Cognitive Bias as Default Setting: An Architectural Argument

A person's biases are the person. They are the accumulated content the person's judgment is based on, present at every moment, shaping every output, observable in how the person judges anything they meet. The cognitive-bias literature has treated bias as something else, an error against a standard the field imported from outside the person being judged. The treatment does not survive the architecture of the thinking it claims to measure. The defaults that produce bias are stored continuously, not triggered by conditions. A person carries their accumulated content whether or not any current input makes it visible. What changes situationally is the input the content meets, not whether the content switches on or off. The cognitive-bias literature, since 1974, has treated bias as deviation from a standard imported by a third-party experimenter, Bayesian probability, evolutionary fitness, reflective rationality. The architecture distinguishes effect from measurement. One person's content reaches another person's cycle when the recipient's own cycle processes the supplied input, which is how teaching, conversation, and group norming work. Measurement is the different claim that a third party can read another person's thinking from outside the cycle that owns it, and this is the move the architecture forbids. The experimenter sits outside the participant, applies an imported standard to a re-description of the participant's behavioral output, and calls the resulting gap a bias. The standard's validity is not in dispute. The application is. The application evaluates the experimenter's re-description against the experimenter's chosen standard, both content in the experimenter's thinking, and the participant's actual judgment process is not in the loop. The imports the literature has been making are therefore not bias measures of the participants to begin with. The architectural claim that no third party reaches inside another person's thinking is not testable from outside thinking, and the unfalsifiability is the architecture confirming itself. The architectural prediction the claim generates about the cognitive-bias literature's record is testable against the field's own published methods and verdicts and is met by them. The asymmetry between the unfalsifiable architectural claim and the falsifiable architectural prediction is what makes the argument hold. Every person's defaults are accumulated content from the layered groups the person has been through. Tuckman's forming, storming, norming, performing structure, developed on small groups, is proposed here to operate at every scale of persistent group, from family to nation, with the norming each group settles into becoming shared content for every member's subsequent judgments in that group's domain. Different group memberships deposit different content, and there is no neutral reference point from which one set of defaults is the unbiased one. A meaningful subset of the existing named-bias literature is, on this reading, not bias at all. The defaults the literature's entries describe are operating normally. What the experimenters measured as deviation is deviation from an imported standard, whether the field's normative target, the culture's expectation, or the researcher's own commitment, that the architecture of thinking forbids. Five cases make the point concrete across two structurally distinct moves. Two named biases dissolve under structural analysis as labeling schemes that hold heterogeneous defaults together with an imported standard the experimenters carried: the bystander effect (held by a moral standard about helping) and the fundamental attribution error (held by a culturally specific standard about how observers should weight situation versus disposition). Three named biases re-describe as the judgment cycle's normal operation on the content the task supplied. The temporal cluster (hindsight, planning fallacy, optimism bias) re-describes as differences in how the past, present, and anticipated future weight against each other at the moment of judgment. Confirmation bias re-describes as the contextualization frame's normal operation on the held hypothesis. Anchoring re-describes as the recognition admission of a numerical signal that the contextualization frame's selective-accessibility operation then carries into resolution. None of the five is an error against an outside measure. **Keywords:** cognitive bias, heuristics and biases, judgment and decision making, bounded rationality, ecological rationality, normative standards, dual-process theory, in-group bias, social identity, Tuckman model, anchoring, confirmation bias, fundamental attribution error, bystander effect