Integrated Information Theory and the Metaphysics of Consciousness

Presenter: Matteo Grasso, State University of Milan

Commenter: Jorge Morales, Columbia University


  1. Thanks for the interesting paper, Matteo! It’s not where you focus your interpretative energies, but I wonder what you make of the “exclusion postulate” from Tononi 2012? It seems that there is trouble brewing there, e.g., in the implication that if we integrated enough societal-level information the individual people in the society would all lose consciousness, even without any change at the individual level or any change in their self-reports of consciousness. (I develop this objection to Tononi in the longer version of my paper for this conference.) Alternatively, if we stick with Tononi 2004-2008, it seems like we end up with many layers of consciousness in a way that seems pretty unintuitive and which, I suppose, probably seemed sufficiently weirdly unintuitive to drive Tononi to modify his view by adding the exclusion postulate. Thoughts?

  2. Interesting article. The analysis looks pretty good to me, except that I didn’t really follow why type-E dualism was excluded as an interpretation. Prima facie there’s a view in the vicinity of type-E dualism that takes the information integration as a fundamental law connecting consciousness to physical states. You seem to suggest that the type-E view is ruled out by the claims that (i) there can be a scientific theory of consciousness, and (ii) the physical world is causally closed. But of course both of these claims are quite compatible with type-E dualism.

    Anyway, I agree that Tononi himself is not a type-E (or type-D) dualist, and that he’s not a type-A or type-B materialist. His writing on the topic can be read in various different ways, but from conversations with him it seems pretty clear that he intends to be something like a type-F monist or perhaps some sort of idealist.

    For what it’s worth, I think it’s not so easy to reconcile IIT itself with type-F monism. Type-F monism standardly requires that the fundamental connection between physics and consciousness lies at the level of fundamental physics, whereas IIT suggests that the fundamental connection is at the level of structural organization above the fundamental level. (The issues here are somewhat connected to the issues that Miguel Sebastian discusses in his paper here on the tension between panpsychism and organizational invariance.) I don’t really see how a version of type-F monism can allow fundamental connections above the fundamental microphysical level (e.g. fundamental phenomenal properties serving as grounds for macrophysical dispositions) without violating microphysical causal closure or else making those phenomenal properties epiphenomenal. And if the fundamental connections are all at the level of microphysics, then it looks like the information integration principle (connecting consciousness in a brain to phi in that brain, say) can’t itself be fundamental. Perhaps it could still be a consequence of fundamental principles that apply only at the microphysical level, but it’s not easy to see how that could work. By contrast it’s much easier to see how a type-E version or a type-B version of the view will go.

  3. I think Mateo’s analysis was comprehensive and clearly presented, I can’t remember reading a metaphysical analysis that was so clearly written. After reading David Chalmers post I above I am wondering how type F monism views the regular phenomena of “assignment” as grounds for a macrophysical disposition, such as the law of identity?

  4. I read David’s original paper PCEG and found it incredibly thought provoking and interesting. My POV as an electrical engineer that type F Monism could be a fundamental component of conscious being as a charge stored memory cell is a fundamental component of computer memory. I am a dualist in two senses, firstly as I sit here I perceive this room totally outside my body although I know it is being created inside my brain possibly by monistic cells which are arranged in incredible structural functions I.e an incredibly complex neocortex which overlays sensory cortex functions and motor functions. The second dualist property comes from David’s Zombie Conjecture, I look inside the zombie and see those identical synapse firings however I theorize those neurons are metabolically interacting to enact some shared inner functions between adjacent cells in the cortical areas of non-zombies causing qualitative phenomenal states.

  5. David, I’m very sympathetic with your remarks. In my comments I suggest that although Tononi uses the term “fundamental,” his description of consciousness as well as the consequences that follow from his view should put IIT closer to type-B materialism.

    I didn’t see either why type-E dualism was excluded as an option for IIT. I didn’t include this in my comments above, but the worry goes like this. Matteo, your definition of Naturalism (N) “Whatever exists is susceptible to explanation through methods which are continuous from domain to domain and which are paradigmatically exemplified in the natural sciences,” rules out Cartesian dualism but it’s not clear to me why it should rule out Type-E dualism (i.e. epiphenomenalism). But if this is so, then Type-E dualism and Type-F monism would be indistinguishable since both accept the four assumptions (naturalism, causal closure, epistemological gap, and ontological gap) . What went wrong? My guess is that naturalism in the way you defined it is not fine-grained enough to distinguish between Type-E and Type-F views. Thoughts?

  6. Hello everyone. First of all let me thank Richard Brown for the organization of this amazing event, it is a great opportunity and a pleasure to participate. Let me also thank Jorge Morales for his great job on my paper, and everyone who is visiting my page as well. I will try to reply to comments as soon as possible.

