Predictive Coding

AKA

Predictive Learning
Predictive Processing

Focus

Anticipating upcoming sensations based on immediate sensations

Principal Metaphors

  • Knowledge is … range of possible anticipations
  • Knowing is … appropriate anticipations
  • Learner is … a sensing agent (organism or a digital technology)
  • Learning is … updating one’s frame of anticipation
  • Teaching is … N/A

Originated

1980s

Synopsis

Predictive Coding asserts that cognitive systems’ movements through the world are based more on educated expectation (top-down, conceptual knowledge) than immediate perception (bottom-up, sensory information). Learning becomes necessary when a Prediction Error occurs:
  • Prediction Error – a mismatch between expected and actual outcomes during perception, learning, or decision-making. A significant Prediction Error should prompt one to update expectations and/or adjust behavior.
The same types of processes are used for different neural networks, including perception, understanding, dreaming, memory and imagination. The fields of computer science and Cognitive Science both utilize Predictive Coding to generate models of cognition that underlie Machine Learning and neural nets. Subdiscourses and important constructs include:
  • Active Inference (Karl Friston, 2000s) – an extension of Predictive Coding emphasizing that organisms minimize prediction error not only by updating beliefs but also by acting on the world. Perception and action are inseparable: organisms move, explore, and intervene to bring sensory input into line with expected states.
  • Bayesian Brain Theory (Geoffrey Hinton, 2000s) – a framework that proposes the brain is constantly generating and updating probabilistic models (or “beliefs” about the world based on prior knowledge and new sensory evidence. (The name comes from Bayes’ Theorem, a  method for updating the probability of a hypothesis in light of new evidence.)
  • Inference – the cognitive system’s “best guess” at what is being perceived – by drawing on past experience to fill in perception gaps

Commentary

Proponents of Predictive Coding make close ties to Enactivism and its conception of the cognitive system in action-oriented terms. However, critiques claim that predictions are based on a metaphor of “stored knowledge,” which is contradictory to the premises of Enactivism. In the same vein, Predictive Coding falls among Postcognitivist Discourses, which problematize the use of computer metaphors to describe cognition.

Authors and/or Prominent Influences

Andy Clark

Status as a Theory of Learning

Predictive Coding provides insights into the complexities of learning.

Status as a Theory of Teaching

Predictive Coding does not attempt to offer any advice for teaching or education.

Status as a Scientific Theory

Predictive Coding is a well-established discourse in Neuroscience and Cognitive Science.

Subdiscourses:

  • Active Inference
  • Bayesian Brain Theory
  • Inference
  • Prediction Error

Map Location



Please cite this article as:
Davis, B., & Francis, K. (2026). “Predictive Coding” in Discourses on Learning in Education. https://learningdiscourses.com.


⇦ Back to Map
⇦ Back to List