When demarcating the boundaries of the mind the demarcation conditions should meet the following criteria:
(1) naturalistically motivated,
(2) non-question begging, and
(3) carve nature at the joints.
Markov blankets have been proposed as providing the conditions for the demarcation of the boundary of mind and world (Clark, Reference Clark2016; Hohwy, Reference Hohwy2016; Kirchhoff & Kiverstein, Reference Kirchhoff and Kiverstein2021; Kirchhoff, Parr, Palacios, Friston, & Kiverstein, Reference Kirchhoff, Parr, Palacios, Friston and Kiverstein2018; Ramstead, Kirchhoff, Constant, & Friston, Reference Ramstead, Kirchhoff, Constant and Friston2021). We argue that Bruineberg and colleagues' distinction between Friston and Pearl blankets exposes why Markov blankets cannot meet the criteria required to demarcate the boundaries of mind.
Pearl blankets are naturalistically motivated and non-question begging but do not carve nature at the joints. This is because Pearl blankets are based upon practices in a range of scientific fields which do not have the purpose of carving nature at its joints, but are methodologically salient.
Friston blankets are not naturalistically motivated. There is an “explanatory leap,” as Bruineberg and colleagues put it, whereby the scientific practices and mathematical rigour of Pearl blankets are erroneously transposed onto Friston blankets by free energy principle (FEP) proponents despite them being disanalogous. Principled reasons and evidence need to be presented to go beyond Pearl blankets to metaphysical claims about reality. For example, in a survey article by Mann, Pain, and Kirchhoff (Reference Mann, Pain and Kirchhoffforthcoming), they effortlessly move between Pearl's restricted sense of a Markov blanket as the total information provided about a node in a Bayesian network by all of the nodes to which it is locally connected, to the “special usage” of Markov blanket as a Friston blanket: A set of nodes that screen off an agent from a set of nodes that are external to the agent. They do so without any explanation of how we get from Pearl's conception to Friston's, this is a very clear example of the explanatory leap that Bruineberg and colleagues have identified.
Furthermore, it is debatable whether the Friston blanket is carving nature at the joints, and it is question begging insofar that it modifies the debate rather than tackles it – transforming the debate from one about the extended mind (Clark & Chalmers, Reference Clark and Chalmers1998) to one about the applicability of mathematics (Bangu, Reference Bangu2012). Friston blankets carry radical metaphysical baggage that is, at present, unjustified and unacknowledged by FEP proponents. For instance, Ramstead et al. (Reference Ramstead, Kirchhoff, Constant and Friston2021) posit a formal ontology in which “traditional” metaphysical enquiry is replaced by mathematics. Rather than eliminating metaphysics, this is a metaphysical view – with a long lineage in Western philosophy. As we have argued previously (Menary & Gillett, Reference Menary, Gillett, Mendonça, Curado and Gouveia2021), the inherent Platonism and Pythagoreanism here is controversial. There are a number of issues that must be tackled in order to make such a view legitimate. For example, if one contends that nature is mathematical, then there are a range of concerns about the relationship between mathematical and physical structures, and whether one is claiming that mathematical entities have causal powers. If Friston blankets are to function as demarcating the actual boundaries of the mind, then the inference is that they have causal powers. But, as mentioned above, FEP proponents often move between methodological and ontological usages of the concept. One cannot simply forgo philosophical argumentation by “doing the math,” as nicely noted by Bruineberg and colleagues. Instead, serious principled reasons need to be given as to why this metaphysical picture is justified and warranted. To be clear, our claim is not that a Friston blanket style-approach to demarcating the boundaries of the mind is untenable, rather it is currently undefended.
