June 25, 2026 · American Pragmatism & the Unity of Knowing and Acting
Pragmatism — Instead of asking "does this idea correspond to reality," ask "what difference does it make in practice"
Pragmatism is America's homegrown philosophy, born in the late 19th century. Its revolution lies in a reversed question: instead of fretting over "does this idea correspond to reality," ask "what observable difference does it make in practice." Peirce used it to clarify meaning, James to redefine truth, Dewey to remake education. And three centuries earlier, on the other side of the globe, Wang Yangming had already struck the same target with "the unity of knowing and acting" — the split between knowing and doing. The shared enemy of all four is the "spectator theory of knowledge" that treats us as mirrors passively reflecting reality.
Charles S. Peirce 查尔斯·皮尔士
West · Pragmatism
How to Make Our Ideas Clear (1878); founder of pragmatism; 1839–1914
Core Thesis · Original Text
Consider what effects, that might conceivably have practical bearings, we conceive the object of our conception to have. Then, our conception of these effects is the whole of our conception of the object. — How to Make Our Ideas Clear
Thesis: the entire meaning of a concept lies in its conceivable practical effects. If two ideas differ in no conceivable consequence, they mean the same thing; a distinction that names no practical difference is empty.
Context & Core Insight
Peirce was a logician and scientist who co-founded the Cambridge "Metaphysical Club." He set out to cure metaphysical disputes that could never be settled. His "pragmatic maxim" is not a worldview but a machine for clarifying meaning: it translates abstract argument into "what difference would we observe?" Later, finding James's use of the method too loose, he renamed his own version "pragmaticism" to set it apart.
Cross-disciplinary cross-reference
Peirce also proposed "abduction" — inferring the best explanatory hypothesis from a surprising phenomenon. This is genuinely isomorphic with the logic of scientific discovery and machine learning's "hypothesis generation": neither deducing from data nor merely generalizing, but "leaping" to a hypothesis that explains the data, then testing it by its consequences. This is also the very spirit of "operational definitions" in physics.
Contemporary Relevance
BigCat scenario: Peirce hands you a razor for cutting empty talk. Facing a clash of two vocabularies or two roadmaps, first ask: "Which observable difference would each produce?" If you can't name a single one, the dispute is a pseudo-problem. In AI strategy and product definition, this one question drags metaphysics back to earth.
Essence · Question
The irreplaceable insight: meaning lies not in the definition but in the consequences — a distinction that names no practical difference is no real distinction.
In the last argument you got drawn into, which observable difference did each side's claim produce? If none, was it worth fighting over?
William James 威廉·詹姆斯
West · Pragmatism · Psychology
Pragmatism (1907), The Principles of Psychology; 1842–1910
Core Thesis · Original Text
The truth of an idea is not a stagnant property inherent in it. Truth happens to an idea. It becomes true, is made true by events. — Pragmatism, Lecture VI
Thesis: truth is not a static "correspondence" between idea and reality but a process: a belief that keeps getting verified by experience, and that guides us smoothly in action, "becomes true." Truth's "cash value" is the practical payoff it cashes out in experience.
Context & Core Insight
James boldly pushed Peirce's maxim for clarifying meaning into a full theory of truth (to Peirce's displeasure). He targeted rationalism's "correspondence theory" — truth as a mirror statically reflecting reality. James countered: how do you know the mirror got it right? Only by whether it "works" in subsequent experience. Truth is therefore a verb, not a noun: it is verified into being, not discovered ready-made.
Cross-disciplinary cross-reference
"Truth happens to an idea" genuinely echoes the predictive-processing view of the brain: the brain constantly generates predictions and uses sensory input to "cash out" or falsify them; a belief is judged by its success at predicting experience. But James also left a famous risk — equating "useful" directly with "true." In the post-truth and AI age, this is exactly the trap to watch: "what soothes me" is not "what predicts accurately."
Contemporary Relevance
BigCat scenario: Judge a belief or model by its "track record of cashing out" in experience, not by how elegant it is. But hold the line James failed to hold: separate "useful to my psychology" from "useful for predicting the world" — the former is comfort, the latter is the cash of truth.
Essence · Question
The irreplaceable insight: truth is a verb — an idea is not "true to begin with" but "made true" in experience.
Your most firmly held judgment — when was the last time experience actually "cashed it out" or falsified it?
John Dewey 约翰·杜威
West · Pragmatism · Education
Logic: The Theory of Inquiry (1938), How We Think (1910); 1859–1952
Core Thesis · Original Text
Inquiry is the controlled or directed transformation of an indeterminate situation into one that is so determinate in its constituent distinctions and relations as to convert the elements of the original situation into a unified whole. — Logic: The Theory of Inquiry
Thesis: thinking begins in a real "indeterminate situation" (a perplexity, a fork in the road); inquiry uses controlled operations to transform the indeterminate into the actionably determinate. Knowledge is the tool of inquiry, not a mirror-copy of ready-made reality.
Context & Core Insight
Dewey pushed pragmatism into education and democracy, targeting the "spectator theory" running from the Greeks to Descartes — treating the knower as an eye standing outside the world, passively transcribing reality. Dewey said: a person is first an agent within a situation, out to solve problems; ideas are "plans of action," valued by whether they sort the difficulty out. Hence "learning by doing" — knowledge is generated in solving real problems, not by force-feeding inert facts.
