Monday 2 July 2012

Shimon Edelman: Being in Time


Abstract: Philosophical and computational considerations, along with neurobiological data, suggest that phenomenal experience is holistic in the sense that it emerges from the dynamics of the entire brain. On this account, your experience of the page in front of you (say) is predicated upon coordinated activity, not just of visual areas alone, but of the rest of your brain as well. Experience thus must be inherently temporally extended, if only because coordination requires time. What is the nature of this coordination and how much time does it take for experience to emerge? Lessons from the science of parallel distributed computation suggest that putting experience -- or, for that matter, any other collective action such as decision making -- on hold until after all of the brain's constituents have had a chance to reach a consensus about it is a recipe for permanent functional paralysis.  To understand why the brain does not have to wait for long (let alone indefinitely) to figure out what experience it is having, we must note that coordination, like experience that emerges, is an ongoing endogenous process modulated by input, rather than a transient ripple in an otherwise quiescent medium. Thus, the input-influenced present turn of the system's trajectory through the activation space -- the embodiment of experience -- is shaped collectively by the system's history, which likely possesses a variety of natural time scales amenable to empirical investigation.

Fekete, T., and S. Edelman, Towards a computational theory of experience, Consciousness and Cognition 20, 807-827 (2011).http://kybele.psych.cornell.edu/%7Eedelman/Fekete-Edelman-ConCog11-in-press.pdf
      Edelman, S., and T. Fekete, Being in Time, extended abstract for the poster presented at the 15th meeting of the Association for Scientific Study of Consciousness (ASSC15), June 2011, Kyoto, Japan.

Comments invited

14 comments:

  1. DANCING IS JUST SUM OF DOINGS -- FEELING IS NOT

    Shimon will give a hypothetical model in which every discriminable JND (Just Noticeable Difference) in feeling space has a counterpart in neural doing space.

    (Shimon is a brilliant theoretician, and similarity is his specialty, to which he has made very important contributions.)

    Trouble is that JNDs (psychophysical judgments) are doings. So are neural doings. So a perfect isomorphism between JND doings and neural doings is just doings/doings.

    JNDs also feel like something. Shimon's correlation holds there too.

    But the hard problem, which Shimon misses, is how and why JND doings feel like anything at all.

    ReplyDelete
    Replies
    1. SHIMON EDELMAN:
      (Tried to reply to Stevan's comments on my talk on the blog, but it refuses to accept my text, so here it is.)

      Stevan, thanks for the compliment. I must, however, repeat the request that I made on your FB wall the other day: please stop just rehearsing the "feeling/doing" slogan and start addressing the technical details of the explanations of one in terms of the other that are on the table (including mine).

      STEVAN HARNAD:
      Feeling/doing is not a slogan. It is a pair of (non-empty) categories, perfectly well-defined. I would be happy to address the technical details of one in terms of the other, but all you have put on the table is doing, waving away feeling with the sound of one sound clapping (which is therefore exactly what I am hearing!)...

      SHIMON EDELMAN:
      Instead of the Turing Test, the Shylock Test: "If you prick us, do we not bleed [or, in the case of Aplysia, ink]? if you tickle us, do we not laugh? if you poison us, do we not die? and if you wrong us, shall we not revenge?" Etc., etc.

      STEVAN HARNAD:
      Meaning what?

      SHIMON EDELMAN:
      The above is just a tongue-in-cheek take on the role of empathy in attributing phenomenality — an issue that came up today in the discussion.

      As we tried to explain in our C&C paper, the more the space of trajectories of a system has a parametrically scale-persistent, intrinsically nontrivial "(w)holey" structure, the more its doings are feelings. This hypothesis explains a bunch of empirical findings, and it also satisfies a bunch of prior logical constraints (such as being intrinsic and computable), again as we discuss in the C&C paper.

      STEVAN HARNAD:
      Shimon, I'd be so grateful if you could explain to me how "the more the space of trajectories of a system has a parametrically scale-persistent, intrinsically nontrivial '(w)holey' structure, the more its doings are feelings." With 6 papers (including yours) to follow, plus dozens and dozens of real-time commentaries to monitor and respond to, it would be a real help if you could explain this intriguing transition to me!

