Does the "gradual brain replacement" thought experiment prove consciousness is independent of substrate?

Question

In "gradual brain replacement," a person's neurons are gradually, over the course of perhaps a few minutes or hours, replaced by nanites, while the person is conscious and talking about the experience.

The nanites are close enough in function to the neurons that the overall behavior and patterns of brain activity of the person are not obviously changed throughout the replacement process. They continue to act and talk like they normally would.

In particular, the nanites are close enough in function to the neurons that the brain circuits that might normally activate to detect a gradual-loss-of-consciousness scenario (e.g. circuits to detect oncoming fainting or sleepiness), do not activate; those brain circuits continue to function as if the brain still was composed entirely of neurons and nothing unusual was happening.

The question the thought experiment is designed to answer is, "does the person suddenly or gradually lose consciousness at any point during this process?"

Because of the assumptions, if you ask the person if they feel like they are losing consciousness, they will tell you, "no, I still feel like me." Functionally, they are outwardly indistinguishable from a normal person.

This is not the end of the story, because what if they are becoming a P-zombie that only acts and talks like a conscious person, but is no longer conscious on the inside? But the scenario is designed to rule this out.

First, it is implausible that the person would suddenly and completely lose consciousness, considering the gradualness of the neuron replacement. If consciousness is lost, it would have to be a gradual fading-out that the person would have time to react to and talk about if able, not a sudden switch. (The rate of neuron replacement should be tuned to a rate so that the person would have time to notice and discuss the fading-out, if and whenever it happens.)

One conceivable result is that the person would notice they are starting to lose consciousness, but simply be unable to mention it. They would lose control of their own speech; their voice would say, "I feel fine, I am perfectly conscious," even though on the inside they are trying to say they are losing consciousness and the process should be halted.

But if they had this internal struggle - if they even noticed they were fading out - wouldn't there need to be some brain circuits lighting up to represent that process of noticing? People don't just feel emotions and notice things without corresponding neural activity. But by assumption of the scenario, those brain circuits are functionally acting just the same as they were before.

To put it another way, there's a link between how the person is subjectively feeling and what they say they are feeling. If this link erodes gradually, then while it's still 75% present they would still be able to report on the 25% reduction.

So we have to conclude they did not notice any loss of consciousness throughout the whole process; throughout the process, as they inspected their state of consciousness, they subjectively seemed just the same as they were at the beginning. So that by the end, with their brain composed fully of nanites, they are still conscious. There was no sudden change in their state of consciousness because the physical replacement was gradual, and there was no gradual change in their state of consciousness because they would have spoken up about it or at least noticed it, neither of which is possible according to the premise.

Does this hold up - does this prove that a person with a brain composed of nanites with function sufficiently similar to that of biological neurons, would possess the same level of consciousness as a person with a biological brain?

---

To make explicit the baseline assumptions:

• There is no extra-physical spirit exerting a causal influence on the world; the brain's patterns of activation and the body's actions are explicable by the neural dynamics in the brain, without causal reference to any supernatural component.
• For humans with biological brains, all subjective experiences have neural correlates.
• There is nothing physically preventing the function of the brain from being re-implemented in nanites.

But I don't assume that consciousness is independent of substrate. This is the conclusion, not an assumption. The assumptions above are not prima faciae incompatible with the idea that consciousness requires a biological brain. I assume (qualia) -> (physical correlate) (that's the second assumption above), but the scenario is meant to show, rather than assume, that (functionally equivalent physical correlate) -> (qualia).

Answer 1

No thought problem provides a proof, as the thought problem has embedded assumptions which are not challenged by an actual test, where the assumptions might not hold.

The explicit assumption that is present in this thought problem is functionalist identity theory -- that if a substrate operates functionally identically to how human brains operate, that it would be identically conscious.

Note this is not the same as neural identity theory. And the logical argument against this presumption, that it is possible that one could functionally do what humans do and NOT be conscious, involves a subtlety different concept of a philosophic zombie than Chalmers' zombie.

Thought problems, to be convincing, basically rely upon concurrence from the intuitions of other philosophers. For those philosophers who are convinced of the unquestionable truth of algorithmic identity theory, this thought problem is convincing. For those who are disciples of Socrates, and seek to identify and challenge the walls of the boxes they think within, this and other thought problems are intrinsically NOT convincing.

