Applying a rational analysis framework of cognition, Binz et al. resolutely embrace the escalating use of meta-learning to re-construct Bayes optimal-learning algorithms and solutions to explain the metaphysical relationship of mind to environment. Such logicomathematical descriptions of cognition non-trivially approximate properties of mind, including agent-based decision processes, to that of ideal statistical inference and the structure of natural tasks and environments. The authors nonetheless fail to satisfactorily introduce Markov blankets, which would expand their cognitive construct beyond standard applications of analytical and numerical tools to demark relations represented in directed Bayesian networks or graphs and variational probabilistic inference. Markov or Pearl blankets, coined by Bruineberg, Dolega, Dewhurst, and Baltieri (Reference Bruineberg, Dolega, Dewhurst and Baltieri2021) to acknowledge the original decades-old epistemic concept developed by Pearl (Reference Pearl1988), earned deserving attention for their utility in decision sciences, offering tractable optimization methods to identify, partition, and understand groups of marginally and conditionally (in)dependent variables associated with complex systems and causal predictions and attributions. In more recent years, however, a trend started by Friston and coworkers (e.g., Hipόlito et al., Reference Hipolito, Ramstead, Convertino, Bhat, Friston and Parr2021; Friston et al., Reference Friston, Parr, Hipolito, Magrou and Razi2021; Ramstead, Badcock, & Friston, Reference Ramstead, Badcock and Friston2018) promotes extension of Markov blankets within a free-energy, active-inference framework to describe the physical interface and reciprocal interactions between agents and their environments. These so-called Friston blankets, a term also coined by Bruineberg et al. (Reference Bruineberg, Dolega, Dewhurst and Baltieri2021) to differentiate their usage from Pearl blankets, cleverly articulate embedded philosophical axioms of mind, body, and environment without fully capturing the logicomathematical rigor and cogency found in typical uses of Pearl blankets and Bayesian inference. Taking a reasonably conventional position on the state-of-art of Friston blankets, theorists supportive of free-energy models must either accept Markov blankets as formal technical innovations that enable practical worthwhile science in absence of known compelling philosophical conclusions about the nature of cognition and life or as dubious mathematics-driven metaphysics interpretations of reality with promising high-impact implications for elucidating cognition and life upon separate experimental biophysical validation. The merits of such a statement seemly convey a fair, albeit critical, edict to the scientific community – one that perhaps discouraged Binz et al. from clarifying Markov blankets and leaves the impression that good productive science enlisting Markov blankets as instruments for inference returns only theoretically mundane findings about cognition and, possibly, life. But, that is not the case and the authors' well-structured arguments may fool readers into believing that this position is true since Binz et al. disappointingly content themselves with only discussing classical Bayesian inference and non-paradoxical cognition, maybe because Friston and coworkers also never stray beyond a classical formulation of their free-energy principle for active inference.

Weaknesses in Binz et al.'s narrow perspective on meta-leaning and cognition may be contrasted and reshaped by exciting findings produced with quantum decision theory, a strict valid quantum-statistical approach capable of defining probabilistic human inference unconfined by the physical mechanical world (Aerts & Aerts, Reference Aerts and Aerts1995; Aerts, Broakaert, Gabara, & Sozzo, Reference Aerts, Broakaert, Gabara and Sozzo2016; Ashtiani & Azgomi, Reference Ashtiani and Azgomi2015; Busemeyer & Bruza, Reference Busemeyer and Bruza2011; Busemeyer, Wang, & Lambert-Mogiliansky, Reference Busemeyer, Wang and Lambert-Mogiliansky2009; Clark, Reference Clark and Salander2011, Reference Clark and Floares2012, Reference Clark2014b, Reference Clark2015, Reference Clark2017; Pinto Moreira, Fell, Dehdashti, Bruza, & Wichert, Reference Pinto Moreira, Fell, Dehdashti, Bruza and Wichert2020; Pothos & Busemeyer, Reference Pothos and Busemeyer2013). Quantum decision theory, with the aid of quantum networks or graphs, quantum Bayesian inference, and other computational features, demonstrates robust successes in modeling and simulating difficult-to-render cognitive phenomena overlooked by Binz et al., especially