    Jorge, thank you very much for your comments. I completely agree with your main remark that type-B materialism is more than a mere conceptual possibility for IIT. Actually I think that embracing type-B materialism would be the easiest way for IIT to account for consciousness from a “traditional” scientific perspective, that is to say a naturalistic and materialistic perspective. Nevertheless, I think that the interesting point to make here is that a quantitative theory which matches neuroscientific experimental results so well, is also compatible with a non-reductive perspective. What I tried to argue for is the idea that, on the one hand IIT theorists could decide to construct IIT as a rigorous scientific theory of consciousness, embracing type-B materialism. On the other, they could construct their theory as a non-reductive theory of consciousness (as they seem to be doing). Since IIT is a mathematical and quantitative theory and it is collecting more and more scientific confirmation, I think the second possibility sounds very interesting. As I claimed in my paper, these two different perspectives lead to very different implications which are to be discussed further. My analysis is nothing but a small contribution. Jorge, as you have well predicted, comments are already fueling the debate about IIT’s metaphysical assumptions. Thank you very much for you remarks.

  7. Dear Eric, thanks for your interesting comment. IIT is developing fast and I think Tononi’s 2012 update is very useful. In my analysis I focus on IIT’s fundamental assumptions, but I’m very interested in deepening IIT’s implications at a broader level. I agree with the fact that Tononi 2008’s position was critical. According to his claims, every object which integrates minimum levels of information is conscious, being a “complex”. In 2008 he was using the concept of “main complex” but apparently that was not enough to explain the hierarchy of sub-complexes and the relationship among them. Moreover, as you underline so well, without exclusion postulate it was difficult to define which complexes count as subjects of experience, and what is the difference between the levels. As Tononi claims, conscious experience depends on the state of the whole system and “complexes can interact but they cannot overlap” (Tononi, 2012, p. 67), but the exclusion postulate generates troubles, as you show with the example of consciousness at societal-level. I strongly believe that these questions must be put on the agenda of its theorists in order to clarify IIT’s implications with respect to panpsychism and the distribution of consciousness in the natural world. As far as I am concerned, the only contribution that my paper brings to this topic is to show that both as type-B materialism or as non-reductive type F-monism IIT suggests a form of panpsychism. Nonetheless, until the nature of the monistic principle posited by IIT is not clarified further, it will be difficult to build claims about the distribution of consciousness upon a solid metaphysical basis. Your paper about these issues is very insightful and I will surely think a lot about it, thank you very much.

  8. David, your comments are very thoughtful, thank you very much. With respect to the first issue that you raise, I agree that type-E dualism is not ruled out by a naturalist perspective, and that it accepts the causal closure of the physical world. The taxonomy I presented is obviously a provisional one, because there are philosophical positions that it doesn’t include. I focused on the positions you named type A, B, D, E and F, but I believe that a complete analysis should include also type-O dualism (centered on overdetermination) and type-I monism (idealism, that as you say has a certain appeal on Tononi). As Jorge pointed out, my definition of naturalism may not be fine-grained enough to distinguish between type-E and type-F. Nevertheless, with respect to type-E dualism I believe that the issue might not be on the definition of naturalism. I think maybe type-E and type-F can be better distinguished through a deeper analysis of the last assumption: the existence of an ontological gap (OG). Indeed both type-E and F accept OG, but for different reasons. For type-E dualism OG is true because there are two separate substances: consciousness is a fundamental “entity” (i.e. metaphysical principle) of reality. On the contrary, type-F monism accepts OG because qualia and conscious experience are fundamental “properties”, irreducible to the physical aspects of reality. I think that the distinction between considering consciousness as a fundamental entity or as a fundamental property is a crucial one and can be misleading. I think you are right when you claim that Tononi’s writing can be interpreted in different (maybe opposite) ways, but it seems clear to me that he rejects dualism, even in a type-E form. As you point out, I should refine my taxonomy to clarify how IIT can be made compatible with type-F but not type-E, for instance by distinguishing more clearly the OG assumption about entities and properties.
    Coming to your last point, I believe that my analysis might be the first step to show that IIT actually is a suitable candidate as a quantitative and neuroscientific theory of consciousness which at the same time embraces type-F monism. It seems to me that (under the hypothesis that IIT embraces monism) IIT’s claim about the monistic principle of reality is not clear. On the one hand, positing consciousness as the very metaphysical principle of reality, IIT would embrace type-I monism. On the contrary, positing consciousness as only a fundamental aspect of a more profound metaphysical principle, IIT would still be a form of type-F monism (in its protophenomenal form). For instance, Tononi says IIT is compatible with the postulate of information as a monistic principle. I am afraid this is not enough as a complete answer to your (tricky) question. However, I think you raised a very interesting point. Since I’m currently working on it I’d be glad to discuss with you further.