Cross-disciplinary cross-reference
Dewey's "inquiry" is an act–observe–revise loop, genuinely isomorphic with the scientific method and agentic AI's "perception–action loop" (active inference / agent loop): the organism acts to dispel uncertainty, then updates on the result — which is also the core of experimental science and agile iteration.
Contemporary Relevance
BigCat scenario: Treat strategy as a hypothesis, not a truth to defend. Don't rush to the "right answer"; first frame the situation, design the smallest actionable experiment, then act–observe–revise. Likewise in parenting: let a child learn by solving real problems rather than stuffing them with inert knowledge.
Essence · Question
The irreplaceable insight: knowledge is a tool, not a mirror — its value is whether it can sort a situation that has you stuck into an actionable one.
Which of your "think it through first" plans would actually be better served by "start moving and think it through in action"?
Wang Yangming 王阳明
East · Confucian Mind-Learning
Instructions for Practical Living (Chuanxilu); master of Ming "mind" Confucianism; 1472–1529
Core Thesis · Original Text
知是行之始，行是知之成。知而不行，只是未知。 "Knowing is the beginning of acting; acting is the completion of knowing. To know and not act is simply not to know." — Instructions for Practical Living
Thesis: knowing and acting are originally one and inseparable. Genuine "knowing" necessarily contains "acting"; to know and not act is not yet to truly know. To know filial piety yet not act on it is never to have truly known it.
Context & Core Insight
Wang answered Zhu Xi's "investigating things to extend knowledge" — Zhu held that one must "know first, then act," exhausting the principles in external things before acting. After his enlightenment at Longchang, Wang denounced this split: it leaves people "never acting, and so never knowing for a lifetime," sunk in empty talk. He pulled "knowing" back from external doctrine to the immediate presence of the innate moral sense (liangzhi) — knowing and acting are two sides of one process, and "extending liangzhi" means honing it on actual affairs, realizing it in action.
Cross-disciplinary cross-reference
This is the issue's key contrast: Wang and American pragmatism share one enemy — "spectator knowledge" (the knowing–acting split, a static truth applied afterward), both tying "true knowing" to action. But the anchor differs: pragmatism places the criterion of "acting" in external experiential consequences; Wang internalizes it as the immediate moral earnestness of liangzhi — one outward and experimental, one inward and ethical. "Knowing and acting as one" also echoes embodied cognition: cognition is the coupling of perception and action, not first representing in the head then outputting movement.
Contemporary Relevance
BigCat scenario: "Knowing but not being able to do it" is the great illusion — Wang says that is simply not real knowing. In the AI age "knowing" is cheap; only knowledge that cashes out in action is scarce. Test whether you truly understand by whether you can do it. In parenting it's even more direct: example outweighs instruction — only a parent who unites knowing and acting can raise a child who does the same.
Essence · Question
The irreplaceable insight: "to know and not act is simply not to know" — an understanding you can't enact is no understanding at all.
The thing you "obviously know" yet have never managed to do — by Wang's standard you never truly "knew" it. Do you agree?
Visualization · Where the Criterion of "Acting" Falls
All four tie "true knowing" to "acting" and reject the spectator theory of knowledge-as-mirror; they differ only in where the criterion of "acting" falls — from Peirce's experimental consequences to Wang's inward moral earnestness.
Four thinkers, one shared enemy: treating us as mirrors passively reflecting reality. Peirce at meaning, James at truth, Dewey at inquiry, Wang at virtue — each strikes through the split between "knowing" and "acting." For the "AI super-individual," this yields one hard discipline: in an age when "knowing" is dirt cheap, only knowledge that cashes out in action counts.
Deeper Reflection
Wang struck through the knowing–acting split three centuries before pragmatism. Great minds thinking alike, or a surface resemblance?
A surface resemblance with a different soul. Both reject the spectator theory and tie true knowing to action; the difference is the criterion of "acting." Pragmatism faces outward, anchored in publicly verifiable experiential consequences; Wang faces inward, anchored in the moral earnestness of conscience. The former asks "does it work?"; the latter, "does it accord with conscience?" The former might count any "effective" act as true, the latter carries a built-in moral anchor — but its weakness is the lack of external public testing.
Why is James's "useful = true" especially dangerous in the AI and post-truth age?
James tied truth to "what works" but conflated two senses of "useful": psychological comfort and predictive accuracy. Algorithms specialize in the former — beliefs that soothe, addict, and harden bias are "useful to you" yet need not be useful for predicting the world. Peirce and Dewey's antidote is to tighten "useful" to "publicly verifiable experimental consequences," not private satisfaction.
Dewey says "knowledge is a tool, not a mirror" — how does this echo the "capability" view of large AI models?
Closely. A large model has no "inner truth corresponding to reality"; its "knowing" is precisely instrumental — whether it cashes out on downstream tasks. This is thoroughgoing Deweyan instrumentalism: understanding is redefined as "the capacity to act effectively in a situation," not mirror-representation. But it leaves the Deweyan puzzle: does knowledge defined purely by "what works" miss some "truth" we still want?
If "the unity of knowing and acting" holds, can AI "know"?
A lovely challenge. By Wang, knowing not realized in action is not true knowing; and AI can output vast statements it "can't act out" — it "knows" the definition of filial piety yet cannot perform it, the emptiest kind of purely symbolic "knowing." But agentic AI has begun to "act" — and as knowing and acting start to unite in it too, Wang's standard may become a yardstick for whether a machine "truly understands."