highest level: General epistemological theories explaining how justified belief in general is possible. Examples include foundationalism, coherentism, probabilism, reliabilism, and direct realism.

• Highest level theories can be seen in two ways:

  • Structural theories of epistemic justification: Describe overall structural relations that produce epistemic justification, explaining how cognition constituents fit together.

  • Analytic theories of epistemic justification: Logical analyses explaining why epistemic justification has its general structure.

• The distinction between structural and analytic theories is often vague; analytic theories explain what would make structural theories true.

• Foundationalism, coherentism, and direct realism are usually structural theories, while reliabilism is often seen as an analytic theory.

• Reliabilism can support a foundationalist structural theory by justifying it as a reliable way of acquiring knowledge, showing these theories are not necessarily incompatible.

• Foundationalists, coherentists, and direct realists often claim their theories are also analytic, meaning their structural accounts are constitutive of the concept of epistemic justification.

• These analytic theories explain justification by enumerating principles that give rise to it rather than an overarching principle like reliabilism.

• A main objection to foundationalist, coherentist, and direct realist analytic theories is that they provide piecemeal, ad hoc analyses, characterizing epistemic justification with a general structure plus many unrelated principles without a unifying account.

• This leads to inelegance and leaves open whether there is a deeper explanation unifying these principles.

• Probabilism and reliabilism aim to provide general, elegant accounts avoiding ad hoc problems, but neither fully succeeds.

• No familiar structural theory of epistemic justification gives a truly unified account, leaving the theoretical situation unsatisfactory.

• The question remains: What is the origin of the complex epistemic structure that yields justified beliefs?

• Insight may come from a careful study of the concept of epistemic justification itself.

• Distinction is important between procedural epistemic justification and a justification concept tied to knowledge and the Gettier problem.

• The focus here is on procedural epistemic justification, which will necessarily include a complex structure with many apparently unrelated epistemic principles.

• The key question is: What unifies these diverse principles as part of the true theory of epistemic justification?

• This question should be placed in a broader context: epistemic cognition is part of rational cognition in general.

• There is a traditional distinction between epistemic cognition (about what to believe) and practical cognition (about what to do).

• Rational agents are those who cognize rationally; correct epistemic cognition is a subspecies of rational cognition overall.

• Asking what makes epistemic procedures correct is a special case of what makes cognitive procedures rational.

• Procedural epistemology is therefore part of a general theory of rationality, which also must account for how goals are selected and how actions are chosen.

Human Rationality

• The question: “What makes rational procedures rational?” can be illuminated by examining philosophical methods for answering how to perform cognitive tasks rationally (e.g., inductive reasoning).

• Philosophers propose general principles (like the Nicod principle, Bayesian inference, hypothetico-deductive method) and test them on concrete examples.

• Testing requires knowing how the example should turn out; e.g., Nelson Goodman’s grue/bleen example demonstrated the need for a projectibility constraint in induction.

• Everyone agrees that inductive reasoning violating projectibility is irrational, but the standard answer for knowing this is “philosophical intuition,” which is vague.

• Compare this to linguists studying grammaticality, who construct theories tested by grammaticality judgments from proficient speakers.

• Similarly, humans can make rationality judgments about whether cognitive acts are rational or irrational; these judgments provide data to test theories of rationality.

• Some philosophers argue rationality is about concepts and logic, making intuitions a kind of Platonic intuition; but this view is not applied to linguistic intuitions, which depend on linguistic convention.

• Learning a language involves acquiring procedural knowledge—knowing how to do something—and there can be multiple procedural knowledges for a task.

• Once procedural knowledge is acquired, one can reliably judge conformity to it, detecting divergences before failure (e.g., knowing when one is about to fall off a bicycle).

• Competent speakers often produce ungrammatical utterances but can recognize and correct them on reflection.

• Similarly, cognition involves procedural knowledge for how to cognize, which includes the ability to recognize when one diverges from it.

• This ability to detect divergence is what “philosophical intuition” about rationality effectively amounts to.

• Performance theories describe how people behave; competence theories describe people’s procedural knowledge for performing tasks.

• People often fail to conform to their own procedural knowledge, so competence and performance can diverge greatly.

• Philosophers’ theories of rationality based on intuitions are best seen as competence theories articulating the rules of our procedural knowledge for cognition.

• In epistemic cognition, these rules correspond to our epistemic norms.

• We have no direct access to these rules but can detect divergences from them and use this to confirm general principles about procedural knowledge.

• Unlike linguistic knowledge, much of our procedural knowledge for cognition is innate, likely built in by evolution to avoid extreme epistemic vulnerability.

• Empirical evidence shows universal patterns in reasoning, such as modus ponens being natural and modus tollens initially unnatural but later learned.

• The rules of our procedural cognitive knowledge are not mere generalizations about behavior (performance) but rules we try to conform to; detecting divergence leads to correction of our cognition.

• Human rationality consists of principles built into our cognitive architecture.

• In epistemology, the diverse principles of a correct structural theory of epistemic justification are unified by being built into human cognition—there need not be a more general characterization explaining why these principles hold.

• The proposal: a theory of human rationality is a competence theory of human cognition, eliciting procedural norms governing cognitive performance.

• When such norms govern epistemic cognition, they are called epistemic norms.

• Procedural epistemic justification consists in holding beliefs in compliance with correct epistemic norms.

• Proposed analysis of epistemic justification:
  A belief is justified if and only if it is held in compliance with the cognizer’s epistemic norms.

• Distinction between acting in accordance with norms and acting in compliance with norms is crucial.

• Acting in accordance means behavior does not violate the norm, but might not be guided by the norm.

• Acting in compliance means behavior follows and is guided by the norm.

• Justification requires compliance, not merely accordance.

• This is a naturalistic analysis of epistemic justification.

• Cognition consists of natural processes governed by internalized norms.

• Epistemic norms are defined as norms that actually govern our cognition.

• No informative logical analysis of the governance process is expected or possible; it is a natural process observed empirically.

• The nature of governance can be clarified by psychological investigation only.

• This is similar to how we understand natural kinds like electrons or magnetism—through empirical investigation, not logical analysis.

• Objection: existing epistemological theories offer piecemeal and uninformative analyses of epistemic justification.

• The proposed analysis escapes this objection by providing a completely general and unified account of epistemic justification as compliance with procedural norms.

• Structural theories arising from this analytic theory will be piecemeal due to the specificity of procedural norms.

• Procedural norms are specific, responding only to directly accessible features of the current cognitive situation.

• Norms cannot appeal to broad general features like “reliability of the process.”

• Example: bicycle riding norms are very specific and situational (e.g., “If losing momentum, push harder”).

• Epistemic norms are similarly specific, e.g., “If something looks red and you have no reason to doubt it, you may believe it is red.”

• No single simple formula can combine all epistemic norms, just like no simple formula governs all bicycle controls.

• Contrast with conventional internalist epistemology: internalists claim epistemic norms describe how we actually reason.

• This internalist claim is misleading because we do not always reason correctly.

• Correct reasoning is precisely conforming to the epistemic norms, not merely how we happen to reason.

• Analogy: bike-riding norms describe how we ride correctly, not all actual riding behavior.

• Traditional accounts face a puzzle: norms describe correct action, but actual behavior can deviate.

• Puzzle resolved by understanding norms as those that actually guide correct behavior (bike riding or reasoning).

• Norms should be elicited from actual behavior, not from some separate external criterion.

• Important difference: this formulation does not take actual reasoning behavior at face value, but recognizes the role of norms in guiding correct behavior.