01 / 13
A Stanford-style primer in thirteen plates

EPISTEMOLOGY
What can be known?

§ · § · §

From Plato's Theaetetus to the replication crisis — a brief survey of the theory of knowledge: its objects, its sources, and the limits of what we may justifiably claim to know.

Plate I  ·  13 slides  ·  read 12 min
Plate II · The classical analysis

02Knowledge as justified true belief


The dominant working definition of propositional knowledge — knowledge that p — held from antiquity through the mid-twentieth century. To know a proposition is to satisfy three jointly necessary, jointly sufficient conditions.1

Definition 2.1 · JTB

S knows that p if and only if…

  1. Belief. S believes that p.
  2. Truth. p is true.
  3. Justification. S is justified in believing that p.
Belief Truth Justification K ← knowledge sits at    the threefold meet
Fig. 2.1Knowledge (K) as the intersection of the three conditions.

1 Plato, Theaetetus 201c-d. Socrates entertains "true judgment with an account" and finds each proposed account wanting.

Plate III · c. 369 BCE

03Plato — the Theaetetus


The first sustained Western treatise on the question what is knowledge? Socrates and the young geometer Theaetetus together test, and reject, three definitions of epistēmē.2

The three attempts

  • Knowledge is perception. Refuted: perception is private and shifting; knowledge must be stable.
  • Knowledge is true belief. Refuted: a juror may believe truly but without insight.
  • Knowledge is true belief with an account (logos). Refuted: what counts as an account?

Why it still matters

The dialogue establishes the agenda for two and a half millennia: knowledge is not mere information, not mere correctness; something further — justification, account, ground — is required.

"True opinion accompanied by reason is knowledge, but that which is unaccompanied is outside the sphere of knowledge." Theaetetus 202c

2 The contrast epistēmē / doxa structures Books V–VII of the Republic as well.

Plate IV · Meditations, 1641

04Descartes — radical doubt


Descartes inaugurates modern epistemology by inverting the question: rather than ask what we know, ask what we cannot doubt. The method of doubt rejects every belief admitting the slightest uncertainty — senses, mathematics, the external world — until something irrefutable remains.3

Argument 4.1 · the cogito

That I am, that I exist, is necessarily true so often as it is uttered.

Even an omnipotent deceiver cannot deceive me into thinking I exist if I do not. The act of doubting presupposes a doubter. Hence: cogito, ergo sum.

The three waves of doubt

  • The senses sometimes deceive.
  • The dream argument: I cannot tell waking from dreaming.
  • The evil demon: a being who systematically misleads me.

What is recovered

From the cogito Descartes attempts to reconstruct knowledge of God, the external world, and mathematics — an architectonic of foundationalism: certainty resting on indubitable first beliefs.

3 Descartes, Meditations on First Philosophy (Paris, 1641), trans. Cottingham 1986.

Plate V · British Empiricism, 17th–18th c.

05Empiricism & the problem of induction


For Locke and Hume, the mind is a tabula rasa: all ideas trace, ultimately, to sensory impressions. Knowledge is the patient construction of experience.4 But Hume then exposes a wound at the heart of the empiricist programme.

Problem 5.1 · Hume's induction

Why expect the future to resemble the past?

No deductive argument warrants induction (the future could differ). No inductive argument warrants it without circularity (we'd be assuming what we wish to prove). The bedrock of empirical science thus has no purely rational foundation — only custom and habit.

Locke (1690)

Distinguishes simple from complex ideas, primary from secondary qualities. Knowledge is the perception of agreement among ideas.

Hume (1739)

Splits all reasoning into relations of ideas (analytic, certain) and matters of fact (synthetic, contingent). Causation is not observed but inferred — constant conjunction, not necessity.