      I am sure the space of trajectories of a system with parametrically scale-persistent, intrinsically nontrivial '(w)holey' structure will prove to be as original and fruitful as your prior work; it's how its doings become feelings that I'm especially interested in.

      SHIMON EDELMAN:
      As to the call for a simplified explanation of our C&C paper: in a nutshell, it's an explanatory move that claims identity between feelings and the dynamics of systems that meet certain mathematically defined criteria. If you buy the in-principle validity of such a move, the rest is details, which I can try to explain. If you don't, this saves me the need to spell out the details :-)

      STEVAN HARNAD:
      If I could buy the in-principle validity of an explanatory move -- by someone whose work I admire -- that claims identity between feelings and the dynamics of systems that meet certain mathematically defined criteria, I would. But unfortunately I can't. So there is indeed no need to spell out the details.

      However, this is not about persuading me, one way or the other. You can be sure that If I wish anyone to succeed in solving the hard problem, I certainly wish it were you, Shimon!

      Delete
    2. FROM SHIMON EDELMAN: Tried to respond to the latest comment on my talk on the Blogspot site, but it keeps rejecting me (even though I have both a Google ID and a Blogspot one). So, here's my comment, in response to Carey YL Huh:

      I said in my talk that the total state of a distributed system is a fiction that can only exist for an external observer, for the simple and incontrovertible reason that instantaneous communication is impossible (as per Einstein's Special Relativity, let alone in a system such as the brain, where transmission speeds are far, far below the speed of light). Wolf Singer's data pertain, of course, to just such an external observer of the system, which, moreover, can only formulate its conclusion (that two events happened simultaneously) post hoc, not in real time. It must also be noted that the mechanisms for neural coordination that appears to an outsider to be "zero-lag" are not at all mysterious (nor do they violate Special Relativity): if two neurons fire at the same time, it is because they are driven by a third one (via equal-delay axons, or whatever). In any case, as Wolf would agree, I think (we discussed these matters briefly), his data do not contradict my view of the role of dynamics in generating intrinsically structured representation spaces, and thereby what the system in question experiences as feelings.

      Delete
  2. FEELING IN TIME

    The need to integrate internal activities across time, in relation to external activities, is already there in today's robots.

    So time-integration explains how/why those robots feel just as well (or badly) as it explains how/why we feel.

    (And those robots don't feel...)

    ReplyDelete
  3. I was thinking about the relation between feeling and doing, when it occurred to me that perhaps feeling is a kind of doing. Indeed, as Doctor Edelman has suggested, perhaps feeling really amounts to the temporal, dynamical dimension of a kind of doing. As a materialist, I would assume that feeling requires a neurological (or at the very least material) basis: only physically implemented systems can have feelings. Now, even if we cannot determine precisely what function is accomplished by feeling, and a fortiori cannot define the exact causal mechanism by which feeling is realized, it remains that feeling must be somehow related to the material substrate.

    Evolution is something organism do—they adapt to their environments and evolve (albeit over the course of a few generations). Functions are obviously something organisms do—in fact, cognitive functions are defined as just that, as what a system can do in terms of information processing and coordinating inputs and outputs. If we are indeed investigating the evolution and function of consciousness, if we go so far as to speculate that perhaps consciousness has an evolutionary, adaptive function, then it seems to me natural to suppose that perhaps feelings are functional in the sense that they are something an organism does.

    And hence, maybe the dichotomy doing—feeling is misleading because feeling is precisely something we do. F. J. Varela once said that “living is sense-making,” by which he meant that the generation of a phenomenal representation is something that organisms actively do. Perhaps, in thinking of experience as non-functional feeling, we are being misled.

    ReplyDelete
    Replies
    1. In the end, is the dichotomy of feeling/doing a cartesian relic ?

      Delete
    2. Well said. I am beginning to think it very well may be such a dualistic artefact.
      And, perhaps unlike Doctor Harnad, I do not think that adopting a such non-dichotomous position (that is, supposing that “feeling” is a kind of “doing”) amounts to merely dismissing the question.

      Delete
    3. Actions are driven by intentions. But when someone argues that feeling is a kind of action, I don't know which kind of intentions is involved (and furthermore I can feel without performing exploratory movements...).

