Confirmation Bias
Reveals the bias that makes all other biases harder to correct.
Onramp · Foundation · Primary Vulnerability
The Codex Lens
If you had to name a single cognitive vulnerability most responsible for the pattern, this would be it.
Confirmation bias is the mechanism by which you lock onto a trajectory and then filter all incoming information to maintain it. It is the reason you feel certain rather than closed: every piece of evidence that reaches conscious evaluation has already been pre-sorted to confirm the existing position. Contradictory evidence is not weighed and rejected. It is not seen at all. It is filtered out before it arrives, by a process you do not experience as filtering. You experience it as thinking clearly.
This is what makes confirmation bias different from simple stubbornness. A stubborn person knows there is contradictory evidence and refuses to engage with it. Under confirmation bias, you do not perceive the contradictory evidence as relevant, or as strong, or sometimes as existing. The bias operates on perception itself, not just on judgment. It shapes what you notice, how you interpret what you notice, and what you remember afterward. By the time you sit down to "think it through," the inputs have already been corrupted.
For the Meridian Range, this is the silent drift. Confirmation bias does not push you toward Control or Decay directly. It pushes you further in whichever direction you are already facing. If you lean toward certainty, it feeds you evidence that your certainty is justified. If you lean toward cynicism, it feeds you evidence that nothing can be trusted. It is an amplifier of existing trajectory, and it makes the amplification feel like reason.
This is also why intelligence does not protect against confirmation bias. It often makes it worse. A more intelligent person finds evidence for what they already believe faster, constructs more sophisticated arguments for their existing position, and spots flaws in opposing arguments while overlooking identical flaws in their own. Intelligence without Scout Mindset is confirmation bias with a higher processing speed.
The Concept
Confirmation bias is the tendency to seek, interpret, and remember information in ways that confirm existing beliefs while ignoring, dismissing, or forgetting information that contradicts them.
It operates across three stages, each compounding the distortion:
Selective seeking. Given a choice of what information to consult, minds preferentially seek sources that are likely to confirm what they already believe. This is not always conscious. It can be as subtle as which headline you click, which expert you find credible at first glance, which search terms you use. The information diet is biased before evaluation begins.
Selective interpretation. The same piece of evidence, presented to people with different prior beliefs, will be interpreted as supporting whichever position the person already holds. Ambiguous data does not produce convergence. It produces divergence. Each side reads the same study and concludes it supports their position. The distortion is perceptual, not moral. Expectation shapes what people see.
Selective memory. Over time, confirming evidence is remembered more vividly and more accurately than disconfirming evidence. The mental record of "what the evidence shows" is not a faithful archive. It is an edited collection that increasingly supports the existing belief, making the belief feel more evidence-based than it actually is.
The combined effect is that you sincerely believe you are following the evidence while operating a filtration system that ensures the evidence always leads to the same place. This is the immune system of false belief. It is the reason beliefs can survive indefinitely in the face of contradictory evidence, and why the holder of those beliefs experiences this survival as vindication rather than as a warning sign.
The Practice
The core discipline is active disconfirmation: deliberately seeking the evidence most likely to prove you wrong.
This runs directly against the grain of how you naturally operate. The natural move, when you hold a belief, is to look for support. The Foundation practice is to look for threat. Not because threat-seeking is comfortable. It is not. But because the bias already provides an overwhelming supply of confirmation. The deficit is always on the disconfirmation side. Actively seeking disconfirmation is correction, not overcorrection.
The disconfirmation question. For any belief you hold with confidence, ask: "What would I expect to see if this were wrong?" Be specific. Not "some evidence might show up" but "if this belief were false, I would expect to find X in the data, or Y in the historical record, or Z in the behavior of the people involved." Then look for X, Y, and Z with the same energy you would use to look for confirming evidence.
The source test. When you encounter evidence that supports your position, apply the same scrutiny you would apply if the evidence supported the opposing position. Ask: "If this exact study, with these exact methods and these exact results, supported the other side, would I still find it convincing?" If the answer is no, the evidence is weaker than your initial reaction suggested. Your reaction was driven by confirmation, not by the quality of the evidence.
The surprise audit. Periodically review your recent updates. When was the last time a piece of evidence surprised you? When was the last time you encountered something that shifted your position, even slightly? If you cannot identify a recent surprise, it is unlikely that your information environment is providing unfiltered input. Either you are in an echo chamber, or your filters are operating at a level that prevents disconfirming evidence from registering. Both are warning signs.
