From cognitivism to autopoiesis: towards a computational framework for the embodied mind

Predictive processing approaches share a common ground in situating top-down, error-minimising predictions as the key locus of information processing, as opposed to classical accounts emphasizing feed-forward feature recognition.

“Recent developments in cognitive science and neuroscience have led to a growth of interest in “predictive processing” theories of mind and cognitive function (Clark 2013; Friston 2010; Hohwy 2013). These diverse approaches share a common ground in situating top-down, error-minimising predictions as the key locus of information processing, as opposed to more classical accounts which emphasize feed-”

• cites Clark 2013 sourceNot checked yet

  Not checked in the free run.

• cites Friston 2010 sourceNot checked yet

  Not checked in the free run.

• cites Hohwy 2013 sourceNot checked yet

  Not checked in the free run.

Natural selection is nature's way of performing Bayesian model selection, offering phenotypes to the environment as plausible hypotheses for explaining sensory exchange with an econiche and selecting those that survive as the hypotheses with the most evidence, i.e. least free energy.

“In short, natural selection is nature’s way of performing Bayesian model selection, offering phenotypes to the environment as plausible hypotheses for explaining sensory exchange with an econiche—and selecting those that survive as the hypotheses with the most evidence (i.e., least free energy).”

Comparator models are agnostic regarding the specific mathematical computation underlying their function, which can be explained by a variety of Bayesian or non-Bayesian mechanisms.

“Crucially, comparator models are agnostic regarding the specific mathematical computation underlying their function, which can be explained by a variety of Bayesian (e.g., variational Bayes, Kalman filtering) or non-Bayesian mechanisms (e.g., reinforcement learning) (Friston 2011; Peters and Schaal 2008; Wolpert et al. 1995).”

• cites Friston 2011Not checked yet

  Not checked in the free run.

• cites Peters and Schaal 2008 sourceNot checked yet

  Not checked in the free run.

• cites Wolpert et al. 1995 sourceNot checked yet

  Not checked in the free run.

Radical predictive processing posits that a localized comparator region of the brain does not necessarily carry out the comparison of expected and actual sensory information for a given domain.

“In either case RPP posits that a localized comparator region of the brain does not (necessarily) carry out the comparison of expected and actual sensory information for a given domain.”

• cites Clark 2013 sourceNot checked yet

  Not checked in the free run.

• cites Clark 2015Not checked yet

  Not checked in the free run.

The nervous system is itself selected by evolution to minimize prediction error within a particular ecological niche.

“This ultimately suggests that the nervous system is itself selected by evolution to minimize prediction error within a particular ecological niche, and may more speculatively support a role for epigenetic mechanisms in propagating cortical function from one generation to the next.”

Comparator models are ultimately functionalist in nature; particular behaviours are explained by appeal to a revision of modality-specific prediction errors, as computed by encapsulated, physically localized brain modules, without specifying a particular computational implementation.

“Such models are thus ultimately functionalist in nature; particular behaviours are explained by appeal to a revision of modality-specific prediction errors, as computed by encapsulated, physically localized brain modules, without specifying a particular computational implementation (Fodor 1983).”

• cites Fodor 1983 sourceNot checked yet

  Not checked in the free run.

Active Inference, as entailed by the Free Energy Principle, is fundamentally enactive and embodied.

“Furthermore, we argue that Active Inference, as entailed by the Free Energy Principle (FEP), is not only fundamentally enactive and embodied, but further offers a synthetic, empirically productive resolution to long-standing disagreements between internalist and externalist viewpoints.”

• AssumptionRests on an assumption: The Free Energy Principle entails Active Inference.

Perception and action are fundamentally affective and embodied in nature, possessing a salience or inherent value for the organism in homeostatic terms.

“Instead of the brain being solely defined by veridical representation, perception and action are now fundamentally affective and embodied in nature, possessing a salience (epistemic) or inherent (pragmatic) value for the organism in homeostatic terms (see also Gallagher and Allen 2016).”

• cites Gallagher and Allen 2016Not checked yet

  Not checked in the free run.

The brain is in the game of predicting the world, but only as a means to the end of embodied self-preservation.

“The brain is in the game of predicting the world, but only as a means to the end of embodied self-preservation.”

Radical predictive processing contrasts in a substantive way with comparator-based formulations in that the entire neuronal architecture within the brain becomes one forward or generative model with a deep hierarchical structure.

“It does however contrast in a substantive way with comparator-based formulations in the sense that the entire neuronal architecture within the brain becomes one forward or generative model with a deep hierarchical structure.”

According to radical predictive processing, prediction error unfolds not only at ontogenetic but also phylogenetic timescales.

“Thus, according to RPP, prediction error unfolds not only at ontogenetic but also phylogenetic timescales; if the brain (and body) constitute a generative model, than those embodied graphs best suited to their environmental niche will be selected by evolution.”

