Life is preserved and enhanced by coordinated selectivity in local neural circuits. Narrow receptive-field selectivity is necessary to avoid the curse-of-dimensionality, but local activities can be made coherent and relevant by guiding learning and processing using broad coordinating contextual gain-controlling interactions. Better understanding of the functions and mechanisms of those interactions is therefore crucial to the issues Clark examines.

The predictive processing framework lacks many of the architectural and implementational details needed to fully investigate or evaluate the ideas it presents. One way to begin to fill in these details is by turning to standard control-theoretic descriptions of these types of systems (e.g., Kalman filters), and by building complex, unified computational models in biologically realistic neural simulations.

Predictive processing models of cognition are promising an elegant way to unite action, perception, and learning. However, in the current formulations, they are species-unspecific and have very little particularly human about them. I propose to examine how, in this framework, humans can be able to massively interact and to build shared worlds that are both material and symbolic.

Clark expresses reservations about Friston's reductive interpretation of action-oriented predictive processing (AOPP) models of cognition, but he doesn't link these reservations to specific alternatives. Neuroeconomic models of sub-cognitive reward valuation, which, like AOPP, integrate attention with action based on prediction error, are such an alternative. They interpret reward valuation as an input to neocortical processing instead of reducing it.

The important roles of prediction and prior experience are well established in music research and fit well with Clark's concept of unified perception, cognition, and action arising from hierarchical, bidirectional predictive processing. However, in order to fully account for human musical intelligence, Clark needs to further consider the powerful and variable role of affect in relation to prediction error.

The Bayesian brain hypothesis provides an attractive unifying framework for perception, cognition, and action. We argue that the framework can also usefully integrate interoception, the sense of the internal physiological condition of the body. Our model of “interoceptive predictive coding” entails a new view of emotion as interoceptive inference and may account for a range of psychiatric disorders of selfhood.

Although predictive coding may offer a computational principle that unifies perception and action, states with different directions of fit are involved (with indicative and imperative contents, respectively). Predictive states are adjusted to fit the world in the course of perception, but in the case of action, the corresponding states act as a fixed target towards which the agent adjusts the world. This well-recognised distinction helps side-step some problems discussed in the target article.

There are aspects of schizophrenia that pose challenges for Clark's model. These include: (1) evidence for excitatory activity underlying self-organizing neural ensembles that support coordinating functions, and their impairment in schizophrenia; (2) evidence regarding hallucinations that suggest they are not due to excessive prediction error; and (3) the critical role of emotional factors as setting conditions for delusion formation.

The approach Clark labels “action-oriented predictive processing” treats all cognition as part of a system of on-line control. This ignores other important aspects of animal, human, and robot intelligence. He contrasts it with an alleged “mainstream” approach that also ignores the depth and variety of AI/Robotic research. I don't think the theory presented is worth taking seriously as a complete model, even if there is much that it explains.

It is often helpful to distinguish between a theory (Marr's computational level) and a specific implementation of that theory (Marr's physical level). However, in the target article, a single implementation of predictive coding is presented as if this were the theory of predictive coding itself. Other implementations of predictive coding have been formulated which can explain additional neurobiological phenomena.

While the target article provides a glowing account for the excitement in the field, we stress that hierarchical predictive learning in the brain requires sparseness of the representation. We also question the relation between Bayesian cognitive processes and hierarchical generative models as discussed by the target article.

The target article sketched and explored a mechanism (action-oriented predictive processing) most plausibly associated with core forms of cortical processing. In assessing the attractions and pitfalls of the proposal we should keep that element distinct from larger, though interlocking, issues concerning the nature of adaptive organization in general.

Classical (Bayesian) probability (CP) theory has led to an influential research tradition for modeling cognitive processes. Cognitive scientists have been trained to work with CP principles for so long that it is hard even to imagine alternative ways to formalize probabilities. However, in physics, quantum probability (QP) theory has been the dominant probabilistic approach for nearly 100 years. Could QP theory provide us with any advantages in cognitive modeling as well? Note first that both CP and QP theory share the fundamental assumption that it is possible to model cognition on the basis of formal, probabilistic principles. But why consider a QP approach? The answers are that (1) there are many well-established empirical findings (e.g., from the influential Tversky, Kahneman research tradition) that are hard to reconcile with CP principles; and (2) these same findings have natural and straightforward explanations with quantum principles. In QP theory, probabilistic assessment is often strongly context- and order-dependent, individual states can be superposition states (that are impossible to associate with specific values), and composite systems can be entangled (they cannot be decomposed into their subsystems). All these characteristics appear perplexing from a classical perspective. However, our thesis is that they provide a more accurate and powerful account of certain cognitive processes. We first introduce QP theory and illustrate its application with psychological examples. We then review empirical findings that motivate the use of quantum theory in cognitive theory, but also discuss ways in which QP and CP theories converge. Finally, we consider the implications of a QP theory approach to cognition for human rationality.

We support the authors' claims, except that we point out that also quantum structure different from quantum probability abundantly plays a role in human cognition. We put forward several elements to illustrate our point, mentioning entanglement, contextuality, interference, and emergence as effects, and states, observables, complex numbers, and Fock space as specific mathematical structures.

One among many misleading quotations about the alleged mysteries of quantum theory is from Feynman (1965): “I think I can safely say that nobody understands quantum mechanics.” Today we know that quantum theory describes many aspects of our world in a fully intelligible fashion. Pothos & Busemeyer (P&B) propose ways in which this may include psychology and cognitive science.

The Hilbert space formalism is a powerful language to express many cognitive phenomena. Here, relevant concepts from signal detection theory are recast in that language, allowing an empirically testable extension of the quantum probability formalism to psychophysical measures, such as detectability and discriminability.

Pothos & Busemeyer (P&B) argue how key concepts of quantum probability, for example, order/context, interference, superposition, and entanglement, can be used in cognitive modeling. Here, we suggest that these concepts can be extended to analyze neurophysiological measurements of cognitive tasks in humans, especially in functional neuroimaging investigations of large-scale brain networks.

I argue that the overly simplistic scenarios discussed by Pothos & Busemeyer (P&B) establish at best that quantum probability theory (QPT) is a logical possibility allowing distinct predictions from classical probability theory (CPT). The article fails, however, to provide convincing evidence for the proposal that QPT offers unique insights regarding cognition and the nature of human rationality.

We propose a way in which Pothos & Busemeyer (P&B) could strengthen their position. Taking a dynamic stance, we consider cognitive tests as functions that transfer a given input state into the state after testing. Under very general conditions, it can be shown that testable properties in cognition form an orthomodular lattice. Gleason's theorem then yields the conceptual necessity of quantum probabilities (QP).

Pothos & Busemeyer's (P&B's) query about whether quantum probability can provide a foundation for the cognitive modeling embodies so many underlying implications that the subject is far from exhausted. In this brief commentary, however, I suggest that the conceptual thresholds of the meaningful learning give rise to a typical Boltzmann's weighting measure, which indicates a statistical verisimilitude of quantum behavior in the human cognitive ensemble.