Pearl blankets are insufficient
One option is to rely on Pearl blankets as a heuristic for demarcation. Some FEP proponents (e.g., Ramstead, Friston, and Hipólito, Reference Ramstead, Friston and Hipólito2020a; Ramstead, Kirchhoff, & Friston, Reference Ramstead, Kirchhoff and Friston2020b) have dabbled with instrumentalism. However, even a deflationary position does not help. Pearl blankets by themselves are insufficient to adjudicate between internalist (e.g., Hohwy, Reference Hohwy2013) and extended (e.g., Clark, Reference Clark2016) accounts of the boundaries of cognition that draw upon the Markov blanket conception. The issue is that a Pearl blanket – as the evidentiary boundary beyond which are hidden variables – can be articulated as a shifting boundary based on an action-orientated engagement with the world (a la Clark) or as a fixed boundary of the skull-bound brain (a la Hohwy). Not only is a Pearl blanket consistent with these opposing positions but it is also insufficient by itself to differentiate between them.
Friston blankets are insufficient
To clearly demarcate the boundaries of the mind one must determine the ontological conditions under which such boundaries are drawn, and this is not a matter of heuristics or instrumental thinking; it is a matter of carving nature at its joints. We note that this scientific realist construal of Friston blankets allows that the scientific models can be approximately true, partial, and probabilistic. However, they nevertheless make concrete ontological claims about the existence of causal phenomena in the world – that is, Friston blankets are real entities. As such, we are forced to return to the challenges discussed above and Friston blankets fail the first and second criteria. We would like those utilising Friston blankets to demarcate the boundaries of mind to be more careful in their shift between Pearl blankets and Friston blankets and to begin to give an account of how these entities function in nature in terms of their causal powers.
In summary, the conundrum for FEP proponents is that both conceptions fail to fully meet the criteria for demarcating the boundaries of the mind. If we adopt a Pearl blanket conception, then we are incapable of differentiating between internalist and extended positions because a methodological heuristic cannot decide an ontological matter. Therefore, we must turn to the more ontologically robust Friston blanket conception. This move involves an explanatory leap that is not currently justified. This either retreats the position back to the unworkable but justified Pearl blanket conception, or requires a commitment to the causal powers of Friston blankets – but this only replaces one set of problems with another (from the extended mind to the applicability of mathematics) rather than actually helping to tackle the original problem.
When demarcating the boundaries of the mind the demarcation conditions should meet the following criteria:
(1) naturalistically motivated,
(2) non-question begging, and
(3) carve nature at the joints.
Markov blankets have been proposed as providing the conditions for the demarcation of the boundary of mind and world (Clark, Reference Clark2016; Hohwy, Reference Hohwy2016; Kirchhoff & Kiverstein, Reference Kirchhoff and Kiverstein2021; Kirchhoff, Parr, Palacios, Friston, & Kiverstein, Reference Kirchhoff, Parr, Palacios, Friston and Kiverstein2018; Ramstead, Kirchhoff, Constant, & Friston, Reference Ramstead, Kirchhoff, Constant and Friston2021). We argue that Bruineberg and colleagues' distinction between Friston and Pearl blankets exposes why Markov blankets cannot meet the criteria required to demarcate the boundaries of mind.
Pearl blankets are naturalistically motivated and non-question begging but do not carve nature at the joints. This is because Pearl blankets are based upon practices in a range of scientific fields which do not have the purpose of carving nature at its joints, but are methodologically salient.
Friston blankets are not naturalistically motivated. There is an “explanatory leap,” as Bruineberg and colleagues put it, whereby the scientific practices and mathematical rigour of Pearl blankets are erroneously transposed onto Friston blankets by free energy principle (FEP) proponents despite them being disanalogous. Principled reasons and evidence need to be presented to go beyond Pearl blankets to metaphysical claims about reality. For example, in a survey article by Mann, Pain, and Kirchhoff (Reference Mann, Pain and Kirchhoffforthcoming), they effortlessly move between Pearl's restricted sense of a Markov blanket as the total information provided about a node in a Bayesian network by all of the nodes to which it is locally connected, to the “special usage” of Markov blanket as a Friston blanket: A set of nodes that screen off an agent from a set of nodes that are external to the agent. They do so without any explanation of how we get from Pearl's conception to Friston's, this is a very clear example of the explanatory leap that Bruineberg and colleagues have identified.