  9. matteo: as i understand things the type-E/type-F distinction is different from the substance/property distinction. type-E dualism is compatible with both substance dualism and property dualism. there are quite a few epiphenomenalist property dualists (e.g. frank jackson in the days of “epiphenomenal qualia”, and i was at least open to this view in “the conscious mind”). these certainly qualify as type-E dualists, but they hold that consciousness involves fundamental properties but not fundamental entities. that was the sort of view i had in mind as one version of IIT: fundamental laws connecting phi to the fundamental property of consciousness. still, you’re probably right that this view is more dualist than tononi intends.

  10. Hi everyone, this is a great discussion!

    I must confess that I do not know a lot about Integrated Information Theory so please forgive me if my comments are somewhat naive or misguided.

    It seems to me that there really are two projects here. One is trying to figure out which metaphysical theories are compatible with what Tononi actually says, the other is trying to figure out which metaphysical theories are compatible with Integrated Information Theory in general (irrespective of what Tononi thinks). It also seems like a lot of the discussion is focused on the former, whereas I am more interested in the latter. Given this I don’t see why we ought to rule out substance dualism. Is there any principled reason that non-physical substances cannot have integrated information? In a Cartesian world there is consciousness, and if consciousness is identical to integrated information then there must be integrated information in a Cartesian world, right? So, though Tononi probably does accept N and CC, there is no reason why a type-D dualist needs to is there?

    The same seems true for type-A materialism. Couldn’t someone hold that it is a priori that when there is high phi there is consciousness? Again, Tononi doesn’t but what is the principled reason for ruling this kind of view out? In fact it is not clear to me that Tononi has ruled out a type-A position. He does admit that Mary can’t know what it is like to see red without seeing it, but certainly once she comes to see red for the first time and thereby acquires the appropriate phenomenal concept she might be in a position to see the a priori connection between the two. It is notable in this connection that Churchland (who seems like a type-A type) makes the very same distinction between being and describing, which again seems to suggest that this is no bar to a type-A position.

    Finally, I am not clear on how type-BS positions are supposed to be different from type-E/F positions (this is related to something that Dave has argued elsewhere). On these kinds of views we have two distinct properties one of which supervenes on the other. If that doesn’t mean that they are identical then what does it mean? The only thing I can think of is that there would have to be some kind of fundamental laws which connect the two properties, and that doesn’t seem like a type-B position to me. In fact it sounds like the line that Dave was suggesting above.

  11. Dear Ken, thank you for the interesting comment. I was thinking about the possible relationship between type-F monism and assignment as grounds for macrophysical disposition (like the identity law), but I’m not sure that I have understood what you are referring to. I’m sorry about that, but I’m not very confident with the things you named, so if you could help me providing some more context I promise I’ll meditate on it and I’ll try to answer as best as I can.

  12. Richard, I thoroughly enjoyed your comment, thank you so much. I will start from your remark about type-A materialism. If I get it right, the only difference between type-A and type-B materialism is about the explanatory gap. Type-A materialists hold that the identity between phenomenal (Q) and physical (P) properties is a necessary “a priori” truth. They deny the existence of an ontological gap (as materialists) but they deny also the existence of an explanatory gap, because they say that we can fully understand and explain the identity P=Q (our present incapacity to explain is just the result of a (temporary) limit which regards our knowledge of the physical substrate of consciousness, e.g. Churchland’s position). However, they have to explain why that identity is so difficult to prove for us, and why it seems difficult to deduce truths about Q from truths about P. On the other hand, also type-B materialists deny OG and accepts that the identity P=Q is necessarily true. However, they claim that probably we won’t ever be able to explain Q in terms of P, either because we are not intelligent enough or because this explanation for natural sciences is out of reach. As suggested by David, one way for type-B materialists is to claim that P=Q is necessarily true (in every possible world), but it is an “a posteriori” truth in the kripkean sense. In other words: it is true even if we cannot explain why, even if we cannot prove the actual identity between specific couples of Ps and Qs. I think IIT is compatible only with type-B materialism because it accepts that the identity P=Q is necessarily true, but it states that it is only knowable a posteriori, and maybe it is even impossible for us to understand the relationship between specific couples Ps and Qs. Indeed, for IIT an explanation of consciousness in terms of integrated information is complete and “ontologically satisfying”, although from that we cannot deduce truths about phenomenal qualities, and we never will.