Radical predictive processing constitutes a strong form of connectionism, in which the overall dynamics of the nervous system accomplish information processing rather than compartmentalized modules.

“RPP constitutes a strong form of connectionism, in which it is the overall dynamics of the nervous system8 that accomplish information processing rather than compartmentalized modules.”

Interoceptive and embodied predictive coding models go beyond extending predictive coding to explain body-awareness to argue that body-related predictions coordinate and contextualize global brain function.

“In summary, interoceptive and embodied predictive coding models not only extend predictive coding to explain body-awareness; rather, they go beyond this weaker claim to argue that body-related predictions coordinate and contextualize global brain function.”

• cites Barrett and Simmons 2015 sourceNot checked yet

  Not checked in the free run.

• cites Seth 2013 sourceNot checked yet

  Not checked in the free run.

• cites Seth et al. 2012Not checked yet

  Not checked in the free run.

Recent evidence indicates that unexpected, unconscious surges of interoceptive arousal reverse the impact of sensory noise on perceptual awareness.

“In support of this argument, recent evidence indicates that unexpected, unconscious surges of interoceptive arousal reverse the impact of sensory noise on perceptual awareness (Allen et al. 2016b).”

• cites Allen et al. 2016bNot checked yet

  Not checked in the free run.

The ability to predict future sensory and embodied states is essential for the efficient control of perception and action.

“The ability to predict future (sensory and embodied) states is essential for the efficient control of perception and action.”

• cites Miall and Wolpert 1996 sourceNot checked yet

  Not checked in the free run.

• cites Wolpert and Flanagan 2001 sourceNot checked yet

  Not checked in the free run.

Radical predictive processing does not necessarily commit one against all functional localization or modularity.

“However, even considering the above, RPP does not necessarily commit one against all functional localization or modularity.”

The inferential weighting of sensory representation by interoceptive precision is a clear departure from the strictly cognitivist stance.

“This inferential weighting of sensory representation by interoceptive precision is a clear departure from the strictly cognitivist stance described earlier.”

Radical predictive processing implies that genetically pre-specified connection patterns laid down during foetal development pre-ascribe some functional specification prior to empirical learning.

“This implies that genetically pre-specified connection patterns—laid down during foetal development—pre-ascribe some functional specification prior to (empirical) learning.11”

Sensory and active states, which constitute the Markov blanket, mediate perception and action locked into a perpetual cycle to upper bound the entropy of both.

“In other words, the sensory and active states (that constitute the Markov blanket) mediate perception and action that are locked into a perpetual cycle to upper bound the entropy of both.”

The organism, body-brain-and-world itself constitutes the 'belief' or generative model that it will survive.

“More exactly, the organism, body-brain-and-world itself constitutes the ‘belief’ or generative model that it will survive; in a very concrete sense, the kinds of limbs and morphological shape one has will constrain the probabilities of the kinds of actions one can engage in.”

Common critiques of predictive processing (circularity, genesis of priors, definition of optimality) arise from the problem of where the brain's prior beliefs come from.

“the commonly levied critiques (e.g., circularity, genesis of priors, and the definition of optimality) of RPP arise from a common problem; from what do the brain’s prior beliefs arise?”

The FEP provides an empirical bridge between the computational and enactive views of the mind, cashed out in terms of specific neuronal and embodied dynamics.

“In doing so, the theory provides an empirical bridge between the computational and enactive views of the mind cashed out in terms of specific neuronal and embodied dynamics.”

The free energy principle (FEP) tackles critiques of predictive processing by providing a normative account of why the brain must engage in embodied predictive processing to maintain its own enactive integrity.

“Indeed, the FEP tackles these issues head-on by providing a normative account of why—through active inference—the brain must necessarily engage in embodied predictive processing if it is to maintain its own enactive integrity.”

The organism's internal states constitute probabilistic beliefs about what actions are most likely to provide evidence for the organism's existence (survival).

“One can formulate this in another way; the organism’s internal states constitute probabilistic beliefs about what actions are the most likely to provide evidence for the organism’s existence (survival).”

The FEP resolves the 'just so' circularity of predictive processing by appeal to enactivism.

“It is this ‘just so’ circularity that the FEP seeks to resolve by appeal to enactivism.”

The Markov blanket induces a circular causality because sensory states depend on hidden states that depend on active states, which depend upon internal states.

“The Markov blanket induces a circular causality because sensory states depend on hidden states that depend on active states, which depend upon internal states.”

The Free Energy Principle furnishes empirically productive process theories, such as predictive processing, by which to guide discovery through the formal modelling of the embodied mind.

“The FEP thus furnishes empirically productive process theories (e.g., predictive processing) by which to guide discovery through the formal modelling of the embodied mind.”

Autopoiesis, or self-creation, enables a system to limit the possible states it visits and thus to survive.

“It is this autopoiesis, or self-creation, which enables the system to limit the possible states it visits, and thus to survive (Varela et al. 1974).”

