A Review of Giulio Tononi’s New Book
Thursday, November 21, 2013
So many questions. How does matter produce mind? Or does it? Is there a soul? What happens when we die? Giulio Tononi’s new book, Phi: A Journey from the Brain to the Soul, seeks answers.
Tononi, a psychiatrist and neuroscientist at the University of Wisconsin, Madison, collaborated for a number of years with Nobel Prize–winner Gerald Edelman, developing with Edelman what was first known as the Dynamic Core Hypothesis, a theory of human consciousness, and later working on his own to develop the Integrated Information Theory (IIT) of consciousness. The essence of IIT is that integrated information is consciousness. As such, it can be quantified and characterized in great detail — but for now it can be quantified only in principle for most systems because doing so in practice is extremely difficult for any conscious system above the most basic level of complexity.
I’ve followed Tononi’s work for a number of years, drawing from his ideas in my own work on theories of consciousness. In particular, I found Edelman and Tononi’s book A Universe of Consciousness (1998) very helpful in shedding light on approaches to quantifying consciousness, and I have enjoyed and appreciated Tononi’s later work on consciousness, even if I don’t agree on all the details.
The key concepts of IIT, discussed in poetic and metaphorical terms in Phi, and in detail in Tononi’s more technical work, may be summarized as: 1) causation is information; 2) information is “integrated” when the whole consists of more than its parts, that is, the whole is irreducible to its parts because information would be lost through such reduction; 3) integrated information is consciousness; 4) we can quantify integrated information, at least in theory if not in practice at this time; 5) we can quantify and characterize actual qualia based on a translation of qualia to a certain type of information space.
IIT is catching on, but it’s taking a while: Tononi’s first publication on IIT was in 1998. This is, perhaps, to be expected for a theory as profound as a comprehensive theory of consciousness. Tononi has some high-profile supporters, including Christof Koch (see here for my interview with Koch), formerly at Caltech and now at the Allen Institute for Brain Science. Even the arch-materialist Daniel Dennett expressed some sympathy for IIT when I asked him about IIT a couple of years ago.
I attended the 17th annual meeting of the Association for the Scientific Study of Consciousness (ASSC) in San Diego in July where I saw Tononi speak twice. I also interviewed him one-on-one as part of my ongoing series of interviews with experts in various fields. The ASSC meeting is the most prominent annual meeting of scientists who study consciousness. I felt like I was witnessing a paradigm shift as it happened, the beginning of a shift from materialist views to frameworks that, while no less scientific, are not strictly materialist because they explicitly include consciousness as a fundamental property — with Tononi’s model being the most prominent in this latter category.
IIT had an extended moment in the sun at this conference, with a three-hour “tutorial” before the conference officially kicked off, a two-hour roundtable discussion midway through the conference, and a poster session by Tononi’s lab colleagues Larissa Albantakis and Masafumi Oizumi, who also took part in the tutorials. That’s some good coverage. Time will tell if IIT is getting a fleeting moment in the sun or is truly the mark of a nascent paradigm shift in this exciting field.
That said, I do feel, along with many others I spoke to at the conference, that while IIT has a lot of promise it still needs work — as Tononi himself would acknowledge. My two primary problems with IIT as it currently stands is that 1) it does not acknowledge its panpsychist implications, and 2) it doesn’t offer a particularly natural solution to the “boundary problem,” the problem of delineating the causal and geographic boundaries of any conscious entity.
With respect to integrated information and consciousness, a key step for Tononi is recognizing that causation equals information. Tononi’s work focuses on information in the Batesonian sense as a “difference that makes a difference.” If we are considering this issue ontologically, however, we see that all differences make a difference at some level of objective reality.* Does a particular electron make a difference to large-scale climate shifts? Almost certainly no, at least not in any manner that is significant at the macro level. But a particular electron surely makes a difference to an atomic clock that includes that electron. In other words, if we view IIT as a theory that makes ontological claims about reality and consciousness as part of reality, we can’t rely on a subjective and potentially arbitrary definition of information, which is what the Batesonian definition offers.
We should, instead, realize that “information” simply refers to causation, as Tononi agrees, and that we can then use information theory to quantify causation of all types. But here’s the key difference in what I’m suggesting: In this ontological interpretation of causation and information, all things are integrated in some manner with other things in the surrounding environment because all things interact causally with other things. They wouldn’t exist if they didn’t interact causally. This interaction/integration can be calculated as some nonzero integrated information value (phi). So in this way, it seems that IIT leads necessarily to panpsychism if we recognize that all differences make a difference at some level of reality. Under this reframing, the key question with respect to consciousness is this: At what level does the locus of consciousness exist — is it low on the chain of being, such as an atom or molecule, or far higher, as in the case of humans with our highly complex consciousness?
Tononi focuses on “differences that make a difference” from the intrinsic perspective, rather than the traditional external perspective (Tononi 2012). However, it seems that we may achieve the same ends if we view information as having a dual aspect: Information is consciousness for the entity at issue (internal perspective) but is described rightly as causation for all other observers (external perspective).