4 Locke (1690); Berkeley, Principles (1710); Hume, Treatise of Human Nature (1739–40), Bk I.

Plate VI · Continental Rationalism

06Rationalism — knowledge a priori


Against the empiricists, Spinoza and Leibniz hold that reason alone can deliver substantive truths about reality. Mathematics is the paradigm: necessary, universal, gained without observation.5

Spinoza (1677)

The Ethics proceeds more geometrico — in the manner of geometry — deriving the structure of substance, mind, and ethics from definitions and axioms. The highest knowledge is scientia intuitiva: direct intellectual grasp of essences.

Leibniz (1714)

Distinguishes truths of reason (necessary, knowable through analysis) from truths of fact (contingent, requiring the principle of sufficient reason). Innate ideas are dispositions, "veined" in the marble of the mind.

Distinction 6.1 · the analytic / a priori grid
  • A priori. Knowable independently of experience. e.g. 7 + 5 = 12.
  • A posteriori. Knowable only through experience. e.g. snow is white.
  • Analytic. True in virtue of meaning. e.g. all bachelors are unmarried.
  • Synthetic. True in virtue of extra-linguistic fact.

5 Spinoza, Ethica (1677); Leibniz, Monadologie (1714); Descartes belongs to this lineage as well.

Plate VII · Critique of Pure Reason, 1781

07Kant — the synthetic a priori


Roused by Hume from his "dogmatic slumber," Kant proposes a Copernican revolution: rather than ask how the mind conforms to objects, ask how objects conform to the mind. Experience is structured by the knower.6

Thesis 7.1 · Kant's central claim

There exist judgements that are both synthetic (informative) and a priori (necessary, universal).

E.g. the propositions of arithmetic and geometry; the principle that every event has a cause. Such judgements are possible because space, time, and the categories of the understanding (substance, causality, unity, …) are forms imposed by the mind on all possible experience.

Phenomena vs. noumena

We know things only as they appear to us (phenomena), structured by our cognitive forms. The thing-in-itself (noumenon) is forever beyond reach.

Why it mattered

Kant reconciles empiricism and rationalism: experience supplies content, the mind supplies form. Modern philosophy of science, mathematics, and cognition all begin here.

6 Kant, Prolegomena (1783) is the readable companion volume.

Plate VIII · Analysis 23, 1963

08Gettier — JTB is not enough


In a three-page paper, Edmund Gettier exhibits cases in which a person has a justified true belief that nonetheless seems not to count as knowledge. JTB is therefore not sufficient.7

Case 8.1 · Smith and the job

Smith is justified in believing Jones will get the job, and Jones has ten coins in his pocket. He validly infers: the man who will get the job has ten coins in his pocket. As it happens, Smith gets the job — and Smith, by coincidence, has ten coins in his pocket. Smith's belief is justified, and true. But did Smith know?

Justified belief Jones gets the job + has 10 coins. (strong evidence) Inferred belief The job-getter has 10 coins. (valid deduction) True — but luckily Smith gets it, Smith has 10 coins. (coincidence) infer obtains
Fig. 8.1The belief is justified, the belief is true — yet it seems an accident that the two coincide.

7 Decades of "post-Gettier" analyses (no-false-lemmas, sensitivity, safety) attempt a fourth condition. None has won universal assent.

Plate IX · Externalism, 1970s–

09Reliabilism — processes, not reasons


One influential post-Gettier response abandons the inward demand for justification as the agent sees it, and asks instead about the process that produced the belief.8

Definition 9.1 · Process Reliabilism

S knows that p iff S's true belief that p was produced by a reliable cognitive process.

A reliable process is one that tends, across a range of relevant counterfactual conditions, to yield true beliefs. Vision in good light is reliable; wishful thinking, guessing, paranoia are not.

Why it appeals

  • Explains animal and infant knowledge without elaborate inner reasons.
  • Naturalises epistemology: cognitive science can study reliability empirically.
  • Handles many Gettier cases: lucky truth is precisely what reliable processes are not.

Standing objections

  • The generality problem. Which "process" — described how broadly — is the relevant one?
  • The new evil demon. A brain in a vat with our exact mental life seems justified, yet its processes are unreliable.

