(See the other posts in this series here!)
As Jonathan Birch points out, the Integrated Information Theory of consciousness (IIT) is an outlier in the field of consciousness studies. Other major theories, such as the Global Neuronal Workspace theory (GNW), are clearly compatible with materialism (i.e., physicalism) about consciousness, and may even be understood to presuppose it. In contrast, IIT seems to distance itself from a number of core materialist assumptions, and make claims that rather point in the direction of idealism.
Birch argues that it may therefore be impossible to empirically test IIT itself against other theories with a different metaphysical starting point, such as GNW. That is, it would be possible to test “conjectures made within these separate metaphysical pictures” but not the metaphysical pictures themselves.
While I largely agree with this conclusion, I still think there is a sense in which IIT itself could be regarded as fairly testable. Its metaphysical commitments can also be interpreted in different ways, and for those who disagree with them, it may also be possible to accept IITs empirical part while leaving its metaphysical part behind.
First, regarding testability, it seems that IIT can be separated fairly cleanly into a metaphysical and an empirical part, and that the empirical part can be regarded as essential to the theory itself (rather than a separate conjecture merely within the framework). The main empirical claim would be that phenomenal consciousness correlates with maximal integrated information, i.e., maximal phi. The metaphysical part then adds further assumptions about the nature of the correlation (such as it being a matter of an “explanatory identity”, understood as different from the identity posited by standard materialism). It also includes a set of the phenomenological axioms that IIT takes the correlation to be a priori derivable from, as well as further assumptions enabling this derivation (more details about IIT are explained here).
The empirical, merely correlational, claim of IIT seems testable against correlational claims of other theories – or, to the extent that it isn’t, this would be due to challenges of a different, largely methodological rather than metaphysical, sort. One challenge is that we need reliable indicators of consciousness that are independent of the theories to be tested, which can be hard to find or agree about (aside from verbal reports and so on, which may be insufficient to settle some decisive disputes). Another challenge, more specific to IIT, is that phi itself is difficult to calculate and measure. A third challenge is that different theories may not always agree about some necessary features of consciousness itself. For example, GNW seems to presuppose that phenomenal consciousness and access consciousness cannot come apart, whereas IIT thinks phenomenal consciousness can exist without access (at least in some senses of “access”). If so, GNW would interpret evidence for the absence of access consciousness as evidence of the absence of phenomenal consciousness, but IIT would not. These are all important challenges, but they are not clearly impossible to overcome.
It may also be possible to accept the empirical part of IIT without its metaphysical part, assuming the empirical part can be adequately supported by empirical evidence alone. The correlation between consciousness and maximal phi could in principle be accounted for in terms of, e.g., materialist identity or realization, a dualist psychophysical law (as explored by Chalmers and McQueen), or Russellian or dual-aspect panpsychism (as I have argued), so the empirical part does not seem to imply any particular metaphysics. Though one might think the metaphysical part of IIT—in particular, the derivation of the correlational claim from the axioms—plays an essential evidential role relative to the empirical part, and there may also be other reasons to take it onboard. So, what are the commitments of the metaphysical part?
I will focus on the implications for the mind–body problem, such as whether IIT commits us to idealism or non-materialism more broadly. To my knowledge, IIT (or its authors) has not explicitly endorsed any metaphysical theory; and in fact makes critical remarks about nearly all of them, including materialism, dualism, idealism and panpsychism (see Tononi and Koch, section 5), so its commitments must be reconstructed from other claims.
One significant hint is IIT’s claim that it may avoid the famous hard problem of consciousness in virtue of “starting from consciousness”, in the sense that instead of asking how consciousness could arise from physical processes (i.e., posing the hard problem), it posits a set of axioms about the essential properties of consciousness derived from first-person phenomenology, and from this derives the properties of its physical counterpart. But this starting point may seem more epistemological than metaphysical. There are also a number of metaphysical theories of consciousness that seem compatible with the possibility of such a derivation (in the sense of not clearly ruling it out), and none that by itself would clearly license it (with the possible exception of analytic functionalism, which it seems clear that Tononi rejects, and is widely regarded as implausible also by others).
Birch notes that Tononi and co-authors also make claims such as: “because my alternatives, reasons, and decisions exist within my experience—as sub-structures within an intrinsic entity—the neuronal substrates of alternatives, reasons, and decisions cannot also exist.” This may sound like a commitment to idealism, by denying the reality of the physical entities such as neurons (or other physical “substrates”).
But there is an alternative interpretation of this and other related claims, according to which they are mainly concerned with mereology rather than the mind–body relation. Briefly, the idea seems to be that a conscious system understood physically, such as a neural system, will causally exclude its own physical parts (and since existence requires causal powers, this means they don’t really exist, according to IIT). Still, consciousness itself always has a physical counterpart, such as a neural system, but this physical system is in a sense physically partless. On the face of it, this is quite a mysterious view, but it is not quite idealism of the sort that denies the existence of the physical world (as opposed to attributing to it a peculiar structure). One way to, perhaps, render it slightly less mysterious is to interpret it as denying the existence of, e.g., neurons in a similar sense that mereological nihilists deny the existence of, e.g., tables (in favor of “particles arranged table-wise”).
Birch also emphasizes IIT’s claim that only consciousness exists intrinsically, while physical entities are extrinsic. With this, I agree that IIT approaches idealism, but not necessarily the kind that regards the physical world as largely an illusion, i.e., subjective or anti-realist idealism. It is more reminiscent of Russellian panpsychism (named after Bertrand Russell, who defended a similar view, and which comes in both idealist and non-idealist versions, see Chalmers). On (the idealist version of) this view, the physical world is fully real, but still fundamentally mental. Very briefly, the idea is that physical properties are purely relational or dispositional (“physics only tells us what things do, not how they are in themselves”, as it’s commonly put), whereas consciousness is intrinsic and can therefore be posited as the relata of all physical relations, or categorical grounds of physical dispositions. The view is also fairly similar to Leibniz’ monadology minus the claim that monads don’t interact, as Birch also describes the metaphysical picture that seems to follow from IIT. However, there may also be some tension between Russellian panpsychism and other claims of IIT (some of which I discuss here and here).
Summing up, I fully agree with Birch that major parts of IIT are metaphysical and hence not empirically testable, and this sets it apart from other theories of consciousness—though with the caveat that materialism, typically presupposed by other theories, is a metaphysical theory as well. Still, one may isolate at least one part, the correlational claim, that is both testable and central/integral to IIT—at least setting aside some challenges to testability of a largely methodological rather than metaphysical sort. So, even though IIT definitively pushes the Overton window for neuroscientific theories, this is not to the extent that the possibility of empirical communication breaks down. The metaphysical commitments of the theory can also be interpreted in different ways, so the metaphysical Overton window need not necessarily be pushed all the way to subjective idealism, but Russellian panpsychism (which has gradually entered the Overton window in philosophy of mind over the last decade or so) could be a natural interpretation.
