Based on our third-wave framing of predictive processing, we argued in our previous post (4) that generative models cannot be unplugged from the world, given that action couples the agent to the environment. It is only via action in the world that the agent is able to tune and maintain its generative model in the way required to keep prediction error to a minimum in the long-term. In our final post we reflect on the implications of this claim for consciousness based on the specific case of culture shock.
In culture shock, subjects describe feelings of distress and alienation from the world. Take Bruce Wexler’s case of 13-year old Eva, who, along with her mother and father, left Poland for the prospects of a better life in Vancouver, Canada (Hoffman 1989, discussed in Wexler 2008). Having spent only three nights in Vancouver, she reports waking up from a dream, wondering:
“…what has happened to me in this new world? I don’t know. I don’t see what I’ve seen, don’t comprehend what’s in front of me. I’m not filled with language anymore, and I have only a memory of fullness to anguish me with the knowledge that, in this dark and empty state, I don’t really exist.” (Hoffman, 1989, p. 180; quoted in Wexler, 2008, p. 175)
Culture shock illustrates how many of the expectations a person comes to embody in their generative model have their origin outside of the individual in patterns of practices. Should the individual move to a new environment the result will be misalignment, and pervasive, hard to suppress prediction error. This is what we see with Eva. To properly explain cases such as culture shock we need to appeal to an extended dynamic singularity comprising Eva’s internal states, the patterns of practice that are enacted within her cultural niche, and her sensory and active states that couple her to the environment (i.e., her Markov blanket). To explain her current experiences one must take into account her involvement in past cultural practices, and the role of these practices in maintaining her generative model. This is because priors are tuned and updated by participating in cultural practices. But emphasising this prior learning history at the expense of action in the here-and-now would be a mistake. It is not just her past that we need to take into account but also her present circumstances, and her orientation to the future in her new cultural environment.
We take the case of culture shock to highlight that attunement to the cultural environment is an integral part of the phenomenology of our everyday conscious experiences. In culture shock this attunement to the everyday world is lost and this leads to a deep disturbance of lived experience. But this attunement or misattunement relies on the ongoing coupling of the generative model to the cultural world through cycles of perception and action. At this stage in our argument the metaphysically inclined may want to push back by pressing us to reflect on a case of neural duplicates Eva and Eva*. They will argue that Eva and Eva* could share the very same experience of culture shock just by virtue of being neural duplicates, irrespective of the environment in which they are embedded.
We suggest it is exactly this possibility of unplugging the brain of Eva from its surrounding cultural environment that the case of culture shock reveals to be implausible. To see this we can begin by asking under what conditions Eva and Eva* could be neural duplicates? Under predictive processing, we come to be the generative model we are in virtue of how we influence the world, and how the world influences us. So for Eva and Eva* to be neural duplicates they must be phenotypic duplicates – they must share the same distribution of functional and physiological states, morphological features, patterns of behavioral dispositions, and the ecological, social, cultural and material niche that makes them the kind of individuals that they are. This suggests that the mere fact that Eva and Eva* are neural duplicates is not sufficient for them to be phenomenal duplicates – they must be duplicate extended dynamic singularities. Eva and Eva* could not be phenomenal duplicates yet differ in their environmental embedding.
To see this last point, let’s assume Eva and Eva* instantiate the same generative model, and generate the same predictions. But now imagine that Eva is back home in Poland while Eva* is living her new life in Vancouver, and experiencing culture shock. What explains Eva*’s experience of culture shock? The difference-maker we suggest is cultural practice. It is the cultural world that explains why Eva and Eva* could be generative duplicates and yet still differ in their phenomenal experience at a specific moment in time.
A familiar objection for us is that our appeal to cultural practices, history and shared worlds just sets the stage for internal mechanisms to do the actual constitutive work in generating perceptual experience. To say otherwise our opponents object would be to confuse ultimate with proximate causes, or causation with constitution. This is the so-called causal-constitutive fallacy reformatted for third-wave extended mind. The causal-constitutive fallacy however relies on there being a fixed distinction between ultimate and proximate causes; or, alternatively, between causation and constitution. We think this is the wrong way to think about the relation between these notions. Indeed, even if one accepts a distinction between the proximate and the ultimate, or between causation and constitution, one need not accept that this distinction picks out a division between privileged or constitutive causes at a time and merely supporting background causes over time. The case of culture shock suggests proximate explanations of how some conscious experience come about depend constitutively on ultimate causes.