    Facing now the issue about the ambiguity between type-E dualism and type-F monism: I think it is a tricky one. As I said, my analysis of IIT is provisional, and I’d like to thank everyone because I have already gathered a lot of insights from this debate. I personally think the assumptions about OG are more fundamental than the ones about EG, for they allow us to distinguish between materialism and substance dualism. In fact, if a theory T accepts OG at the level of substances (let’s say t accepts “substance OG”) then it is for sure compatible with substance dualism. To the contrary, if T rejects “substance OG” then substance dualism is for sure excluded. However, in this last case, T can still be compatible with materialism and idealism (type-I) if it rejects “properties OG” as well. On the other hand, if T accepts “properties OG” then it is compatible with type-F monism. As David says, this last position can be considered also a kind of type-E “property” dualism. However, since it is not a substance dualism (remember, it rejects “substance OG”), I am inclined to think that this position is closer to type-F monism (the only one that accepts “property OG” rejecting “substance OG”) than to type-E dualism (which it seems to me accepts both “property” and “substance OG”). If we have to guess from Tononi’s words, we would say that IIT rejects substance dualism (for it rejects “substance OG”), nevertheless it seems compatible with property dualism (for it might accept “property OG”), therefore it seems more compatible with type-F monism than with type-E dualism. Maybe it is only a matter of definition, in fact I agree with what David said in his comment about the compatibility with IIT and property dualism.

    At this point, I think that a plausible reply to your remark about the difference between type-BS materialism and type-E/F could take the same shape. The only difference that I see is that according to type-BS materialism, consciousness is a property that supervenes on the physical ones. Conversely, according to type-E substance dualism, consciousness is not a property at all, instead it is an epiphenomenal substance without causal power. For type-F monism and type-E “property” dualism (if this positions are distinguishable, which is something that I am skeptic about) consciousness is not a property that supervenes on physical properties, but it is a fundamental property itself, just alike the physical ones.

    Coming to the “second project” you named, apart from the exegesis of Tononi’s words, IIT’s metaphysical assumptions are not clearly defined so I think we can speculate a lot about its compatibility with almost every position. I was thinking hard about your idea that there can be integrated information also in a Cartesian world. I agree with that. Tononi defines information as reduction of uncertainty, therefore information can be generated from every system, physical or not, which satisfy the conditions to generate Shannon’s information. The point is that if we accept substance dualism, then we have that information is generated both from physical and non-physical systems. It could makes sense, but in that case the same law would regulate the generation of (integrated) information in two very different substances. There is nothing inconsistent in this scenario. Nonetheless, given the central role of information, maybe we should consider an alternative perspective, more parsimonious and perhaps more interesting. Indeed, under the hypothesis that information is the common property which links physical and mental substances, we should better consider the chance that information is itself something more than a mere property. As David suggested with his “double-aspect theory”, embracing an “it from bit” perspective type-F monism could define both physical and phenomenal properties as different aspects of a unique metaphysical substance, that is to say, information. Under this hypothesis, phenomenal and physical properties are both fundamental and mutually irreducible aspects of information, that is the only metaphysical principle. In my opinion, this form of non-reductive monism that embraces property dualism is closer to type-F monism than to any other position. Moreover, since Tononi’s theory is compatible with this perspective, it seems plausible to suppose that IIT could represent a “scientific” version of the double-aspect theory, which provides also the mathematical laws and neuroscientific data on the correlation between consciousness and information.

  13. Vicpanz, thanks for your comment. I’ve found very interesting the second sense in which you are dualist. I was wondering if you identify your position compatible with property dualism and if you see it closer to type-F monism or to dualistic positions like type-D/E. Moreover, you say in non-zombies neurons interact metabolically to enact shared inner functions between cortical cells causing qualitative phenomenal states. Now, under the zombie hypothesis zombies are physically and functionally indistinguishable. If so, also the processes you describe would be identical. Therefore, how do you explain zombies’ lack of phenomenal consciousness?

  14. Matteo: Thank you for the reply and also thanks to Richard for the conference and his reply on the subject. As Richard pointed out you could make the PHI theory true for any position and I happen to agree. In reality I am closer to Churchland’s position but I accept dualism….Huh?,,as you say. Well as I originally pointed out we are made for the dualist perspective (The room is outside us, the manifest image, thunder and lightning are two events etc.) In reality we are made from monist material that evolved itself into this dualistic mind we have. As you say given every physical fact if P–>Q and we can conceive a Zombie then physicalism is false (as our dualistic minds see it) or given thunder there doesn’t have to be lightning. However call me a “deeper functionalist” and conceive of every particle in nature not as a physical particle (which it may be) or a unit of consciousness (which it may be) BUT as a FUNCTION. Given all functions F, if F–>Q can you conceive a Zombie and I would say NO because we cannot see or conceive the processes from deep in the cell to the outer functions which Type A materialists call conscious, semiconscious, unconscious etc. functions. Afterall based on the scientific image scientists delve deeper into the physical world based on mathematics or the mapping of function.

    So now you ask what is my theory of how the monads of function can turn themselves into the eventfulness of conscious phenomenal biological mind spaces which cobble themselves into higher and higher functions?….Give me some time.

  15. Thanks for your reply Vicpanz, your ideas are very interesting. Now I see that you clearly rejects both materialism and idealism, as well as substance dualism. Despite functionalism is compatible with type-B materialism your position seems closer to type-F monism. However, if you see *function* as the basic principle of reality I wonder what makes you think your position is close to dualism (of substances or properties). I’m very interested in your solution of the whole issue and I will keep an eye on your future research for sure.