8 Externalist alternatives: Nozick's tracking account; Sosa's virtue epistemology; Williamson's knowledge first.

Plate X · The skeptical hypothesis

10Skepticism — the brain in a vat


The modern descendant of Descartes' demon. A brain, suspended in nutrient fluid, receives electrochemical inputs from a supercomputer that perfectly simulate ordinary experience. From the inside, you cannot tell the difference. So — how do you know you are not such a brain?9

Vat Simulator Inputs — "I see a hand." → Outputs — motor signals ←
Fig. 10.1The skeptical scenario: experience is preserved exactly; its causes are not.
The skeptical argument
  1. If I know I have a hand, I know I am not a brain in a vat.
  2. I do not know I am not a brain in a vat.
  3. ∴ I do not know I have a hand.

Responses include contextualism (Lewis, DeRose), contrastivism, semantic externalism (Putnam), and Moorean common-sense rejection of premise 2.

9 The argument generalises Descartes' demon and anticipates the Matrix-style scenarios of contemporary philosophy of mind.

Plate XI · Social epistemology

11Testimony, expertise, the wisdom of crowds


Most of what any one person knows — the date of the French Revolution, the structure of DNA, the existence of Antarctica — was learned from testimony, not direct observation. Knowledge is irreducibly social.10

Testimony

Is testimonial belief justified by default (anti-reductionism, Reid) or only when independently corroborated (reductionism, Hume)?

Expertise

The novice/expert problem: how does the layperson rationally identify whom to trust, when by hypothesis she lacks the expertise to evaluate the claims?

Aggregation

Condorcet's jury theorem. If voters are independent and better than chance, majority verdicts approach certainty as n → ∞ — provided independence holds.

Caution 11.1 · epistemic pathologies of groups
  • Echo chambers — insulation from countervailing testimony.
  • Information cascades — later agents rationally ignore their evidence in deference to earlier ones, propagating error.
  • Epistemic injustice (Fricker) — testimony discounted on the basis of speaker identity.

10 Miranda Fricker, Epistemic Injustice (Oxford, 2007), distinguishes testimonial from hermeneutical injustice.

Plate XII · Contemporary epistemology

12Bayesianism, humility, and the replication crisis


Twenty-first century epistemology has turned increasingly formal and empirical. Belief is modelled in degrees, updated by evidence; the actual practices of science are scrutinised for the systematic errors they incubate.11

Definition 12.1 · Bayesian updating

P(H | E) = P(E | H) · P(H) / P(E)

Beliefs are credences in [0, 1]. Rational agents update by conditionalisation: posterior ∝ likelihood × prior. Epistemic humility = a prior that is not pinned to 1 or 0.

The replication crisis

The Open Science Collaboration's 2015 effort replicated only 36% of 100 high-profile psychology results. Subsequent failures in cancer biology, economics, and social priming followed. A live case study in fallible collective knowledge production.

Diagnoses

  • Publication bias toward novel positive findings.
  • Garden of forking paths — flexibility in analysis.
  • Underpowered designs and unregistered hypotheses.
  • Status incentives misaligned with truth.
"It ain't what you don't know that gets you into trouble. It's what you know for sure that just ain't so." attrib. Mark Twain

11 Open Science Collaboration, "Estimating the reproducibility of psychological science," Science 349.6251 (2015): 943.

Plate XIII · Further reading

13References & video


Primary & classical

  1. Plato, Theaetetus, c. 369 BCE.
  2. Descartes, Meditations on First Philosophy, 1641.
  3. Locke, Essay Concerning Human Understanding, 1690.
  4. Hume, An Enquiry Concerning Human Understanding, 1748.
  5. Kant, Critique of Pure Reason, 1781 / 1787.

Modern

  1. Gettier, "Is Justified True Belief Knowledge?", Analysis 23 (1963).
  2. Goldman, "What is Justified Belief?", 1979.
  3. Putnam, Reason, Truth, and History, 1981.
  4. Fricker, Epistemic Injustice, 2007.
  5. Open Science Collaboration, Science 349 (2015).

Video lectures


End of plate XIII  ·  Finis.