causal judgment errors or paradoxes unaccounted for by classical decision theory, including conjunctive and disjunctive fallacies, the Allais paradox, and the Ellsberg or planning paradox (Atmanspacher & Römer, Reference Atmanspacher and Römer2012; Busemeyer, Pothos, Franco, & Trueblood, Reference Busemeyer, Pothos, Franco and Trueblood2011; Clark, Reference Clark2021b; Favre, Wittwer, Heinimann, Yukalov, & Sornette, Reference Favre, Wittwer, Heinimann, Yukalov and Sornette2016; Moreira & Wichert, Reference Moreira and Wichert2016a, Reference Moreira and Wichert2016b, Reference Moreira and Wichert2018b; Pothos & Busemeyer, Reference Pothos and Busemeyer2009, Reference Pothos and Busemeyer2013). The great explanatory power of quantum decision networks (Bianconi, Reference Bianconi2002a, Reference Bianconi2002b, Reference Bianconi2003; Bianconi & Barabási, Reference Bianconi and Barabási2001; Li, Iqbal, Perc, Chen, & Abbott, Reference Li, Iqbal, Perc, Chen and Abbott2013) permits expression of quantum Markov chains and blankets (Brandao, Piani, & Horodecki, Reference Brandao, Piani and Horodecki2015; Moreira & Wichert, Reference Moreira and Wichert2018a; Qi & Ranard, Reference Qi and Ranard2021; Sutter, Reference Sutter2018; Wichert, Pinto Moreira, & Bruza, Reference Wichert, Pinto Moreira and Bruza2020), which bound suboptimal free-energy classical inferential paradoxes from optimal free-energy quantum inferential solutions. In this context of meta-learned cognitive inference, both quantum Bayesian inference and Markov blankets provide epistemic tools, analogous to variational Bayesian inference and Pearl blankets, to legitimately approximate irrational human decision making and cognition within a cognitive-completeness constraint. That constraint, not considered by Binz et al., strongly limits degrees of freedom for emergence of meta-learned subjective physical reality (cf. Blume-Kohout & Zurek, Reference Blume-Kohout and Zurek2006; Brandao et al., Reference Brandao, Piani and Horodecki2015; Clark, Reference Clark2014a, Reference Clark2017, Reference Clark2019, Reference Clark2020, Reference Clark2021b, Reference Clark2023; Yearsley & Pothos, Reference Yearsley and Pothos2014), a scenario which helps affirm the idea that Markov blankets may yield meaningful philosophical conclusions regarding the nature of cognition and life.

Cognitive completeness (Tressoldi, Maier, Buechner, & Khrennikov, Reference Tressoldi, Maier, Buechner and Khrennikov2015; Yearsley & Pothos, Reference Yearsley and Pothos2014) encapsulates a black-box approach that isolates any studied cognitive system from the formidable environment-significant measurement problem of quantum mechanics. Some scientists insist the scalable neurophysiological contents of this black box map onto classical or quantum cognitive states relevant to particular sets of respective rational or irrational decisions and their corresponding outcome probabilities (Clark, Reference Clark2017, Reference Clark2021a; Wang & Busemeyer, Reference Wang and Busemeyer2015; Yearsley & Pothos, Reference Yearsley and Pothos2014). For example, if one abandons the constraint of cognitive realism – the assertion that all (meta-learned) cognitive events emerge from classical deterministic neurophysiology – then cognitive systems and their decisional outcomes may be completely described by dimensions or sets of similarity classes on the set of all probability distributions over deterministic and indeterminate (or stochastic) neuropsychological variables and associated environmental settings. Further, arguably more fundamental cognitive completeness and Markov blanket partitions imply non-disturbing measurements should be non-invasive on brain physiology and cognition, with disturbing measurements affecting outcomes of hidden neurophysiological and cognitive variables in a quantum-sensitive manner. Disturbing measurements violate temporal Bell or Leggett–Garg inequalities, signifying violations of classical (meta-learned) cognition, such as Markov-blanketed inferential paradoxes at suboptimal free energy. Restructuring these cognitive phenomena within quantum decision theory and quantum Markov blankets creates opportunities for irrational decision making to be organized into explainable manifolds at optimal free energy, tantamount to quantum active Bayesian inference where observable incompatibility or inequality violations are acceptable (Clark, Reference Clark2021a). Such theoretical elegance in describing complex cognition forces classical and quantum aspects of brain structure and function into a newer realm of logicomathematical formalism with verifiable, important real-world consequences.