  16. Honestly Matteo I believe there is a solution to the problem so let me start you with this point: The phylogeny of human bicameralism is the phylogeny of bipedal organisms. If we analyze the human person from the sensory motor POV, the language of movement is networks generated in the higher lobes of the brain and fed back into the motor neurons. Without a sensibility of time, it would be impossible to generate this language, which is why I alluded to time and consciousness in the previous. We could say this silent language of movement is our inner zombie which requires conscious time experience.

    The reason why I think consciousness, PHI, eventful ness, time etc. is an embedded property of function because the more complex the physical function, the more space and extension it occupies so a biological cell is a higher function than an inorganic crystal etc. I theorize that neurons achieve higher extension when they fire like a Fermi chain reaction. Do they somehow align to form a supercell and higher extension into space? A millimeter of space can perhaps generate a large enough cluster supercell to cause time emergent phenomenal experience? I would theorize that the cells interlock or take control of each other.

  17. Dear Matteo,
    I am impressed by the skill with which you have addressed the original agenda. However, I am concerned about the validity of a premise. You say that IIT embraces naturalism. As Richard Brown points out, there is a distinction between its adherents saying they embrace naturalism, if they do, and IIT being compatible with an embracing of naturalism. The crucial question is whether IIT is itself a naturalistic theory. If not then it would seem to be incompatible with your N. My concern probably overlaps with other’s queries about type E.

    My understanding is that in the current scientific environment a naturalistic theory needs to be based on **local** causation. There is another, local, sense of integration in brains or computers but Tononi’s concept of integration is not local. So of itself it can do no causal work. The causal work must be done by the basic local physiological events. That seems to indicate that IIT may be a useful précis of the physiology, but having no extra causal explanatory power it is arguably not in itself a naturalistic theory. If it aims to provide some explanation over and above the local physiology I am fairly sure it is not a naturalistic theory. I may be wrong here but I think it has to be thought hard about. The fact that the theory comes from a neurobiology lab does not necessarily make it naturalistic. The fact that it fits with empirical evidence may reflect more on the way it has been devised to fit that evidence than anything – this is the common situation in science.

    One specific query I have is about the definition of integration. It is a simple objection, that again may be wrong, but it worries me. In a computational system a null signal is as important as an active signal, and with inhibitory signals it gets worse. So cells that are not firing are, from the information point of view, as integrated as cells that are not firing; i.e. the whole brain is integrated if anything is. IIT seems to put weight on synchrony of firing, but as Bjorn Merker has recently pointed out, synchrony is only of computational relevance at the point of **local** causal integration of convergent signals downstream. Synchrony of distributed events is just the steam whistle EEG trace that tells us the spike train is coming to some well-timed local integration downstream. It may be important that there are some well-timed null signals arriving at the same point as well.

    These are just some brief points, but my current thought is that IIT is not a naturalistic theory in the sense that I, as a biologist, recognize. If it were a theory of local causal integration then naturalism would be much more reasonable. (I also, like Eric, have worries about maxima and minima and the monsters that threaten if we swallow an iPod that happens to have a bigger PHI than we do.)

  18. Dear Jonathan, thank you very much for your remarks. As you and Richard have pointed out, I’ve based my analysis mostly on what Tononi says about IIT, and I’ve speculated with my interpretations only where Tononi has not been clear enough (e.g. about the assumptions that might resolve the ambiguity type-B Vs type-F). You are perfectly right, IIT could not be a naturalistic approach to consciousness regardless of what Tononi says about it. However, I think there is an ambiguity about the definition of “naturalism”. If N definition is based on *local* causation, as you claim, then probably IIT is not a naturalistic approach. As you say, this definition of naturalism is the one accepted in scientific fields like biology. Nevertheless, it seems to me that this definition of naturalism rhymes with reductionism, but that this is not necessary. Indeed, from a philosophical point of view, one can embrace naturalism without at the same time embracing reductionism. For instance, this is the case of non-reductive monism. Now, if it is true that Tononi’s concept of integration is non-local, then maybe IIT should reject naturalism as you have defined it, that is to say naturalism based on causal explanatory power. But, from a philosophical point of view, I think in this case IIT should reject only reductionism. As David pointed out, in fact, even type-E epiphenomenal theories could be compatible with naturalism. In my analysis I excluded the compatibility of IIT with this perspective, but for other reasons (because IIT rejects dualism). Obviously, if the concept of integration will turn out to be non local, then we should consider IIT incompatible with reductionism, or naturalism in the sense you define it, but I don’t think this is the aim of Tononi. A part from that, I agree that without causal explanatory power, IIT would remain a non scientific theory, even if interesting from a philosophical point of view. I hope that IIT will be developed keeping an eye on these issues, and I hope that the dialogue with the philosophers will be fruitful in facing up to these problems.
    With respect to your remark about null signals I completely agree. However, I remember Tononi insisting on the fact that the “silent” state of neurons is as informative as an active one, because given the same repertoire of possible activations, silence rules out exactly the same amount of alternative states. In fact it is only when integration breaks down (like during deep sleep or anesthesia) and the repertoire is narrowed, that both activity and silence are differently informative. Where is Tononi’s mistake about that?
    Even if IIT seems to me appealing from a philosophical point of view (maybe especially as a non-reductive theory), I agree with you that as a scientific theory of consciousness it would be much more reasonable to ground integration upon local causation. I thank you very much for your comment, I’ll continue to think hard about the objections that you’ve raised.

