Definition
Predictive Processing (PP) is a unifying framework in contemporary cognitive science, principally associated with Karl Friston (free energy principle) and Andy Clark (Surfing Uncertainty, 2016). Its core claim: the brain is not a passive stimulus-response machine but a hierarchical generative model that continuously produces top-down predictions about its sensory inputs and updates those predictions in proportion to precision-weighted prediction errors propagating bottom-up.
Three components are central:
- Priors — probability distributions encoding the system's current best guess about hidden causes of sensory states. Priors are themselves the output of higher-level priors, all the way up.
- Prediction error — the discrepancy between predicted and actual sensory signal at each hierarchical level.
- Precision — the inverse-variance weighting that determines how much influence a given prediction error has on belief update. Precision is itself dynamically estimated and is the cognitive-science correlate of attention.
Under PP, perception, action, attention, and learning are unified: all are forms of prediction-error minimization. Friston's Free Energy Principle generalizes this: any self-organizing system that maintains its boundary against entropy must, in effect, minimize a quantity equivalent to variational free energy — equivalent to surprise, equivalent to prediction error.
Context
Predictive processing supplies the mechanistic substrate that makes epistemic annihilation legible at the level of neural computation. The Kompendium treats PP as the bridge between philosophical / phenomenological description and operational, testable mechanism:
- EA-1.1.x cognitive-input techniques (selective exposure, information overload, source poisoning) operate by starving or saturating the prediction-error signal — denying the priors the corrective evidence they need to update.
- EA-1.3.x immunization techniques function by artificially raising prior precision so that no incoming evidence can shift them. The victim's belief becomes Bayesian-immune by design.
- Gaslighting (a paradigmatic technique catalogued throughout the kompendium) is precisely adversarial precision manipulation of the victim's interoceptive and self-monitoring channels — the victim cannot trust their own perceptions because the precision-weighting of self-generated signal has been collapsed.
- The NABLA Infinity epistemic pipeline implements PP-aware defenses: explicit prior tracking, precision auditing, and detection of pathologically immune belief states. See /glossary/epistemic-pipeline/ for the 16-level architecture.
- The kompendium at /kompendium/07-teoreticke-reflexe/ develops the PP framing as one of three theoretical pillars (alongside phenomenology and critical theory) for understanding why EA is so devastating: it does not merely insert false beliefs, it rewires the inference machinery itself.
- The Decision Engine calibration loop is itself a precision-management system — Brier scores and ECE measure how well the platform's priors track reality, providing PP-style accountability.
See Also
- Epistemická anihilace — Phenomenon explained
- Kognitivní anihilace — Cognitive-process counterpart
- Metakognitivní destrukce — Reflective-precision damage
- NABLA Infinity — Epistemic framework
- Epistemic Pipeline — 16-level reasoning architecture
- Kompendium — Full taxonomy
- Teoretické reflexe — Theoretical foundations
Connect & Contribute
Created by Tomáš Korcak (korczis) | Open Source under GHL