Supplemental elaboration on the answer:

This is primarily because our world is immensely more complex than as assumed in philosophic thought problems. For example, Greek (and Chinese) "element" theory postulated four elements (5 for China). Yet when we examined matter, we found over 200, and "fire" is definitely not among them.

Medieval psychology/physiology theory postulated four "humors" that drove our health and psychology. But just for psychology, our mental world is so much more complex than ANY four features could produce -- Just look at Wikipedia's list of biases: https://en.wikipedia.org/wiki/List_of_cognitive_biases

Empirical investigation almost always finds our world is immensely more complex than any philosopher imagines while setting out a thought problem. What this boils down to is generally that the thought problem has unidentified assumptions in it, that upon investigation do not actually hold in our world.

supplemental comments on specific assumptions

The OP amended the question, to add a list of specific assumptions. I will extend the answer to comment on them.

There is no extra-physical spirit exerting a causal influence on the world; the brain's patterns of activation and the body's actions are explicable by the neural dynamics in the brain, without causal reference to any supernatural component.

This is basically an assertion of causal closure of the physical. There are logic based reasons to be highly suspicious of causal closure of physics, including Hempel's Dilemma, and the inability of physics (or Science, or empiricism generally) to justify itself, so this is a logically suspect assumption. Additionally, a short list of active approaches within philosophy to try to explain minds, includes multiple approaches that do not share this assumption: Is it possible for there to be an AI chatbot that is a philosophical non-zombie?

From this list, the stronger versions of emergent physicalism, algorithmic identity theory, emergent algorithmic identity, most neutral monism, idealism, and all interactive dualisms do not share this assumption.

For humans with biological brains, all subjective experiences have neural correlates.

From my prior list, fewer philosophy of mind research programmes would disagree with this than with causal closure, as most algorithmic identity thinking plus emergent physicalism would agree with this assumption, as would a lot of neutral monists, and even a fair number of of idealists. However, the physicalist "new mysterians" would disagree, so one would lose a faction of physicalists.

There is nothing physically preventing the function of the brain from being re-implemented in nanites.

This assumption would lose John Searle, and any other physicalist who holds that there is some unique biological rather than functional feature to consciousness.

In addition to this list, two assumptions in the original text are not universally accepted, as pointed out in Nir's rebuttal paper to Chalmers:

First, it is implausible that the person would suddenly and completely lose consciousness, considering the gradualness of the neuron replacement. If consciousness is lost, it would have to be a gradual fading-out that the person would have time to react to and talk about if able, not a sudden switch.

Phase changes happen in our world. Denial that they are "plausible" relative to consciousness is just an assumption, not proven, and is essential to make this thought problem go.

One conceivable result is that the person would notice they are starting to lose consciousness, but simply be unable to mention it. They would lose control of their own speech; their voice would say, "I feel fine, I am perfectly conscious," even though on the inside they are trying to say they are losing consciousness and the process should be halted.

It is also logically possible that one would NOT notice, as self-introspection degrades along with the fading consciousness.

It is also useful to ask what a failed test would tell the OP? What conclusion should the OP draw if the subject describes consciousness fading? That nanites are unable to replicate consciousness? OR would the OP declare that they must have been mistaken and THESE nanites must not have completed their development, so the hypothesis did not fail? And instead some additional functionality should be added to nanites 2.0, or 3.0 etc. until the test is passed?

Answer 2

This appears to be a rephrasing of Chalmers' "Absent Qualia, Fading Qualia, Dancing Qualia" (1996) paper. Given your 12.5K score on Philosophy Stack Exchange, I presume you are familiar with it.

I have previously written a short unpublished paper titled "Yet Another Objection to Fading and Dancing Qualia" that, in my view, dismantles these thought experiments.

Here is a simple point regarding why a sudden change can occur. You write:

First, it is implausible that a person would suddenly and completely lose consciousness, given the gradual nature of neuron replacement. If consciousness is lost, it would likely fade out gradually, allowing the person time to react and discuss the experience, rather than switching off abruptly.

However, consider a one-of-a-kind ancient and priceless Chinese vase, floating a meter above the floor, suspended by a rope. Imagine cutting the fibers of the rope one by one.