Furthermore, it is debatable whether the Friston blanket is carving nature at the joints, and it is question begging insofar that it modifies the debate rather than tackles it – transforming the debate from one about the extended mind (Clark & Chalmers, Reference Clark and Chalmers1998) to one about the applicability of mathematics (Bangu, Reference Bangu2012). Friston blankets carry radical metaphysical baggage that is, at present, unjustified and unacknowledged by FEP proponents. For instance, Ramstead et al. (Reference Ramstead, Kirchhoff, Constant and Friston2021) posit a formal ontology in which “traditional” metaphysical enquiry is replaced by mathematics. Rather than eliminating metaphysics, this is a metaphysical view – with a long lineage in Western philosophy. As we have argued previously (Menary & Gillett, Reference Menary, Gillett, Mendonça, Curado and Gouveia2021), the inherent Platonism and Pythagoreanism here is controversial. There are a number of issues that must be tackled in order to make such a view legitimate. For example, if one contends that nature is mathematical, then there are a range of concerns about the relationship between mathematical and physical structures, and whether one is claiming that mathematical entities have causal powers. If Friston blankets are to function as demarcating the actual boundaries of the mind, then the inference is that they have causal powers. But, as mentioned above, FEP proponents often move between methodological and ontological usages of the concept. One cannot simply forgo philosophical argumentation by “doing the math,” as nicely noted by Bruineberg and colleagues. Instead, serious principled reasons need to be given as to why this metaphysical picture is justified and warranted. To be clear, our claim is not that a Friston blanket style-approach to demarcating the boundaries of the mind is untenable, rather it is currently undefended.
Pearl blankets are insufficient
One option is to rely on Pearl blankets as a heuristic for demarcation. Some FEP proponents (e.g., Ramstead, Friston, and Hipólito, Reference Ramstead, Friston and Hipólito2020a; Ramstead, Kirchhoff, & Friston, Reference Ramstead, Kirchhoff and Friston2020b) have dabbled with instrumentalism. However, even a deflationary position does not help. Pearl blankets by themselves are insufficient to adjudicate between internalist (e.g., Hohwy, Reference Hohwy2013) and extended (e.g., Clark, Reference Clark2016) accounts of the boundaries of cognition that draw upon the Markov blanket conception. The issue is that a Pearl blanket – as the evidentiary boundary beyond which are hidden variables – can be articulated as a shifting boundary based on an action-orientated engagement with the world (a la Clark) or as a fixed boundary of the skull-bound brain (a la Hohwy). Not only is a Pearl blanket consistent with these opposing positions but it is also insufficient by itself to differentiate between them.
Friston blankets are insufficient
To clearly demarcate the boundaries of the mind one must determine the ontological conditions under which such boundaries are drawn, and this is not a matter of heuristics or instrumental thinking; it is a matter of carving nature at its joints. We note that this scientific realist construal of Friston blankets allows that the scientific models can be approximately true, partial, and probabilistic. However, they nevertheless make concrete ontological claims about the existence of causal phenomena in the world – that is, Friston blankets are real entities. As such, we are forced to return to the challenges discussed above and Friston blankets fail the first and second criteria. We would like those utilising Friston blankets to demarcate the boundaries of mind to be more careful in their shift between Pearl blankets and Friston blankets and to begin to give an account of how these entities function in nature in terms of their causal powers.
In summary, the conundrum for FEP proponents is that both conceptions fail to fully meet the criteria for demarcating the boundaries of the mind. If we adopt a Pearl blanket conception, then we are incapable of differentiating between internalist and extended positions because a methodological heuristic cannot decide an ontological matter. Therefore, we must turn to the more ontologically robust Friston blanket conception. This move involves an explanatory leap that is not currently justified. This either retreats the position back to the unworkable but justified Pearl blanket conception, or requires a commitment to the causal powers of Friston blankets – but this only replaces one set of problems with another (from the extended mind to the applicability of mathematics) rather than actually helping to tackle the original problem.
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Conflict of interest
None.