Wednesday 4 July 2012

Michael Shadlen: Consciousness as a Decision to Engage

      Abstract: Consciousness encompasses a variety of functions and properties, such as awakening, awareness, and subjective aspects of both perception and volition (e.g., qualia and authorship, respectively). It remains to be seen whether these diverse functions are related to one another through common neural mechanisms, and if so how. Here, we advance the thesis that the neural mechanisms that give rise to conscious states share features with neural mechanisms that underlie simpler forms of decisions. The neurobiology of decision-making provides detailed insight into how the brain deliberates and reasons from evidence to make choices. The underlying mechanisms, mainly studied in animals, could support a variety of complex cognitive functions that probably operate independently of many aspects of consciousness. For example, many complex decisions in humans rely upon wakefulness but not upon awareness or authorship. In animal studies, decisions are typically embodied: they can be described as selection among possible actions. By substituting 'circuits' for 'actions' in the preceding phrase, we generalize from 'deciding to do' to 'deciding to consider' or, more generally, 'deciding to decide to. . . .' This is an appealing notion from the perspective of brain evolution, because it allows us to recognize ideation as an elaboration of a simpler sensory-motor design. We propose that many of the functions of consciousness are simply ways of engaging the environment. Thus consciousness might be mediated by (non-conscious) decisions to engage, as in awakening, or to engage in a certain way, as when attaching narrative to action. Although the neural mechanisms underlying 'decisions to engage' are unknown, they are likely to involve intralaminar (and matrix) thalamus and processes that 'decide' to turn other circuits on. This idea invites an analogy between the functions of brain regions that project to matrix thalamus, including the 'default system', and the role of parietal cortex in perceptual decisions. While highly speculative, we think 'decision to engage' provides a biologically plausible and computationally coherent hypothesis about the neural correlates of consciousness.  (From Shadlen & Kiani 2011)

      Shadlen MN, Roskies AL. 2012. The Neurobiology of Decision-Making and Responsibility: Reconciling Mechanism and Mindedness. Frontiers in Neuroscience 6
      Shadlen MN, Kiani R. 2011. Consciousness as a decision to engage. In Characterizing Consciousness: From Cognition to the Clinic? Research and Perspectives in Neurosciences, ed. S Dehaene, Y Christen, pp. 27-46. Berlin Heidelberg: Springer-Verlag
      Shadlen MN, Kiani R, Hanks TD, Churchland AK. 2008. Neurobiology of Decision Making: An Intentional Framework. In Better Than Conscious?: Decision Making, the Human Mind, and Implications for Institutions, ed. C Engel, W Singer, pp. 71-102. Cambridge: MIT Press
      Gold J, Shadlen MN. 2007. The neural basis of decision making. Annu Rev Neurosci 30:535-74
      Kiani R, Shadlen MN. 2009. Representation of confidence associated with a decision by neurons in the parietal cortex. Science 324:759-64
      Yang T, Shadlen MN. 2007. Probabilistic reasoning by neurons. Nature 447:1075-80
      Shadlen MN, Movshon JA. 1999. Synchrony unbound: a critical evaluation of the temporal binding hypothesis. Neuron 24:67-77

Comments invited


  1. This was a very informative presentation, but I wonder whether Shadlen might be making some unjustified assumptions when he jumps from claims about neuroanatomy to claims about consciousness and decision-making. In the abstract, he writes: "The neurobiology of decision-making provides detailed insight into how the brain deliberates and reasons from evidence to make choices." But, as one person noted, it is not clear that the experiments which he discussed really tell us how and where the decision was taken in the monkey's brain when it decided to go for the red or green circle. All the experiments showed was that some neurons were good indicators of the monkey's decision and were fine-tuned to choose the circle which brought a reward. But, for all we know, the decision-making process might have already started before the neurons in LIP react (or the neurons' reaction might be only the first step of the decision-making process). If that is right, then we still don't know what decision-making in this case really involves.

    1. All reasonable points. I did not tell you everything we know about the process in these tasks. The importance of the reaction time version of the task derives from the constraints it places on the estimate of decision time on single trials. Please consider those fits of the monkeys accuracy and reaction times on the random dots experiment (the slide with the equations). It produces an estimate of the time that that was used and not used to reach the decision. The latter corresponds beautifully with the physiology. That is, the stereotyped dip before the decision related activity begins (colored curves begin to separate by motion direction, strength [slide with Jamie Roitman's picture]) and the time at the end after the responses come together (the sign of termination) but before the actual eye movement. In these experiments, the agreement is unlikely to be off by more that 10-20 ms. There's much more I can say about these delays. But there are many reasons you should regard these neural responses as informative about the decision process.

      There's plenty we don't know, of course, even about these neurons. I describe the process as integration, but we don't know how that is achieve at the circuit level. I showed you evidence for a terminating bound (or threshold), but we don't know how that threshold crossing is implemented and suspect it is not in LIP (although LIP stops integrating when it occurs; so there is communication).

