As Merker et al. note, contemporary consciousness science exhibits some traits of pre-paradigm science, in the Kuhnian sense. For example, proponents of different theories of consciousness still disagree on whether the real explanandum (i.e., the phenomenon to be explained) of consciousness science is phenomenal or access consciousness (Block, Reference Block1995, Reference Block2007), and on whether phenomenal consciousness can be empirically investigated or not. Given that there is no consensus on the methodology for detecting consciousness among all the possible confounds (Overgaard & Fazekas, Reference Overgaard and Fazekas2016; Tsuchiya, Wilke, Frässle, & Lamme, Reference Tsuchiya, Wilke, Frässle and Lamme2015), there is no agreement on how to empirically study phenomenal consciousness as such: The scientific study of consciousness seems to have hit an impasse.
Integrated information theory (IIT) purportedly has the potential to overcome this impasse. This is because IIT's approach starts from phenomenological data, rather than experimentally gathered data. IIT individuates the explanandum of consciousness science via phenomenological observation rather than via cognitive access, and, therefore, it does not require an account of the relationship between consciousness and cognitive accessibility, at least for theory-building purposes. Thus, it looks like the IIT approach, which aims to explain consciousness by starting from consciousness itself, might be able to avoid the problems affecting other research programmes in consciousness science and, consequently, to propose itself as the dominant paradigm in consciousness science.
However, according to Merker et al., IIT cannot do so, because it is not a theory of consciousness. They claim that there is a mismatch between the phenomenological realm individuated by IIT's axioms and the formalism individuated by IIT's postulates, and therefore, the alleged identity between consciousness (i.e., the phenomenon picked out by the axiomatic approach) and Φ (i.e., the informational measure resulting from the postulates) cannot be justified – it is a case of mistaken identity.
The authors focus mainly on the axioms of integration and information, but their analysis seems to rest on two misconstruals of the core notions of IIT. First, they argue that the phenomenological property of unity under a point of view is not represented in the formalism. But IIT's first axiom, that of intrinsic existence, states that to exist intrinsically is to have causal powers upon itself: The notion of unity under a point of view should, therefore, be assessed under the intrinsic aspect of integrated information. In IIT, phenomenal unity is purportedly captured by the irreducible causal powers of a system, but the unity under the system's point of view is purportedly captured by the exertion of the system's powers upon itself. Merker et al.'s analysis, however, does not include this intrinsic aspect of integrated information, and therefore, their criticism seems misconstrued.
Second, Merker et al. claim that IIT conflates the notions of informativeness and that of informational capacity in accounting for the property of phenomenological differentiation. But the information axiom just states that every experience is specific: The specific form of an experience works as phenomenological principium individuationis, whereas a specific cause–effect profile of its physical substrate works as physical principium individuationis of that experience. In IIT, differentiation means “specificity,” but Merker et al.'s analysis of the information postulate does not consider this particular meaning.
Thus, Merker et al.'s thesis that there is a mismatch between axioms and postulates is not convincing, because it is based on inaccurate interpretations of IIT's axiomatic approach. However, I agree with the authors that the translation of axioms into postulates is a crucial issue for IIT: If we want to explain consciousness by starting from phenomenology, we need to make sure that the explanatory identity (Haun & Tononi, Reference Haun and Tononi2019, p. 5) between Φ and consciousness is a unique link, properly extracted from consciousness itself.
In this light, Merker et al.'s point on IIT's mistaken identity can be best reformulated in terms the non-uniqueness problem for IIT (Hanson & Walker, Reference Hanson and Walker2021): If the path going from consciousness to Φ can take different routes, the explanatory validity of Φ as a measure of consciousness can be questioned.
In order for IIT to maintain that Φ uniquely matches consciousness, and not, say, network efficiency, IIT must establish a principled way to justify why other interpretations of Φ are not correct. I suggest that the real problem for IIT is that this aspect of the theory is currently missing: Because IIT does not specify the logical reasoning fuelling the inference of the postulates from the axioms, we do not even have the instruments to tell whether such an inference is valid or not. And, if we cannot tell whether the formalism is validly inferred from phenomenology, then we cannot tell whether there is a unique and proper match between Φ and consciousness.
From a logical viewpoint, the inference from axioms to postulates can be either a deduction or an abduction (Bayne, Reference Bayne2018, p. 2). If it is a deduction, then each axiom must be accompanied by at least a “bridging premise” that connects it to the corresponding postulate. However, I believe abductive reasoning is more in line with how IIT proponents think of the relation between axioms and postulates (Koch, Reference Koch2019, p. 75). In this case, uniqueness amounts to selecting the candidate postulate that, given independent background knowledge and best-maker criteria, best explains the corresponding axiom. But, because we are not given this relevant information, disagreement about the explanatory validity of Φ, and therefore about the potential of IIT as a theory of consciousness, can emerge. This is a shortcoming in IIT, but not of the fatal kind proposed by Merker et al.
Were IIT able to clarify this crucial aspect of its theoretical structure, the path between phenomenology and Φ might prove to be unique. IIT would then indicate a clear explanandum of consciousness science (i.e., phenomenal consciousness as picked out by the axiomatic approach), clear first principles (i.e., axioms and postulates), and a clear (abductive) methodology for mapping the explanandum into the explanans (i.e., the Φ-based formalism). Contrary to what Merker et al. argue, IIT does seem to have the potential – albeit not fully realised yet – to advance consciousness science.