• cites Varela et al. 1974 sourceNot checked yet

  Not checked in the free run.

Any biological organism will possess ergodicity: by placing an upper bound on entropy, it will necessarily occupy some states more often than others, so that the average probability of being in a given state equals the probability of being in that state when observed at random.

“By placing an upper bound on entropy, an organism will necessarily occupy (the neighbourhood of) some states more often than others. This means that the average probability of being in a given state is equal to the probability of the system being in that state when observed at random.”

The Free Energy Principle attempts to dissolve tension between internalist and externalist accounts of cognition by providing a formal synthetic account of how internal representations arise from autopoietic self-organization.

“As a final point, we illustrate how the Free Energy Principle (FEP) attempts to dissolve tension between internalist and externalist accounts of cognition, by providing a formal synthetic account of how internal ‘representations’ arise from autopoietic self-organization.”

• AssumptionRests on an assumption: The Free Energy Principle entails Active Inference.

Because the entropy of the Markov blanket is, by ergodicity, the time average of surprise or negative Bayesian log model evidence, sensory and active states will appear to maximise Bayesian model evidence.

“Because the entropy of the Markov blanket is, by ergodicity, the time average of surprise or negative Bayesian log model evidence, sensory and active states will appear to maximise Bayesian model evidence.”

Reinforcement learning for value-based decision making is tautological because a cost function guides 'optimal' behaviour and cost is defined operationally by whatever an agent chooses.

“For an acute example of this tautology, consider the case of reinforcement learning for values based-decision making, where a cost function guides ‘optimal’ behaviour, and cost is defined operationally by whatever an agent chooses.”

The existence of a system depends on conserving its boundary, known technically as a Markov blanket, so that it remains distinguishable from its environment.

“In short, the very existence of a system depends upon conserving its boundary, known technically as a Markov blanket, so that it remains distinguishable from its environment—into which it would otherwise dissipate.”

Bayesian decision theory is circular because any behaviour can be described as optimal by writing down a prior that prescribes it in a 'just so' fashion.

“Here, we can see the circularity inherent to Bayesian decision theory; any behaviour can be described as optimal, because one can always write down a prior that prescribes any behaviour in a ‘just so’ fashion.”

• cites Bowers and Davis 2012 sourceNot checked yet

  Not checked in the free run.

Living organisms have a necessary tendency to resist the second law of thermodynamics by maintaining an internal structure or dynamics in the face of constant change.

“the FEP highlights the necessary tendency of living organisms to resist the second law of thermodynamics; i.e., to maintain an internal structure or dynamics in the face of constant change.”

• AssumptionRests on an assumption: The second law of thermodynamics applies to living organisms and they must resist it.

The FEP derives a normative, a priori first principle from a provable definition of living systems.

“Rather than the post-hoc definition of priors or cost functions, the FEP derives a normative, a priori first principle from a provable definition of living systems.”

The FEP recasts autopoiesis as a self-fulfilling prophecy in which an organism constitutes, in the generative sense, a belief that it will prevail within certain embodied and environmental conditions.

“The FEP recasts this as a kind of self-fulfilling prophecy, in which an organism itself constitutes, in the generative sense, a belief that it will prevail within certain embodied and environmental conditions.”

A cell persists by creating and maintaining a boundary through which it interacts with the environment, thereby maintaining the integrity of the boundary.

“For example, a cell persists in virtue of its ability to create and maintain a boundary (cell-surface), through which it interacts with the environment, thereby maintaining the integrity of the boundary.”

• cites Varela et al. 1974 sourceNot checked yet

  Not checked in the free run.

My internal representations-the generative model of the world embodied in the web of neural connections-are causally coupled to my homeostatic needs and the environmental niche within which my brain has evolved.

“The implication is that my internal representations—the generative model of the world embodied in the web of neural connections—are causally coupled to my homeostatic needs and the environmental niche within which my brain has evolved.”

The body itself is a prior boundary condition, or a conditioning factor, in the overall generative model defined by my Markov blanket.

“The body itself is thus a prior boundary condition, or a conditioning factor, in the overall generative model defined by my Markov blanket.”

The body-brain system has evolved to constitute a generative model, which specifies the types of behaviours and environments in which I am likely to engage.

“The body-brain system has evolved to constitute a generative model, which specifies the types of behaviours, and environments in which I am likely to engage.”

The free energy bound is not absolute, computed solely in the head, but instead relative to the embodied nature of the organism as selected by evolution by the type of body and environmental niche inhabited by the organism.

“This bound is not absolute, computed solely in the head, but instead relative to the embodied nature of the organism as selected (via evolution) by the type of body and environmental niche inhabited by the organism.”

The kinds of limbs and morphological shape one has constrain the probabilities of the kinds of actions one can engage in.

“in a very concrete sense, the kinds of limbs and morphological shape one has will constrain the probabilities of the kinds of actions one can engage in.”