Embedded Brains and Conscious Experiences

Here is a thought experiment I’ve been perplexed by.1

Embed a brain within a larger neural superstructure so that each neuron of the original brain is connected not only to all of the neurons it was previously connected to, but also a large number of extraneous neurons in the superstructure. The superstructure is itself densely interconnected in some arbitrary way. Make it so that the new connections (connecting the brain to the superstructure) are mildly chancey, meaning that whether a neuron fires isn’t strictly determined by the signals it receives. Even if the neuron receives a large number of incoming signals, it has a small chance of not firing. This seems to be more or less the case for normal human neurons, but if we have any doubts about this, we can withhold those doubts from the new neural connections.

Being so enmeshed will fundamentally and radically change the original embedded brain’s dynamics. When you stimulate its visual centers in the manner the brain normally receives visual information, the neurons will not just send signals up to the later layers of the visual hierarchy, but also to the extraneous neural superstructure in which it is embedded. Given that the neural structure is tightly interconnected, and given that its neurons connect back to other neurons in the brain, the effect will be to inject noise into the rest of the brain. This is enough to make visual experiences basically impossible: when the visual centers of the embedded brain are tickled, the activity spreads out into the neural superstructure and back into later areas of the brain in a disorganized fashion a bit like a seizure.

Given that the neural connections are chancy, the new connections may fail to change any particular pattern of neural behavior over a short time. The superstructure neurons may happen not to fire no matter what signals they receive. During such a time, the superstructure would have no effect on the inner brain’s activity. Since the neural connections are chancy, it may be very unlikely that all the new connections should fail simultaneously, but it is still possible.

It makes conceptual sense to ask about what the embedded brain would experience (if anything) during such a time of unlikely non-interference from the superstructure when it was receiving normal perceptual information to its sensory centers. Suppose that we provided it with the appropriate stimulation akin to looking at a red apple. The neurons within the original brain might display the same pattern of firing of the same brain if not embedded. This would proceed throughout the whole brain, giving rise to the neural equivalence of self-reflective thoughts about the experience.

What, if anything, would the embedded brain feel? Would it seem to it just as if it had the experiences of a red apple? Would it instead feel something else? Would it feel nothing at all?

This thought experiment is interesting in part because intuitions pull in a different direction from what I take the major going theories to predict. Most developed theories, insofar as they predict anything, seem to predict that the brain would have no conscious experiences at all.

The Integrated Information Theory is not a theory I find particularly compelling, but it does a good job of illustrating the problems. IIT holds that consciousness arises from the integrated information within a system. Conscious experience requires a cohesive, unified informational state across a complex network. Under this framework, the neural superstructure would be part of the relevant network, and would disrupt the brain's informational integration by introducing external pathways that fragment what needs to be a unified conscious state. The random neural connections scatter information beyond the brain's original boundaries, preventing the kind of tight informational coherence that IIT requires for consciousness. Therefore, IIT would predict that the embedded brain, despite maintaining its internal structure during periods of superstructure inactivity, would lack consciousness because it can no longer achieve the necessary level of integrated information.

Some of the more plausible developed theories rely on functionalism and functionalism depends on causal structure. The superstructure prevents the embedded brain from having the causal structure that it needs to have according to any reasonable story of what that causal structure consists of. So the embedding within the superstructure alters likely causal patterns and thereby prevents consciousness even in unusual circumstances.

Would it? I like the following idea: Anything that is an atom-for-atom duplicate of a conscious human brain over a period of time would also be conscious, even if you add other stuff around it. If you take a brain and stick it in a vat instead of a body, and feed it the same input signals, it would continue to have experiences just as if it were in a body. If, instead of sticking it in a vat, you stuck it in a neural superstructure, it would continue to have experiences insofar as its internal dynamics persisted unchanged, even if this was objectively unlikely because of the new neural connections.

This idea seems to entail that functionalism is false. The embedded brain, by virtue of its new connections, might exhibit different internal dynamics. It is only because of the unlikely inactivity of the extra neural connections that the embedded brain acts like an ordinary brain.

We might try to say that the embedded human brain would still have the right structure to have conscious experiences because it does if we conceptually exclude the superstructure. The superstructure is just more neural tissue and there is no clear reason why we should separate a mass of neurons within it as consisting of its own functional system on the basis of its resemblance to an ordinary human brain. The superstructure changes the embedded brain’s dynamics even if it preserves the unlikely possibility of an unchanged pattern of behavior. So there doesn’t appear to be a principled reason for exclusion.

Perhaps we should treat every collection of neurons as being their own functional systems, so every collection of embedded neurons would be candidates for having conscious experiences, if they play the right roles only in the absence of outside (neural or otherwise) influences. They would display patterns of activity spurred on by external influence, but the subsets could still be evaluated for significance relative to their functional roles in isolation. Then we could say that the embedded brain would have the right causal dynamics for ordinary conscious experiences because it had the right dynamics-in-absence-of-the-superstructure.

This approach risks opening the floodgates to a proliferation of conscious entities. We encounter what might be called a "problem of many minds." Just as Peter Unger famously asked whether a cloud minus one atom remains a cloud and so we have many clouds whenever we have one, we may ask: if various neural subsystems within a human brain could theoretically sustain consciousness when isolated from interfering influences, should we conclude that they are all simultaneously conscious even while embedded within us?

Consider that any individual neuron in our brain could be removed without destroying our overall cognitive function. If we accept that consciousness can persist despite external interference as long as the right causal dynamics exist "in isolation," then we might be forced to conclude that countless overlapping neural configurations within each human brain could constitute separate conscious experiences. Every possible subset of consciousness-relevant neurons would represent a distinct conscious entity – the full brain minus one neuron, minus two neurons, and so forth. This threatens to populate each human head with a vast multitude of conscious beings. Given that these neurons play causal roles, we should even expect these entitites to sometimes have slightly different experiences, if enough of the relevant neurons are excluded.

I don’t see a nifty resolution to these tensions. We seem to face three possibilities: deny consciousness to the embedded brain within the silent superstructure, allow for arbitrary exclusions in causal mechanisms to define cognitive systems and thereby proliferate minds, or find some alternative to functionalism that doesn’t encounter the same problem.