What Is Knowledge

WHAT IS KNOWLEDGE? epistemology — the study of knowing. what separates KNOWING something from merely believing it, or guessing right? 1 · KNOWLEDGE = JUSTIFIED TRUE BELIEF (the classic def.) BELIEF TRUTH JUSTIFICATION KNOW you KNOW p only if: you believe p, p is TRUE, and you're JUSTIFIED (have good reason). all three, or it's not knowledge. 2 · THE GETTIER PROBLEM — JTB isn't enough a STOPPED clock you glance at a clock: it reads 3:00. you believe 'it's 3:00' — justified (clocks are usually right) — and it really IS 3:00. but the clock STOPPED 12 hours ago! justified + true + belief... yet you don't KNOW — you got lucky. the justification is disconnected from what makes it true. so JTB is NECESSARY but not SUFFICIENT for knowledge. PATCHES people have tried: • NO FALSE LEMMAS — your reasoning used no false step • RELIABILISM — belief from a RELIABLE process counts • CAUSAL / VIRTUE — the fact must CAUSE the belief / it's due to your intellectual skill, not luck. (none fully settles it.) 3 · WHERE DOES KNOWLEDGE COME FROM? RATIONALISM from REASON alone, prior to experience (a priori). math, logic. Descartes, Leibniz EMPIRICISM from EXPERIENCE, the senses (a posteriori). science. Locke, Hume vs KANT's synthesis: the mind has built-in structures (space, time, causation) that SHAPE raw experience into knowledge. the five usual SOURCES: perception · reason · memory · testimony · introspection a priori — knowable by thought alone (7+5=12). a posteriori — needs the world (water boils at 100C). analytic — true by meaning ('bachelors are unmarried'). synthetic — adds real content about the world. the split between these is epistemology's oldest fault line. 4 · SKEPTICISM — can we be sure of ANYTHING? brain in a vat? Descartes' method of DOUBT: reject anything you could POSSIBLY be wrong about. an EVIL DEMON could be faking all your senses — modern versions: brain-in-a-vat, the SIMULATION. how could you tell from the inside? you can't. but one thing survives: 'I THINK, THEREFORE I AM' (cogito). even if deceived about everything else, the DOUBTING itself proves a doubter exists. Descartes' one indubitable anchor. GLOBAL skepticism (know nothing) is hard to refute but self-undermining. most philosophy since is a reply to it: we can't have CERTAINTY, so we settle for JUSTIFIED belief that's good enough to act on. (→ science, board 05.) the value of doubt: it's the engine of inquiry — never trust an unexamined belief. (cf. systematic debugging: reproduce, don't assume.) 5 · WHAT JUSTIFIES THE JUSTIFICATION? every reason needs a reason... (Agrippa's trilemma): 1) INFINITE REGRESS — reasons all the way down, forever, or 2) CIRCULARITY — reasons that loop back on themselves, or 3) it stops at an ARBITRARY, unjustified foundation. the two main escapes: FOUNDATIONALISM — some beliefs are BASIC/self-evident; the rest rest on them. (a pyramid on bedrock.) COHERENTISM — no bedrock; beliefs justify each other as a consistent WEB. (a raft, not a pyramid.) AI TIE: does an LLM KNOW things? it has belief-like patterns, often TRUE, but no grounding in experience — HALLUCINATION = confident belief with no truth or justification. RAG adds a SOURCE (a justification / reliabilist check). 'does it truly KNOW, or just predict the next token?' → board 04. whether prediction-from-text can amount to KNOWING is the live question the AI track raised. we answer it in board 04. EPISTEMOLOGY IN ENGINEERING (shop / Agent) • 'it works on my machine' = BELIEF without justification. a passing TEST = justification (reliabilist) — but a flaky green can be Gettier: true-and-passing by luck, not because the code is right. • OBSERVABILITY is perception for systems; LOGS are testimony/memory; a REPRO is the causal link that turns a hunch into knowledge (why 'reproduce, don't guess' beats confident belief). • LLM outputs are JUSTIFIED-sounding true-ish beliefs — treat them as testimony from an unreliable witness: demand a source (RAG/citation) before you 'know', exactly as reliabilism prescribes. JARGON, DECODED epistemology — the theory of knowledge JTB — justified true belief (the classic definition) Gettier case — JTB that isn't knowledge (true by luck) a priori / a posteriori — known by reason / by experience rationalism / empiricism — reason-first / experience-first reliabilism — knowledge = belief from a reliable process cogito — 'I think therefore I am'; Descartes' bedrock foundationalism / coherentism — bedrock beliefs / a web skepticism — the doubt that we can know (with certainty)