We know we need at least 95% of the fibers intact to support the vase's weight. Thus, we can be certain that, at some point, the rope will break, the vase will fall and shatter, and you will be held legally responsible for its negligent destruction.

Would you argue in court that you did not anticipate the rope breaking suddenly because you had logically ruled out that possibility through armchair philosophy?

A similar argument can be made against the idea that neurons can be replaced one by one with units that perfectly mimic them. I am willing to expand on this point should there be interest, but for now, the above example suffices.

Answer 3

Neurons are examples of nanites themselves. They are tiny physical entities obeying the laws of physics. If the subject should report any weird sensations, you would have to conclude that your nanites are not perfectly identical (in their physical function) to the neurons. The mind/matter mystery is already present without introducing nanites; the nanite narrative is simply an aid to stress the physicality of the brain.

Answer 4

I've encountered this idea. I've also encountered its inverse. In essence, this is the same mistake as the philosopher zombie mistake; the result is assumed.

However this time it is theoretically a testable hypotheses; that deserves the name forbidden experiment (which is already taken but I digress...)

You might predict the consciousness remains intact as you replace the neurons is consciousness remains. You might predict consciousness is lost and we can't tell the difference. You might predict consciousness is lost and we can tell the difference. I consider the second hypothesis untenable. With a mind of the subject intent on running the experiment and not hiding the result this should not happen; therefore I discard it. If we were really doing this I would take note that our test could be fooled, but we're not going to do this.

However there is a fourth case not easily guessed. Consciousness might be retained and only when you copy the mind state you find out the copy isn't conscious. Designing a test for that even assuming a cooperating mind is no mean feat. The spectre of this result hangs over your thought experiment; resulting in it proving nothing at all. Too bad.

I actually predict the worst possible outcome; coping appears to retain consciousness for some individuals and loses it for others. It's difficult to reason over the material; but I bid you consider that I actually came up with this prediction on my own, and it is not of wanting to hide from the truth I came up with it, but it actually makes sense to me this would be divergent here.

Answer 5

First, I want to say that the thing you want to conclude is something I agree with, i.e. that consciousness is substrate-independent. However, the question isn't whether this is true but whether the argument proves it, and I think it doesn't - there are definitely reasonable objections that someone could make. For example:

1. The argument assumes that it is possible to replace someone's neurons with nanites without affecting their behaviour. I believe this to be true (in principle at least) but I have definitely encountered the view that it is not. That is, one can reasonably believe that consciousness is a fundamentally biological phenomenon, to the extent that any artificial system, no matter how it's constructed, will not be able to precisely replicate the behaviour of a conscious system - it might get close but, on this view, it would always be possible to tell the difference.

2. The argument presents one apparently-possible way in which someone could gradually lose phenomenal consciousness while behaving as if they still have it, namely an "internal struggle" in which they notice their loss of consciousness but are unable to say so. But, for someone who thinks phenomenal consciousness is substrate-dependent, this might not be the only possibility. Consciousness could just gradually "fade out", for example. The experience might be similar to being half-asleep, vaguely aware that words are coming out of your mouth but without any great attention paid to what they mean and no awareness that anything is wrong. This lack of attention wouldn't be detectable by any circuits failing to light up - the circuits are still there and still lighting up, they're just made of nanites and hence not contributing to consciousness.

3. Another view I've encountered before is that a brain made of nanites would still have consciousness, it's just that its content would be different. This is still a form of substrate-dependence, even if it doesn't claim a zombie would be created in this experiment. On this view, I suppose that the subject's qualia would gradually change into different ones, with their memory changing to match, so as not to affect their behaviour. When asked to describe the redness of an apple they will still say the same things as they did before the treatment, but the underlying quale - the part of the experience that can't be expressed in language - will be different.

There might be other reasonable objections that can be made. Overall, as with most thought experiments I think it's a good argument for sharpening intuitions if you're predisposed to agree with its premises, but it's hard to call it a proof.

Answer 6

Your assumptions are very strong and essentially affect the outcome.

Thomas Breuer has shown that from the point of view of any observer, a system properly containing himself (such as his brain) does not follow the same physical laws as the rest of universe, due to self-reference. This is a fundamental, mathematical result. It means that such system has physically indistinguishable states that nevertheless affect future. So, the wavefunction is not usual, and Breuer calls it "subjective decoherence".