  19. Hi Matteo, thanks for your very helpful response! A couple of thoughts in response.

    Re type-A: What I was trying to suggest was that it is not clear that Tononi himself couldn’t endorse a type-A view. He seems to try to give an argument that consciousness is a priori associated with phi. This is what I take the ‘photodiode’ and ‘camera’ thought experiments to be trying to get at. Taken at face value this looks like a denial of the explanatory gap. All we have done is some a priori reasoning and ended up with the conclusion that consciousness is integrated information (in fact he seems to endorse this when he says, “According to the IIT, consciousness is one and the same thing as integrated information. This identity, which is predicated on the phenomenological thought experiments at the origin of the IIT, has ontological consequences,” (from the 2008 provisional manifeso paper)).

    So, what then are we to make of his comments about Mary? You take these comments to suggest that he endorses an explanatory gap but I don’t think that this is what they point to. Rather, it seems, what he is really getting at is that knowledge of phi is not the same thing as being a system with phi. Once we have had the relevant experience it then is a priori that consciousness is phi (as per the thought experiments). This is why I was suggesting the comparison to other type-A materialists. The Mary thought experiment, for them, merely shows us that describing the conditions for consciousness are not the same as implementing them but does not suggest an explanatory gap (and it doesn’t automatically suggest that the knowledge of the identity is a posteriori *as long as the relevant experience plays a merely enabling role,* as for instance in allowing one to have the appropriate concepts to allow the a priori reasoning in the thought experiments).

    Re type-E and F: you say “…type-F monism (the only one that accepts “property OG” rejecting “substance OG”) than to type-E dualism (which it seems to me accepts both “property” and “substance OG”)”

    I don’t think this is right (as Dave himself pointed out in an earlier comment). Type-E dualism is simply committed to some kind of dualism plus the claim that consciousness is epiphenomenal. One could accept substance dualism and epiphenomenalism (maybe Parallelism is an example of this from the history of philosophy) OR one could accept property dualism plus epiphenomenalism (this was Huxley’s view and is probably the more standard kind of epiphenomanlism that we find in history of this view). And as far as I understand that view they do accept supervienence claims (which really are very weak since they merely require that a difference in one set of properties must be accompanied by a difference in the ‘base’ set of properties) You say you are skeptical that we can actually distinguish these two views but I am not sure why you say that. Any thoughts on this would be appreciated!

    Re type-D: I found your suggestions here very interesting but a couple of things occurred to me as I was thinking about this. Type-D dualism is the view that the two substance (or properties) interact with each other. So you could have a type-D property dualism if one held that physical states produce non-physical properties *which then in turn have physical effects* OR you could have the more familiar kind of interactive substance dualism. But what struck me was you claim that such a view is committed to the claim that there is integrated information in both the physical and non-physical system. I wonder about that since, if integrated information is identical to consciousness, then if there is integrated information in the physical system then there is consciousness in that system, but that conflicts with the commitment to an ontological gap. On the other hand I was thinking that maybe the integrated information might somehow be a property of the two substances interacting, and if so there might be room for a kind of hylomorphism about consciousness from an IIT perspective…this last thought is rather sketchy but in any case I am interested to hear what you think about any of this.

  20. Dear Matteo,
    Thank you for your clear and thought-provoking response. The issue of reductionism is interesting. Do I claim that IIT is not reductionist? I think maybe the opposite. This may sound strange but I think it is worth looking at this carefully.

    Wikipedia says reductionism is ‘…philosophical position that a complex system is nothing but the sum of its parts, and that an account of it can be reduced to accounts of individual constituents’. IIT makes use of a concept of integration that applies to a whole system. However, it is entirely derived from, and thus arguably reducible to, an account of its parts. Compare this with explanations in modern physics based on local causal interactions, for which **there is no concept of parts**. There may be fields, but no parts. People assume that physics reduces to an atomic level, but that was the blind alley of Rutherford’s era. The reflection of light off a horizontal mirror cannot be explained in terms of interactions with billiard ball electrons in atoms because billiard balls are not horizontal. We have gone back to a **partless** account. I am oversimplifying, but I think the drift is right. The need for the complexity of percepts to be explained in terms of something local with **no parts** is the whole basis of the Monadology, which I see as close to the spirit of modern physics. All relations are local and causal, so there can be no relation of ‘being part of’ (which IIT depends on) in physics, as far as I know. This poses a knotty problem for philosophy of mind, but if we can handle mirrors now, maybe we can handle minds too.

