Predictive processing from cognitivism to autopoiesis
Active inference under the free-energy principle is enactive and embodied, unlike modular comparator models.
Source
From cognitivism to autopoiesis: towards a computational framework for the embodied mind
What they did
They sort predictive-processing theories by commitment to embodiment and argue that Active Inference, as entailed by the Free Energy Principle, is not a cognitivist comparator scheme but an autopoietic, empirically usable synthesis of internalist and externalist views.
What they found
Comparator models remain modular cognitivism; radical predictive processing is connectionist; the FEP treats nested Markov blankets as constitutive coupling of brain, body, and niche rather than a sceptical wall hiding the world.
The limits
What it doesn't show
It does not identify which neural systems encode domain-specific uncertainty, nor settle whether distributed 'representations' still count as representations.
Key terms
- Predictive processing
- Family of theories that treat error-minimising top-down predictions as the core of cognition.
- Comparator model
- Cognitivist scheme in which encapsulated modules compare predicted and actual sensory states.
- Radical predictive processing
- Connectionist view that predictions and errors cascade across a global neuronal hierarchy.
- Free Energy Principle
- Normative claim that organisms persist by minimising surprise (variational free energy).
- Active inference
- Enactive process theory in which action as well as perception reduces expected free energy.
- Markov blanket
- Statistical partition into internal, sensory, active, and external states that individuates a system.
Flashcards
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
Comparator models of prediction are, philosophically,
Common questions
Are all predictive-processing theories embodied?
No. Comparator models are modular cognitivism; only some variants, especially FEP/active inference, are argued to be enactive and embodied.
What is radical predictive processing?
A connectionist picture in which a canonical microcircuit motif, not a localised comparator, carries predictions and prediction errors across the hierarchy.
Does a Markov blanket hide the agent from the world?
The authors say no: blankets are nested and glue brain to body to econiche; they do not license radical scepticism.
What does the FEP give enactivists?
A formal guideline to discovery for building testable models of brain–body–world coupling, not a finished process theory.
More on Embodied mind