My argument in the previous paragraphs goes beyond how Tononi frames IIT because he doesn’t agree that all things have some nonzero phi, and, in our conversation, he still shies away from the panpsychist label as bad politics. But it seems to me that the panpsychist consequences of the theory are incontrovertible if we accept that all differences make a difference at some ontological level. As stated, it’s simply a matter of which level of physical organization is the locus of consciousness. And if all things are integrated with other things in some manner, then all things have a nonzero phi, even if the calculated phi is, in the vast majority of the universe, negligible compared to biological life and the more obviously conscious entities like animals. Is there a nonzero phi in an atom, as Koch has suggested? Is there a nonzero phi in a macromolecule? A bacterium? It would seem that the answers to all these questions has to be yes.
This, then, is my first main gripe with Tononi’s current framing of his theory: the theory leads inevitably to panpsychism, so rather than avoid the issue why not transform what may seem to be a weakness of the theory and turn it into a strength, judo-like? Panpsychism is not so absurd anymore, with a number of respected thinkers proclaiming support for some version of panpsychism in recent years. Koch, Tononi’s regular collaborator, has come out of the closet, so to speak, as a panpsychist, so why not Tononi?
Tononi favorably discusses Bruno, Spinoza, and Leibniz, all of whom were fairly strong panpsychists of various stripes. Tononi embraces key ideas of all these thinkers, but he rejects their panpsychist ideas as being inadequate in various ways, particularly from the perspective of a quantitative framework for consciousness. Tononi does not seem to have read Whitehead, however, who falls firmly within the panpsychist tradition of Spinoza and Leibniz, butWhitehead developed a far more detailed system with respect to consciousness and its relation to the physical world. [Tononi has since informed me that he read some Whitehead in his youth but not delved deeply.] Whitehead’s system is not a quantitative framework, but it does provide strong conceptual support for a coherent panpsychist worldview. In short, Tononi’s framework could use a good dose of Whitehead.
With respect to my second gripe, the boundary problem is concerned with the mechanism/process by which each conscious entity is formed in each moment. What is the physical and psychic boundary of each conscious entity, and why?
Finding the locus of consciousness, the locus for computing phi, is the essence of the “combination problem” or “boundary problem.” Is the Internet a “mere aggregate” and more akin to a pile of sand — that is, undifferentiated, unconscious, and uninteresting from the macro level — or more akin to a human mind with a highly differentiated and interesting type of consciousness? We don’t know, but any good theory of consciousness will, at least in theory, be able to provide good answers to these questions.
Tononi’s solution points to the local maxima of integrated information as the physical process that determines the boundaries of each conscious subject. That is, whatever configuration within a given system that yields the highest integrated information will itself become the only higher-level conscious entity within that system, and it will exclude any potential constituent conscious entities. This is Tononi’s “exclusion principle,” an essential component of IIT.
I’ve tackled this problem in my own work, and, while my approach is similar in key ways to Tononi’s work, I’ve suggested that there may be a solution to the boundary problem that springs from the nature of time and natural speed limits on the flow of causation/information. I think the nature of time is integrally related to the nature of consciousness, and IIT’s current framing doesn’t seem to take this into account.
If we accept that there is a smallest unit of time — that is, that time is quantized, as matter and energy already are quantized in quantum theory — then the universe consists of a series of snapshots, the latest springing from the last in a perpetual process that is the grandest movie we can imagine. (For those readers savvy about physics who object to this serial view of time, different interpretations of relativity theory allow for a traditional notion of simultaneity.)
Just like in a movie of human provenance, the movie that is the entire universe unfolds, in my suggested ontology, one frame at a time. Each corner of the universe is informed by the causal/informational pathways available to it from the immediately prior snapshot. In this view, the universe is nothing but causation/information. But there is a limit to the speed of such causal/informational pathways, and it is this background speed limit that produces the boundary of each conscious entity. That is, each conscious entity can only be a product of the causal/informational flows it can receive within the speed limits that nature imposes. Biological life has allowed for the leveraging of these causal/informational flows in such a way that conscious entities can exist at a far higher scale of consciousness and complexity than would be the case without such leveraging.
In sum, Tononi’s new book is an interesting read, offering poetic prose at times and shedding light on one of the key debates of our time: What is the nature of consciousness? It seems clear that Tononi has already secured himself a place in the modern pantheon of the philosophy of mind. Time will tell, however, whether a paradigm shift truly is underway. Nevertheless, Tononi’s book, as with IIT more generally, feels like it would have benefited from further fleshing out of key ideas and descriptions.
A final personal note, while not often mentioned in academic discussions, is that I couldn’t avoid, during my time with Tononi, the impression of being in the presence of a great man. I was able to get a good feel for Tononi’s personality and abilities during his talks, our interview, and over dinner. Tononi is obviously incredibly intelligent, highly articulate to the point of rattling off very long numerically differentiated paragraphs about all sorts of topics, and highly accomplished in his two main fields (sleep science and consciousness). But he is also highly affable, likeable, and egalitarian. I place great stock in these latter qualities.
* Tononi is making ontological claims with IIT: “Ontologically, the information postulate claims that, from the intrinsic perspective of a system, only differences that make a difference within the system exist.” Tononi 2012, p. 297.
Hunt, Tam. “Kicking the Psychophysical Laws Into Gear: A New Approach to the Combination Problem.” Journal of Consciousness Studies, 18, No. 11–12, 2011, pp. 96–134.
Tononi, Giulio. Phi: A Voyage from the Brain to the Soul. New York: Pantheon, 2012. Print.
Tononi, Giulio. “Integrated information theory of consciousness: an updated account.” Archives Italiennes de Biologie, 150: 290-326, 2012.