A man loses his ax. He suspects his neighbor stole it. He watches his neighbor walk, and it looks like the walk of a man who stole an ax. He watches his neighbor speak, and it sounds like the speech of a man who stole an ax. He observes his neighbor's expressions, his movements, his manner — and every feature confirms it: this is an ax-thief. Then the man finds his ax, buried in his own yard where he left it. He looks at his neighbor again. The walk no longer looks like that of an ax-thief. The speech no longer sounds like it. The expressions reveal nothing incriminating. The neighbor is precisely the same person he was the day before; everything the man is observing is identical. What changed is the premise.1
This is the paradigm case in the Lü-shih Ch'un-ch'iu — a Chinese philosophical compendium compiled approximately 239 BCE — and it is a more precise description of confirmation bias than most modern treatments offer. The sickle-thief story (the text's version uses an ax, sometimes rendered as sickle in other translations) does not merely point out that people selectively attend to confirming evidence. It identifies the mechanism: the interpretation of incoming perceptual data is governed by the premise in advance of the data. The neighbor's behavior did not change. The perceptual apparatus processing it was the same. The premise changed, and so every output of the apparatus changed with it.
Kahneman and Tversky identified this as a defining feature of System 1 processing in the 1970s and 1980s. The Lü-shih Ch'un-ch'iu identified it as a defining failure mode of political judgment in 239 BCE.1
The Lü-shih Ch'un-ch'iu's catalog goes further than the paradigm case — it lists forty-six sources of cognitive obfuscation (pi) that corrupt assessment. The text attributes much of this taxonomy to Hsün-tzu (ca. 310–235 BCE), whose chapter "Untangling Obfuscations" (Chieh-pi) is the most systematic early treatment of what we now call cognitive interference in political and intelligence contexts.
Hsün-tzu's classification of obfuscating factors includes: desire (which causes a man to see what he wants to see rather than what is there); aversion (which causes him to see what he fears whether it is there or not); beginning (which causes him to fixate on the initial frame and ignore contradicting evidence as it arrives); end (which causes him to interpret earlier evidence in light of a later conclusion he has already drawn); near (which makes the proximate seem more significant than the distant); far (which causes the remote to seem more reliable than the near, through a prestige of distance); breadth (which causes superficial comprehensiveness to be mistaken for depth); and narrowness (which causes specialist attention to miss what lies outside its category).
This is recognizably a pre-modern taxonomy of cognitive biases — not merely as logical fallacies but as systematic distortions in the perceptual processing of evidence, rooted in motivational states (desire, fear), framing effects (beginning, end), and attention patterns (near, far, breadth, narrowness).1
The Chinese texts address the sickle-thief problem not as an abstract epistemological puzzle but as a catastrophic failure mode in political and military judgment. The frame is always: how does a ruler, commander, or advisor correctly assess a situation when their prior beliefs — about a minister's loyalty, an ally's intentions, an enemy's capabilities — have already primed their interpretive machinery?
The intelligence problem that confirmation bias creates is not simply that you might misread a situation. It is that the very effort to assess carefully might increase the confidence of the misreading. The more data you collect from a premise-governed observational process, the more confirming evidence you accumulate, and the more certain you become of a conclusion you arrived at before you began. The Lü-shih Ch'un-ch'iu documents this as a specific danger: the careful observer who has formed a strong prior is, paradoxically, more dangerous than the inattentive one — because they have more confirmatory observations to point to, and more confidence in the wrong conclusion.1
This is also why the chih jen (knowing men) tradition (see Knowing Men — Chih Jen) insisted on structured, multi-domain protocols rather than open-ended behavioral observation. Confucius's three-step framework and King Wen's Six Indications both function as anti-confirmation-bias structures: they specify in advance what will be observed and in what sequence, precisely to prevent the observer from allowing their prior assessment of the subject to govern what they see.
The six-method and eight-method active test sequences (see Active Testing Protocols) serve the same function: they impose a predetermined stimulus on the subject and record the response, rather than allowing the observer to select which behaviors to attend to. Pre-specified behavioral tests resist premise-governance more effectively than open observation — which is why the tradition developed them.1
The confirmation bias problem also powers the Contrary Employment doctrine (see Contrary Employment Doctrine). A sophisticated opponent who understands that trained observers have established interpretive heuristics can exploit precisely the premise-governance that confirmation bias describes. If the observer believes they are seeing a retreating force (because they saw the canonical signals of retreat), every subsequent piece of ambiguous data will be interpreted as confirming the retreat. The manufactured premise governs all subsequent observation. Contrary Employment is confirmation bias deployed as a weapon: induce the false premise, then allow the observer's own interpretive machinery to do the rest.1
The Chinese tradition documented confirmation bias as a practical failure mode in political and military judgment. Modern behavioral science (Kahneman/Tversky/Festinger) documents it as a descriptive feature of human cognition with experimental corroboration. These are the same phenomenon approached from different methodological directions, and what their convergence reveals is significant:
The Chinese tradition discovered confirmation bias through the failure of political assessment — through watching sovereigns and generals make consistently wrong judgments after forming strong priors. Modern behavioral science discovered it through controlled laboratory experiments in which subjects consistently exhibited the bias under controlled conditions. One is a naturalistic observation of catastrophic real-world instances; the other is a controlled demonstration of the mechanism in low-stakes experimental contexts. Both arrive at the same structural description. Their convergence from opposite methodological starting points constitutes independent corroboration: this is not merely a logical fallacy that could be avoided by more careful reasoning; it is a property of how human perceptual processing works under motivated or primed conditions.1
Confirmation bias as ancient problem connects to two other domains where the same distortion in premise-governed observation appears with different surface features but identical underlying structure.
Psychology: Bodily Repression Mechanism — the repression mechanism that Lowen describes operates partly through what the Chinese tradition would recognize as a desire-based obfuscation: the organism does not want to see certain emotional truths (grief too large to bear, rage too dangerous to express, fear too threatening to acknowledge), and so the perceptual processing that would generate those recognitions is systematically suppressed. This is the sickle-thief problem applied intrapsychically: the premise ("this feeling cannot be mine") governs all subsequent observation of one's own internal states. The Chinese tradition's forty-six obfuscating factors include desire and aversion as the two most powerful; Lowen's repression mechanism is precisely these two factors applied to the observation of one's own body. The cross-domain insight: confirmation bias is not only an interpersonal phenomenon (misreading others); it is also an intrapersonal one (the premises governing self-observation are as premise-governed as observations of the external world).
History: Estrangement Techniques — the estrangement doctrine works by installing a false premise (the minister is treacherous, the ally is defecting) that then governs all subsequent observation. Once Hsiang Yü's premise about Fan Tseng was corrupted, every action Fan Tseng took that could be read as treacherous was read as treacherous — just as the man with the lost ax read every behavior of his neighbor as ax-thief behavior. The estrangement operator does not need to manufacture specific confirming evidence; they only need to install the premise. The sickle-thief story is the operating principle of estrangement.
The Sharpest Implication
The Lü-shih Ch'un-ch'iu's sickle-thief paradigm implies something that neither the Chinese tradition nor modern behavioral science has fully confronted: the problem is not that we sometimes form premature premises that govern our subsequent observation. The problem is that all observation is premise-governed — that there is no observation without interpretive priming, and therefore no confirmation-bias-free assessment of any complex situation. The tradition's response was to develop structured observational protocols that minimize premise-governance. But the protocols themselves were developed from premises. The only honest response to the sickle-thief story is that the man found his ax, and became less certain — not more. The absence of the confirming premise produces less certainty, not more knowledge. And we are almost never in the position of finding the ax.
Generative Questions