    People think of science as reductive because it involves dissecting, especially biology. However, that is just to focus on an event of interest. The dynamics of an atom are analysed at atom level but the dynamics of a girder are now seen as existing at girder level. I think there is often conflation between reductive and causal. Naturalistic accounts are causal, even if the meaning of cause has shifted. Reduction is choice of level of explanation, not part of the explanation. Maybe we should bin ‘reductionism’ as a red herring. When we do find rules of correspondence between qualia and nerve cells it will not be a reduction because one will not be part of the other. It may not be causal either. However, that does not mean it cannot be part of a naturalistic account because all causal accounts have implicit within them that this correspondence obtains at the point of observation where physical causation is complete (as in Newton’s famous letter to Oldenburg). As Russell put it, it will just be a translation: Mary’s new dictionary. Our problem is in finding a partless local causal neural event to open the page at.

    I note your comment that Tononi recognizes the equivalence of silent or null signals and active ones. I am then puzzled as to how a repertoire gets narrowed if the number of synapses remains constant. But this is detail.

  21. Hi Matteo,

    Enjoyed the paper. Perhaps this has been said above, but I worried a little bit that your description of type-F monism sounded more like property dualism. One important feature of the Russellian view (or at least an interesting version of it) is that phenomenal/proto-phenomenal properties are the very properties denoted by the predicates of physics. So we get a kind of identity between physical and phenomenal/proto-phenomenal properties, but in which the phemomenal concept rather than the physical concept reveals the nature of those properties.

    Also, two small worries:
    A. The definition of causal closure was a bit stronger than I’ve seen it before. The claim is usually that all physical events have sufficient physical causes (and then in conjunction with a ban on overdetermination, we reach the conclusion that irreducible mental properties can do no causal work).
    B. Although Kripke is no doubt an influence on type-B physicalists, I would dispute the claim that Kripke himself is a type-B physicalist. ‘Naming and Necessity’ offers an argument against physicalism.

  22. Richard, thanks again for your insightful remarks. On type-A materialism I agree that Tononi’s thought experiments show that “consciousness = integrated information” is an a priori truth. I don’t know if my interpretation of type-A is too extreme, but it seems to me that this position also claims that once we will know everything about the physical features of the brain, then there won’t be anything unexplained about the emergence of consciousness. IIT, as a form of materialism, could accept the identity consciousness = integrated information, but it would still claim there are limits to our knowledge. IIT provides a mathematical (or geometrical) model of qualia (as sub-spaces of the Qualia-space characterised by informational relationships that univocally specify each and every experience), and it states that once we have a description in terms of informational relationships within the Q-space, we have a complete description of the qualitative properties of a specific conscious state. Qualia are “nothing but” activation states that rule out a specific set of alternative states of a specific repertoire. If we interpret IIT as materialist, there is nothing missing in this description from the ontological point of view (there is no OG). However, the mathematical description of a specific quale doesn’t allow us to know “what it is like” to experience that quale first-hand. While for the scientific description of water there is nothing missing both from the ontological and from the epistemic point of view, for consciousness (under the “IIT as type-B” hypothesis) there is nothing missing from the ontological point of view, but there is something gappy on the epistemic side that scientific descriptions cannot face up to. Again, maybe I’m interpreting type-A as much more extreme than it really is, so I’d be glad to hear your thoughts about it.

    Coming to type-E I totally agree with you and I’m starting to think that it is just a matter of definitions. In my analysis I’ve classified type-D and E dualisms as “substance dualisms” in order to distinguish them from monisms. One of the most important lessons that I’ve learned from this discussion of my paper is that, in my taxonomy, I should distinguish substance and property type-E dualism, and I am very thankful to you for this result. However, it still seems to me that property dualism, with consciousness as a fundamental property besides the physical ones, is also compatible with a form of type-F monism. In fact, as David says in his 2002 paper:
    “From another perspective, [Type-F monism] can be seen as a sort of dualism. The view acknowledges phenomenal or protophenomenal properties as ontologically fundamental, and it retains an underlying duality between structural-dispositional properties (those directly characterized in physical theory) and intrinsic protophenomenal properties (those responsible for consciousness). One might suggest that while the view arguably fits the letter of materialism, it shares the spirit of antimaterialism.”
    We can distinguish this form of type-F monism from type-E property dualism for the fact that the latter posits phenomenal properties as epiphenomenal while type-F doesn’t. For this reason, I tend to consider IIT as compatible with type-F monism but not with type-E property dualism. Moreover, since David says also type-D dualism is compatible with property dualism, maybe there is an issue in the starting distinction of these positions. Nevertheless, I agree that in my analysis I should distinguish better the various positions and their basic assumptions. I hope that this clarifies a little my claims about type-E/F, but since this debate has been very fruitful until now, I hope to hear more of your thoughts about that.