The diagnostic question is: "When was the last time evidence changed my mind about something I cared about?" If the answer is not recent and specific, confirmation bias is likely operating unchecked.
In the Wild
An investor had held a position in a company for two years. Every quarterly report, she focused on the metrics that supported her thesis and skimmed past the ones that did not. Revenue was growing. She told herself the story was intact. A colleague asked her a simple question: "If you did not already own this stock, would you buy it today at this price?" She realized she could not say yes. The debt-to-equity ratio had deteriorated steadily for five quarters. She had seen the numbers every time. She had not registered them as a problem because they contradicted a conclusion she had already committed to. She sold the position. Two quarters later, the company issued a profit warning and the stock dropped 40%. The data had been there all along. Her filters had made it invisible.
A hiring manager believed that candidates from certain universities performed better. Over five years, he had assembled a mental catalogue of successful hires that confirmed this. When asked to review actual performance data across all hires, the pattern disappeared. Candidates from his preferred universities performed no better on average. But his memory had been doing selective archiving for years: the successes from preferred schools were vivid and available, the failures were vague, and the successes from other schools had never been mentally tagged as evidence against his belief. His hiring decisions had been shaped by a dataset that existed only in his biased memory.
A couple in a long-running disagreement about household responsibilities each believed they were doing more than their share. Each could cite specific, vivid examples of times they had picked up the slack. Neither could easily recall the times their partner had done the same. They were not lying to each other. They were both experiencing selective memory: their own contributions were salient, their partner's contributions were background. When they spent a week writing down every task each person completed, the actual distribution was close to equal. The bias had been manufacturing a grievance out of a perceptual distortion.
Pick a belief you hold with confidence. Something that matters to you. Now ask yourself: when was the last time you looked for evidence that it was wrong? Not evidence that confirmed it, which your mind supplies automatically. Evidence against it. If you have not looked recently, the bias is running. It does not mean you are wrong. It means you do not know whether you are right, and you have been feeling certain anyway. That gap between certainty and knowledge is where confirmation bias lives.
Where This Comes From
The Codex did not discover confirmation bias. It is one of the most extensively studied phenomena in cognitive science. What follows is the intellectual history and where to go for deeper study.
Peter Wason identified and named confirmation bias through a series of experiments in the 1960s. His most famous experiment, the 2-4-6 task (1960), demonstrated that people overwhelmingly test hypotheses by looking for confirming instances rather than by attempting to falsify them. Subjects were given a rule to discover and consistently tested only examples they expected to work, rarely testing examples they expected to fail, which would have been far more informative. Wason's original papers remain fascinating reading for how cleanly they demonstrate the phenomenon.
The decades following Wason's initial work expanded the understanding from a laboratory finding into a pervasive feature of real-world reasoning. Raymond Nickerson's 1998 review, "Confirmation Bias: A Ubiquitous Phenomenon in Many Guises," synthesized hundreds of studies and established confirmation bias as perhaps the single most consequential cognitive bias, arguing that it accounts for a significant portion of disputes, errors, and poor decisions across every domain of human activity. Nickerson's review is the definitive academic survey for anyone who wants the full research picture.
Ziva Kunda's work on motivated cognition (1990) deepened the understanding of the mechanism: confirmation bias is not random error but directional distortion, driven by the goals and identity investments of the thinker. This connection between confirmation bias and identity is central to the Codex's framing and is addressed directly by Identity Decoupling in the Foundation. For the motivational roots of bias, Kunda's "The Case for Motivated Reasoning" is the key paper.
The rationalist community, particularly through the work of Eliezer Yudkowsky and the Less Wrong community beginning in 2006, translated the research into practical frameworks for counteracting the bias in real time. Philip Tetlock's forecasting research (Superforecasting, 2015) provided the empirical demonstration that people who actively counteract confirmation bias through deliberate updating produce measurably better predictions than those who do not.
There is a counterpoint: some researchers argue that confirmation bias has adaptive value in specific contexts. In environments where rapid commitment to a hypothesis is more valuable than exhaustive testing, a confirmatory strategy can be efficient. The problem is not that confirmation bias exists. The problem is that it operates in contexts where it is catastrophically maladaptive: in complex, high-stakes, long-time-horizon decisions where getting the answer right matters more than getting an answer quickly. These are precisely the contexts the Codex addresses.