Now, let's look if your assumptions conform to the result by Breuer.

• There is no extra-physical spirit exerting a causal influence on the world; the brain's patterns of activation and the body's actions are explicable by the neural dynamics in the brain, without causal reference to any supernatural component.

This manifestly contradicts the result by Breuer. There are physically-indistinguishable states that influence physical outcomes.

• For humans with biological brains, all subjective experiences have neural correlates.

I do not see this directly contradicting Breuer's result.

• There is nothing physically preventing the function of the brain from being re-implemented in nanites.

Hmmm. This depends. Whose brain we are talking about? The one of the observer or of a random person?

And what is meant by "re-implemented"? Does it mean a complete quantum-mathematical copy can be created with all the same states? Or just a machine capable of imitating brain functionality?

Due to no-cloning theorem, a complete physical copy of a quantum system is impossible, but a new system can be a copy of the old one if the old one is destroyed.

---

So, to sum it up, your assumptions contradict the case where the brain belongs to observer himself due to self-reference. If your assumptions hold, the observer is outside of the experiment and only watches it from the distance. In this case it is completely possible that the subject of the experiment is a philosophical zombie from the beginning, before any transformation. The patient will report that there is no loss of consciousness, etc.

Now, what happens if the patient of the transformation is the observer himself? This is more interesting, though your assumprions are violated.

But it is difficult to say what would be the outcome and how the flow of qualia would go on. A factor of quantum immortality also would play a role.

I can only speculate that:

• If the brain is classically scanned and copies of the parts of brain are created step-by-step, the result will not be the original observer, because you cannot copy the quantum state by classical scanning like X-ray. You cannot copy indistinguishable states this way either. The experiment is equivalent to a machine-assisted suicide, although bearing in mind quantum immortality, the outcome is unpredictable.

• If the quantum states of brain are transferred into nanites using quantum teleportation, the result should be technically the same as when you eat food and particles in your brain get replaced with other particles over years (in 7 years your body has most of the atoms replaced).

So, short summary:

• The person is not the observer, classical information transfer: the result will be visibly successful, but some quantum correlations may be lost.

• The person is not the observer, quantum information transfer: the result is absolutely the same person.

• The person is the obsever, classical information transfer: machine-assisted suicide (possibly, unsuccessful due to quantum immortality).

• The person is the observer, quantum information transfer: successful quantum digital transformation.

Answer 7

I would say that the assumptions under which the experiment operates prevent it from proving independence of substrate, core to the experiment is the assumption that this:

The nanites are close enough in function to the neurons that the overall behavior and patterns of brain activity of the person are not obviously changed throughout the replacement process. They continue to act and talk like they normally would.

In particular, the nanites are close enough in function to the neurons that the brain circuits that might normally activate to detect a gradual-loss-of-consciousness scenario (e.g. circuits to detect oncoming fainting or sleepiness), do not activate; those brain circuits continue to function as if the brain still was composed entirely of neurons and nothing unusual was happening.

is possible. But this already assumes that there is a substrate which is sufficiently similar to the human brain that consciousness continues; and so all it can ever demonstrate is that if such a substrate exists then you can switch to that substrate without removing consciousness.

It does this in two ways: firstly it assumes that there will be no change to "the behaviour and patterns of brain activity" which, imo, already assumes that the new substrate allows consciousness (even if you believe a P-zombie is possible, the brain activity of a being lying about having consciousness is different from that of a being actually experiencing it). This is then confirmed by the next paragraph which guarantees the conclusion by stating that there can be no gradual loss of consciousness.

Answer 8

"If they even noticed they were fading out [and becoming a P-zombie], we would be able to detect those thoughts" seems incompatible with the other assumptions.

Suppose: There is no extra-physical spirit exerting a causal influence on the world; the brain's patterns of activation and the body's actions are explicable by the neural dynamics in the brain, without causal reference to any supernatural component.

The construction of this assumption prohibits us from distinguishing a P-zombie from a person with a mind. Suppose our subject says or thinks "Wait! My mouth said I felt normal, but I feel like I'm slipping away!" Either this is

• A medical problem with the nanites causing physical thoughts in the brain, causing them to feel faint, etc. We have assumed this not to be the case because our nanites replicate the neurons perfectly.
• A purely mental experience (that of being P-zombified) which has, just now, exerted a causal influence on the world.