Breaking with the temptation to treat proximate and ultimate causes as mutually exclusive factors in explanation implies that ultimate causes are sometimes part of the constitutive explanation of consciousness. The third-wave perspective on predictive processing entails a new metaphysics of the constitution of conscious experience as diachronic – a metaphysics of constitution that does not treat this dependence relation in purely synchronic terms. Diachronic constitution allows for a view where the agent and the wider cultural niche cannot be cleanly unplugged from one another in a way that would allow for a purely neural explanation of consciousness. In other words, conscious beings like Eva are not like Pac-Man. Conscious beings cannot throw away the world and rely wholly and exhaustively on their brains for the construction of their experiences. Conscious beings cannot be unplugged from the extended dynamic singularity that forms in the agent’s coupling with the world because conscious beings are extended dynamic singularities.
References
Hoffman, E. (1989). Lost in Translation. New York: Penguin Books.
Wexler, B. (2008). Brain and Culture: Neurobiology, Ideology, and Social Change. Cambridge, MA: The MIT Press.
Hi Julian, Michael–many congratulations on the book
Just a quick question here. If I understand your positive proposal, it relies on a substantial notion of ‘coupling’; this is needed to undermine the claim that the Markov blanket stops at sensory/action interfaces. Thus the substantial claim at the top of this piece: “It is only via action in the world that the agent is able to tune and maintain its generative model in the way required to keep prediction error to a minimum in the long-term.”
Yet I’m not sure what the claim of coupling is supposed to amount to. I understand there’s a history in this debate of suggesting that coupling means constitution; but there’s an equally rich tradition arguing that coupling just means causal-statistical inference. If coupling is understood more in terms of the latter, then I’m not sure I see how some of the strong claims made go through.
So here’s the question! Could you say a little bit about what coupling amounts to? -and if it is supposed to ground a claim of constitutiveness in some way, could you say a little about the metaphysical dependency relationship at stake?
Hi Andrew, Nice of you to stop by, and many thanks for raising this question! You’re entirely correct to say that the notion of coupling is often invoked in these kinds of discussions, and certainly in the more general literature on 4E cognition.
We approach the notion of coupling from a number of different perspectives, both here over these blog posts and in the book. We cast the notion of coupling is in terms familiar to those working within predictive processing, by drawing a distinction between the generative model and the generative process. The latter is what does the work of coupling the generative model to the outside world. This is sometimes thought to be akin to the perception-action cycle foregrounding the importance of action for perception. The generative model, on the other hand, is a probabilistic model of the generative process – i.e., a statistical mapping from causes to observations conditioned on action.
We also think of coupling in temporal terms. It’s pretty standard to draw a sharp division between ultimate causes and proximate causes, as the former maps onto diachronic explanation and the latter onto synchronic explanation. We think this fixed and sharp distinction is problematic, given that it excludes such things as cultural practices from playing a substantial role in the generation and maintenance of perceptual experience. We argue that sometimes embedding ultimate causes within proximate explanations is needed to properly explain the phenomenon of interest.
How would we handle the causal-constitutive distinction? We try to dismantle it. The
causal- constitutive fallacy turns on causation and constitution being two entirely distinct relations of dependence. Causation is temporal or diachronic. Constitution is atemporal or synchronic. We argue that this sharp distinction is, well, too sharp. Synchronic constitution works well in the domain of material constitution. It does not really shoehorn all that nicely in dynamical domains such as cognitive science. We suggest that an argument for extended consciousness requires a diachronic notion of constitution. This notion blurs any sharp lines between causation and constitution. Specifically, it suggests that a species of constitution is a species of causation, especially a nonlinear and dynamical notion of causation.
Diachronic constitution thus shows why the causal-constitutive fallacy is not an objection against our account: the CC fallacy turns on there being a clear and sharp distinction between causation and constitution. We show that there is no such clear
and sharp distinction in dynamical sciences like cognitive science.
Finally, we add to all of this a story about a phenomenology of attunement. Something we’re trying to bring out in the case of culture shock where disattunement to the environment results in feelings of alienation, for example.
Hello Julian and Michael! I’ve really enjoyed this thread. Congrats on the book!
I’m wondering whether you thought about an error theory at all? That is, if our minds are so radically externally and socially constituted, why do we tend to be so wrong about this? Why are internalist, Cartesian intuitions so compelling?
I suppose this really involves two questions:(1) is internalism truly a kind of default self-understanding – a part of the deep, unreflective “manifest image”, or is it just a philosophical gloss to which Euro-American academic philosophers are prone – part of what Dennett calls the “folk ideology of the manifest image” – something like our reflective, enculturated self-understanding, not our more basic, unreflective self-understanding? Relevant here, I suppose, are cross-cultural studies. (2) in either case, what explains its apparent persistence, at least in some communities, despite its (on your view) radical falsity?