      I hope this helps.

  2. I enjoyed Shadlen's presentation and his incredibly impressive experimental design. Although I appreciate his straightforward approach in the search for 'neurological' consciousness- I think the claim that a philosophical consciousness can be written-off is stretching this data.
    The 'post-decision wagering' experiments give insight into the probabalistic computations a neuron performs when presented with a distinct visual task, but I do not think this represents the entire decision-making process. If this neuron is indeed the site of visual appraisal, there must additionally be an entire feedforward and feedback relation with memory/reward centers modulating the monkey's confidence index (if you will).
    Nor does this explain how, in the context of a less structured scenario (say, swinging through a tree and having to make accurate visuo-spatial analyses of potential prey), one would choose what to flag and attend to in the first place.

    1. Thanks for this comments.

      1. My point is not to write off P-con. The thesis is that it is mediated by neural mechanisms that are similar to the ones that mediate the more pedestrian N-con. One can understand both as a decision to engage. This insight is useful because, if correct, it implies we know much more about the process than one might imagine. That's because we know so much about the neurobiology of decision making.

      2. Your points about the limitations of our paradigms are well taken. For example, I did not talk about the decision to choose the sure-target on the post-decision wagering task. I showed you that the evidence for this choice is (i) the state of evidence used to make the left-right decision and (ii) elapsed time. I didn't show you all the physiology in support of this claim, but that's in the Science paper. It's an open question whether LIP or orbitofrontal cortex or ACC or vPMC or ... (or any combination) are involved in converting this evidence and elapsed time into a commitment to take the sure-target or not. We don't know yet.

      3. Re " would choose what to the first place." That's a key insight. Yes, we do not understand how context leads to a decision to engage motion in one part of visual field as something that bears on the salience of one or another target elsewhere in the visual field. This is what I refer to as the circuit selection problem. But notice the way I'm framing the problem. It sounds tractable, to me anyway, when I put it this way, which brings us back to point 1.

  3. Apart from the fact that Shadlen offers one of the most powerful and eloquent interpretation of a neural/behavioral dataset of this conference, I believe that he also brings an important distinction between two types of consciousness (i.e. Neurology-Consciousness vs. Philosophy-Consciousness) that seem to evade even some of the most prominent speakers of this school (namely Searle when he argued that he is conscious (Philosophy-conscious) because he can move his arm at will (Neurology-conscious)).

  4. Xavier Dery ‏@XavierDery

    @mahdick I don't think you can "clock" the brain that simply, as it would require a good sample of single neurons, and given the number...

    3:31 PM - 4 Jul 12 via Twitter for Android


    Richard Mah ‏@mahdick

    This might be a stupid question, but what (if it exists) is the clock rate of the brain? Can we overclock it? #TuringC #couldbegood

    3:12 PM - 4 Jul 12 via Twitter for iPad

    1. Please see Mazurek ME, Shadlen MN (2002) Limits to the temporal fidelity of cortical spike rate signals. Nat Neurosci 5:463–471.

  5. Xavier Dery ‏@XavierDery

    Shalden: Wow, seeing perception as the blind man's stick, poking the world in order to get practical answers... I'm okay with that! #TuringC

    3:49 PM - 4 Jul 12 via Twitter for Android

    1. This is from Merleau-Ponty, Phenomenology of Perception. Sure hope I said so. To me this is a guide for where to look in the brain, but it leaves open all the "how" issues, as noted by some of the commenters at the top.

  6. I think that Shalden's views can be related to MacIver's in terms of anticipation and decision making being key to consciousness... Probably to be followed in my paper for the institute!

    1. I'm only just seeing this 14 months after you posted it. But please send me your paper (shadlen at columbia dot edu).

  7. From the abstract: "While highly speculative, we think 'decision to engage' provides a biologically plausible and computationally coherent hypothesis about the neural correlates of consciousness."

    Allegedly, one of the benefits of mindfulness meditation is that the practitioner becomes less prone to (emotional or otherwise) reactivity in his/her daily life. The way this is said to happen is through a practice of deciding to engage with the world as it is, through practicing acceptance. It sounds a lot like the quote above.

    Could it be that clear perception of the world allows for more freedom in decision making? Or, in the terms we've been discussing: a better integrated and more accurate sense-system might allow for a better adjusted motor output?

    Is it conceivable that this is an ability beings develop with evolution and it is still possible to fine tune it through things like meditation?

  8. Perhaps this is a bit of a naive comment, but I really liked Dr. Shalden’s experimental design concerning the confidence rating in monkeys. What a great way to get a little more insight in their ‘decision-making’ and how they feel about the choices they make!

  9. While interacting in our day-to-day life, we need to act or react to bodily processes and the happenings in the world, sometimes instantly, to provide us beneficial outcomes.

    Consciousness is designed by the evolutionary process to allow data from such interactions that requires judgmental power to become available for making decisions, thereby benefiting from the capability of making free will decisions (If there were no free will, there was no requirement of consciousness).