The question of how we come to hold beliefs — and why we defend them against evidence — sits at the intersection of epistemology, cognitive science, and evolutionary psychology. This essay attempts a synthesis.
The Foundational Problem
Every belief system must eventually rest on something that isn’t itself justified by another belief. This is the regress problem: if belief B is justified by belief A, what justifies A? You have three options: infinite regress (A is justified by A₁, which is justified by A₂, and so on), circularity (A is eventually justified by B), or foundationalism (some beliefs are basic — they require no further justification).
The scandal of philosophy is not that so many disagreements remain unresolved, but that the tools for resolving them are themselves disputed.
Descartes famously attempted to ground his entire epistemology in a single indubitable foundation: cogito ergo sum. The act of doubting proves a doubter exists. From there, he attempted to reconstruct the external world through careful deduction. The project largely failed on its own terms — but it established a framework that dominated Western epistemology for three centuries.
Coherentism as Alternative
Coherentism abandons the search for bedrock. Instead, beliefs are justified by their coherence with other beliefs. A belief is rational to the extent that it fits into a mutually supporting web of other beliefs. No single belief is privileged; the system hangs together holistically.
The practical implication is significant: coherentism predicts that deeply held beliefs will be remarkably resistant to isolated counter-evidence. A single disconfirming fact rarely overturns a coherent belief system — instead, it gets explained away, reinterpreted, or assigned low reliability. The system protects itself.
Predictive Processing and Bayesian Brains
Contemporary neuroscience has converged on a model that maps surprisingly well onto coherentism: the predictive processing framework. On this view, the brain is fundamentally a prediction machine. Rather than passively receiving sensory data, the brain continuously generates predictions about incoming information and updates based on prediction error.
Posterior = Prior × Likelihood / Evidence
P(hypothesis | data) ∝ P(data | hypothesis) × P(hypothesis)The subjective experience of perception is, on this account, largely a product of prior expectations — not raw sensory data. This explains a host of perceptual phenomena: why optical illusions persist even after we “know” what they are, why expectation shapes taste, why pain can be modulated by placebo.
Implications for Intellectual Honesty
If coherentism is correct — and if the brain is a prediction machine that actively resists updating — then intellectual honesty is not a natural state. It is a discipline. It requires building deliberate practices that counteract the system’s natural conservatism:
- Actively seek disconfirming evidence rather than confirming evidence
- Track prediction track records (write down predictions before outcomes)
- Distinguish between “this is consistent with my model” and “this confirms my model”
- Budget time for seriously engaging with the strongest version of opposing views
Next in this series: Language and Belief — can you hold a belief for which you lack the vocabulary?