    At last, I find your comment about type-D very thought-provoking, thank you! I thoroughly enjoyed your idea of hylomorphism between the two interacting substances. It makes me think in a new way about what I said to you in my previous comment. Indeed, the hylomorphism you theorize could rest upon the very concept of information itself. Since information is defined as a reduction of uncertainty or entropy, it could be generated from whatever substrate. If we accept this, there are at least two possibilities that come to my mind: (1) the relevant kind of information that constitutes consciousness (integrated information?) can be generated both by systems which are exclusively physical or exclusively non-physical (we could have both physical or disincarnated consciousness), or (2) information can be generated by each substance alone, but consciousness cannot, for the relevant kind of information that constitutes consciousness (integrated information?) is only the one produced by the interaction between the two substances. The second option would be coherent with the hylomorphism you are proposing. These thoughts are even more sketchy than yours but I find the whole idea very interesting and I’d be glad to discuss it further with you.

  23. Dear Jonathan, thanks for your interesting reply. What I had in mind for reductionism is more similar to the definition of “ontological reductionism” on wikipedia, that is to say a definition closer to the one of physicalism. For instance, I’m referring to the one of the IEP: “The type of reductionism that is currently of most interest in metaphysics and philosophy of mind involves the claim that all sciences are reducible to physics. This is usually taken to entail that all phenomena (including mental phenomena like consciousness) are identical to physical phenomena.” ( ) or the one on the MITECS: “reductionism within the cognitive sciences holds that neuroscientific theories will explain the success of psychological theories and, therefore, will reveal that psychological states and processes are nothing but bodily states and processes.” ( ). Given these definitions, I think IIT could be a naturalistic theory of consciousness even rejecting reductionism. I find very interesting your thoughts on locality and causality and I’d be glad to hear more about the issue they might represent for the philosophy of mind.
    Coming to the last point, I think IIT would reply that the repertoire gets narrowed even if the number of synapses remains constant and the brain seems equally active. Tononi writes: “Due to changes in intrinsic and synaptic conductances triggered by neuromodulatory changes (e.g., low acetylcholine), cortical neurons cannot sustain firing for more than a few hundred milliseconds and invariably enter a hyperpolarized down-state. Shortly afterward, they inevitably return to a depolarized up-state (Steriade et al., 2001). Indeed, computer simulations show that values of ⌽ are low in systems with such bistable dynamics (Fig. 4F, Balduzzi and Tononi, 2008). Consistent with these observations, studies using TMS, a technique for stimulating the brain non-invasively, in conjunction with high-density EEG, show that early NREM sleep is associated either with a breakdown of the effective connectivity among cortical areas, and thereby with a loss of integration (Massimini et al., 2005, 2007), or with a stereotypical global response suggestive of a loss of repertoire and thus of information (Massimini et al., 2007).” (Tononi, 2008: 223). To provide a quotation about your previous question on null signals, Tononi says: “The assumption that neural elements that are active are broadcasting information often goes hand in hand with the corollary that inactive elements are essentially doing nothing, since they are not broadcasting anything. According to the IIT, this is not correct. In the general case, being ‘‘off’’ is just as informative as being ‘‘on.’’ An element that fires specifies previous states that would have made it fire and rules out other states. Similarly, an element that does not fire rules out previous states of affairs that would have made it fire and thereby contributes to specifying the actual repertoire.” (Balduzzi & Tononi, 2009: 14-15).

  24. Philip, thanks for your thoughtful comments. Your first remark is very relevant, in fact most of the discussion here is focusing on property dualism. While I agree that IIT could be compatible with this position, in my last reply to Richard I’ve expressed my concerns about the ambiguity between type-D/type-E property dualisms and the form of type-F monism which posits the duality between structural-dispositional properties and intrinsic protophenomenal properties. I surely have to work on this distinction to improve my analysis.
    Coming to the definition of causal closure, thanks for your remark. I will think hard how the analysis of positions would be affected using the definition of causal closure that you have suggested, and I’d be very glad to hear your opinion, of course.
    About your point B, I’ve suggested that Kripke’s position is compatible type-B materialism for the same reason why I think IIT is compatible with this position, that is to say because type-B materialism is not necessarily committed to physicalism. As Chalmers says: “type-B materialism holds that phenomenal states can be identified with certain physical or functional states” (2002, p.14) and my claim is resting on the “functional version” of this position. However, if you think the issue about Kripke’s categorization remains I’d be glad to hear more about it. Thank you very much for your precise remarks.

Comments are closed.