Answer 9

I think this is a straightforward question. Can we copy/replicate conciseness?

People were able to replicate fire, then lightning. At some point, people will replicate the conciseness to some degree.

Would it get us closer to answering the question about the origins of life and self-aware conciseness? Probably not.

Answer 10

I am also an armchair philosopher that has always loved this problem, and I think there is a big problem, even with your assumptions.

Go to the end if you want to avoid the list of problems even with my solution.< I'm new here (by some definitions) so there's probably some convention that I don't know.

Let's clarify a couple (perhaps additional) assumptions: -The person is awake. Without this, the process could be functionally identical to just instantly freezing a person (to absolute zero) and painstakingly copying everything about every neuron. Hopefully this would satisfy the whole no-cloning theorem issue as a method for non-gradual brain cloning. -The neuron-to-nanite conversion is not truly instant. There is a limit to the speed with which you can do anything. My solution might fall apart though if you also gradually replace each neuron and replace the functions of subsections of neurons with the nanites. -Our goal is to perform this process so that there would be no (perceptible) difference at all between the person and their actions in the timelines where neurons were replaced and where they weren't replaced. If there is a difference, this could indicate a modification of consciousness. This point is very much debatable, though I believe that this in some way mean a loss of free will (I'm questioning whether we have it anyways despite this).

Now for my solution: If it takes any amount of time for each conversion to take place, it means a nanite will be "desynced" from the neurons around it upon the initial replacement. In order to catch up, it will probably either a) Have to guess what it was supposed to be doing by watching the behavior of surrounding neurons. b) Watch the neuron's behavior before replacing it and use the information to train another neural network to use immediately after the replacement process. Either way, it will HAVE to approximate a neuron, at least initially. After all, a guess is only an approximation, and a neural network just approximates a function. And it will have to undergo this process, since, after all, the person is conscious and their neurons are constantly doing SOMETHING that builds their whole consciousness. Even if the nanite is FUNCTIONALLY identical, it will be difficult to guess what its initial state should be. In a real world experiment, this will have to encounter error. Maybe there won't be enough error to damage consciousness until like a couple percent, since we can survive concussions without even blacking out some of the time. This error will have to build up, though, and our final result (though calling them a result may be quite de personifying), even if .001% off, will, according to chaos theory, quickly become wayyy off from what the original would have done in any situation, indicating a modification of their free will. Maybe, in the future, if replacing your brain with nanites so that you could upload it easier or change your perception of time becomes an industry, there would be an arms race over who could get the least final error.

Sorry if there are any spelling mistakes I wrote this on a phone at 4am when I decided to check my Google news feed in the middle of the night.

Answer 11

It might be the case that to create "sufficiently good nanites that can 100% replace the brain without the person noticing" would end up requiring your nanites to be in some critical way identical to normal biological cells (I mean in this in a ridiculous sense, i.e. needs to use carbon, needs to have fat exterior, needs to have DNA etc...)

Now I don't personally believe the above is true, but IF IT WERE, (and it could be!) then we certainly would be forced to conclude "the human conscious experiment is CERTAINLY tied to the substrate".

The Gradual Brain Replacement thought experiment uses "we could replace neurons with sufficiently different system X" to make its point. We actually don't know how different "X" can be to make this work. If physics and chemistry is such that "X" is constrained to be almost identical to what we call a neuron today, then the SAME Gradual Brain Replacement thought experiment actually proves that consciousness is very much tied to the substrate.

Answer 12

As I understand the role of nanites in your thought experiment is to replace biological neurons by artifical neurons.

Your thought experiment is similar - but much more complex and presumably not practicable due to the complexity of the nerve system of the brain - to implanting to a patient an artifical heart. I remember the time of the first heart transplantations, first with donor hearts and later with artificial hearts. There were some discussions before whether the new heart would change the consciousness of the receiver.

These discussions were triggered by the traditional view that the heart is the location of emotions and of all kinds of feelings. The transplantations show that this assumption is false. Instead, they show in a striking manner:

The function of the organ does not depend on its physical substrate.

I would extrapolate the experience from actual heart transplantations to the hypothetical transplantations of biological neuronal nets from your thought experiment.