As Merker et al. note, contemporary consciousness science exhibits some traits of pre-paradigm science, in the Kuhnian sense. For example, proponents of different theories of consciousness still disagree on whether the real explanandum (i.e., the phenomenon to be explained) of consciousness science is phenomenal or access consciousness (Block, Reference Block1995, Reference Block2007), and on whether phenomenal consciousness can be empirically investigated or not. Given that there is no consensus on the methodology for detecting consciousness among all the possible confounds (Overgaard & Fazekas, Reference Overgaard and Fazekas2016; Tsuchiya, Wilke, Frässle, & Lamme, Reference Tsuchiya, Wilke, Frässle and Lamme2015), there is no agreement on how to empirically study phenomenal consciousness as such: The scientific study of consciousness seems to have hit an impasse.
Integrated information theory (IIT) purportedly has the potential to overcome this impasse. This is because IIT's approach starts from phenomenological data, rather than experimentally gathered data. IIT individuates the explanandum of consciousness science via phenomenological observation rather than via cognitive access, and, therefore, it does not require an account of the relationship between consciousness and cognitive accessibility, at least for theory-building purposes. Thus, it looks like the IIT approach, which aims to explain consciousness by starting from consciousness itself, might be able to avoid the problems affecting other research programmes in consciousness science and, consequently, to propose itself as the dominant paradigm in consciousness science.
However, according to Merker et al., IIT cannot do so, because it is not a theory of consciousness. They claim that there is a mismatch between the phenomenological realm individuated by IIT's axioms and the formalism individuated by IIT's postulates, and therefore, the alleged identity between consciousness (i.e., the phenomenon picked out by the axiomatic approach) and Φ (i.e., the informational measure resulting from the postulates) cannot be justified – it is a case of mistaken identity.
The authors focus mainly on the axioms of integration and information, but their analysis seems to rest on two misconstruals of the core notions of IIT. First, they argue that the phenomenological property of unity under a point of view is not represented in the formalism. But IIT's first axiom, that of intrinsic existence, states that to exist intrinsically is to have causal powers upon itself: The notion of unity under a point of view should, therefore, be assessed under the intrinsic aspect of integrated information. In IIT, phenomenal unity is purportedly captured by the irreducible causal powers of a system, but the unity under the system's point of view is purportedly captured by the exertion of the system's powers upon itself. Merker et al.'s analysis, however, does not include this intrinsic aspect of integrated information, and therefore, their criticism seems misconstrued.
Second, Merker et al. claim that IIT conflates the notions of informativeness and that of informational capacity in accounting for the property of phenomenological differentiation. But the information axiom just states that every experience is specific: The specific form of an experience works as phenomenological principium individuationis, whereas a specific cause–effect profile of its physical substrate works as physical principium individuationis of that experience. In IIT, differentiation means “specificity,” but Merker et al.'s analysis of the information postulate does not consider this particular meaning.
Thus, Merker et al.'s thesis that there is a mismatch between axioms and postulates is not convincing, because it is based on inaccurate interpretations of IIT's axiomatic approach. However, I agree with the authors that the translation of axioms into postulates is a crucial issue for IIT: If we want to explain consciousness by starting from phenomenology, we need to make sure that the explanatory identity (Haun & Tononi, Reference Haun and Tononi2019, p. 5) between Φ and consciousness is a unique link, properly extracted from consciousness itself.
In this light, Merker et al.'s point on IIT's mistaken identity can be best reformulated in terms the non-uniqueness problem for IIT (Hanson & Walker, Reference Hanson and Walker2021): If the path going from consciousness to Φ can take different routes, the explanatory validity of Φ as a measure of consciousness can be questioned.
In order for IIT to maintain that Φ uniquely matches consciousness, and not, say, network efficiency, IIT must establish a principled way to justify why other interpretations of Φ are not correct. I suggest that the real problem for IIT is that this aspect of the theory is currently missing: Because IIT does not specify the logical reasoning fuelling the inference of the postulates from the axioms, we do not even have the instruments to tell whether such an inference is valid or not. And, if we cannot tell whether the formalism is validly inferred from phenomenology, then we cannot tell whether there is a unique and proper match between Φ and consciousness.
From a logical viewpoint, the inference from axioms to postulates can be either a deduction or an abduction (Bayne, Reference Bayne2018, p. 2). If it is a deduction, then each axiom must be accompanied by at least a “bridging premise” that connects it to the corresponding postulate. However, I believe abductive reasoning is more in line with how IIT proponents think of the relation between axioms and postulates (Koch, Reference Koch2019, p. 75). In this case, uniqueness amounts to selecting the candidate postulate that, given independent background knowledge and best-maker criteria, best explains the corresponding axiom. But, because we are not given this relevant information, disagreement about the explanatory validity of Φ, and therefore about the potential of IIT as a theory of consciousness, can emerge. This is a shortcoming in IIT, but not of the fatal kind proposed by Merker et al.
Were IIT able to clarify this crucial aspect of its theoretical structure, the path between phenomenology and Φ might prove to be unique. IIT would then indicate a clear explanandum of consciousness science (i.e., phenomenal consciousness as picked out by the axiomatic approach), clear first principles (i.e., axioms and postulates), and a clear (abductive) methodology for mapping the explanandum into the explanans (i.e., the Φ-based formalism). Contrary to what Merker et al. argue, IIT does seem to have the potential – albeit not fully realised yet – to advance consciousness science.
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Conflict of interest
None.