      Delete
  4. After Dr. S. Edelman's talk, which neural activities make it into phenomenal experience? I get that silent neurons contribute to the 'holey' trajectories, but what about inhibitory neurons which are often quite active during brain rhythms etc.? Do cells have to be computing cells to be part of consciousness or could they also be modulatory ones too?

    Dr. S. Edelman's reply:
    First off, everything that makes up a brain computes (have a look at my "Computing the Mind", Oxford U Press, 2008). The rest of the question is problematic: cut out part of the brain, and the functioning of the rest will change, including the phenomenology that it generates.

    Carey Huh:
    Thank you Dr. Edelman. I also would like to know whether you equate the neural correlates you refer to with phenomenal experience (___is experience) or rather you are saying that the neural correlate generates it (___ generates experience). The latter implies a causal relationship and could we have even a delay between the two processes (correlate occurs before the experience, though I am not sure if we are able to measure this delay?).

    Dr. S. Edelman's reply:
    The former (which means that metaphysically I am fully committed to the identity stance; I explain my view of the role of metaphysics in science here http://kybele.psych.cornell.edu/%7Eedelman/Edelman-on-Shanahan.pdf and here http://kybele.psych.cornell.edu/~edelman/Edelman-six-challenges-FTPP12.pdf )

    ReplyDelete
  5. Dr. S. Edelman explicitly stated in his talk that there is no such thing as zero lag. I wonder what he means by this, since Dr. Singer described at least 3 mechanisms for zero lag in neural synchrony during his talk?

    ReplyDelete
    Replies
    1. SHIMON EDELMAN: Tried to respond to the latest comment on my talk on the Blogspot site, but it keeps rejecting me (even though I have both a Google ID and a Blogspot one). So, here's my comment, in response to Carey YL Huh:

      I said in my talk that the total state of a distributed system is a fiction that can only exist for an external observer, for the simple and incontrovertible reason that instantaneous communication is impossible (as per Einstein's Special Relativity, let alone in a system such as the brain, where transmission speeds are far, far below the speed of light). Wolf Singer's data pertain, of course, to just such an external observer of the system, which, moreover, can only formulate its conclusion (that two events happened simultaneously) post hoc, not in real time. It must also be noted that the mechanisms for neural coordination that appears to an outsider to be "zero-lag" are not at all mysterious (nor do they violate Special Relativity): if two neurons fire at the same time, it is because they are driven by a third one (via equal-delay axons, or whatever). In any case, as Wolf would agree, I think (we discussed these matters briefly), his data do not contradict my view of the role of dynamics in generating intrinsically structured representation spaces, and thereby what the system in question experiences as feelings.

      Delete
  6. (Originally posted on FB. I am reposting it here only for credits. My question has already been answered.) Dr. Edelman said at one point in his lecture that he abandoned the identity theory (as defended by people like J.J.C. Smart among others) to embrace a theory inspired by dynamics system theory that considers states of the whole brain over time. But I am not sure that identity theorists (or token identity theorists anyway) would really deny that. Dr. Edelman’s proposal to consider the dynamics of the whole brain over time is not really against the spirit of the identity theory. Sure, identity theorists often used the claim that "C-fibre firing = pain" as an example of the type of identity they had in mind. But as long conscious mental states are identified with brain states (be they states of the whole brain over time or brain states localized in space and time), then his position is compatible with the identity theory. Have I misunderstood him?

    ReplyDelete
  7. Even though the activity of the brain is temporally distributed, we should not forget that experimental protocol aims at studying clear transitions between brain states. Usually an experimental protocol will have a initiation phase, during which we clear any previously ongoing computation/brain states, so that the brain is in a "known" state or fixed initial condition. Then we proceed with the execution of the task that will bring the brain from it's initial condition to a second state. It is during the transition period that most of the scientist record their data (as far as I know). Further, most of them don't neglect temporal information... Not a critic on the talk, just a comment.

    ReplyDelete
    Replies
    1. I think that Edelman's point was not about that kind of temporal resolution. You talk about changes that occur in the length of an experiment, while (as I understand it at least) his view requires to examine the changing state of the brain at a finer level, as it unravels through time. Again, from what I get, it is this ever changing state that could be key in solving the issue we are debating in this institute.

      Delete