Identity and the Copying of Minds

Alexander R. Pruss


Department of Philosophy

Georgetown University


November 20, 2006


Draft of Work in Progress: I am not entirely convinced by the arguments here, but I think they’re worth thinking about.


1. Introduction

            Assume for simplicity that human mental states are constituted by brain states (if dualism holds, copying of brain states may need to be replaced with copying of soul states).  According to psychological continuity theories of personal identity, if the personality and memories of a human person A were copied into the brain of B while the brain of A were destroyed, and no other copies were made, then A would survive in the B-body and would be identical with the post-operative person occupying the B-body. By considering the mechanics of copying in a case where A and B have qualitatively identical brains, I will argue that the conclusion is false, and hence the psychological continuity theories that imply this identity claim are false. 

            Psychological continuity theories typically distinguish connectedness from continuity.  Simplifying by restricting myself to the case of memory (the forward-looking case of intentions emphasized by Parfit (??ref) is analogous), if t2>t1, a person x at t1 is connected to a person y at t2 if and only if y has a first-personal quasi-memory of something that befell x at t1.  If t2<t1, we say x at t1 is connected to y at t2 if and only if y at t2 is connected to x at t1, and by default we deem x at t to be psychologically connected to x at t if x exists at t.  A quasi-memory is just like a memory, minus the requirement, if there is one(??ref), that one can only have a first-personal memory of something that one has oneself done.  Psychological continuity, then, is the transitive relation generated by psychological connectedness: x at t1 is psychologically continuous with y at t2 provided there is a chain of psychological connectedness in between.  Psychological continuity theorists think that, at least absent branching(??ref), psychological continuity is necessary and sufficient for personal identity.

            My argument will not work against a psychological continuity theorist who believes that quasi-memories need to be connected to the quasi-remembered events by a causal connection of such a sort as can only exists within a single brain.  The endorsement of mind-copying thought experiments seems to be, however, one of the main reasons for accepting psychological continuity theories over brain and bodily identity theories (??ref). 

I shall assume I am dealing with a psychological continuity theorist who thinks that a substantive connection needed between a quasi-memory and what it is the quasi-memory of.  A swamp-man in China arising coincidentally right after I perish and with apparent memories of what I did but with no causal connection to me will not count as having quasi-memories of what I did.  However, the connection between a quasi-memory and the quasi-remembered event is not so tight as to rule out mind-copying thought experiments.  This, I take it, is a description of the view of a typical psychological continuity theorist.

2. The argument

Suppose that we have a way of describing a brain state as a finite sequence of N bits, where a bit is something that can have value 0 or 1.  If brains can be described by mathematical physics, we can indeed do this, at least up to any required degree of precision, and presumably there comes a point at which we have enough precision to capture the mental state that supervenes on the brain state (e.g., in capturing the state of a digital computer, we may need not note the exact voltage level in a given area, but simply to note whether the voltage is close to the value representing zero or close to the value representing one).

Now I will describe a process whereby the data of the brains a and b of two people are copied into a single output brain c.  Call the two initial persons A and B, respectively.  We thus read in the data from the brains a and b bit by bit, then feed the corresponding bits from each of these brains into a machine that takes two bits, one from a and one from b, and outputs one bit.  We take the output bit and impose it on c, while destroying the parts of the original brains that have already been read. 

I will call the machine that computes the one output bit from the two input bits “the Combiner”.  The Combiner, I posit, has two settings, the X setting and the Y setting.  The setting is put in ahead of time, e.g., by flipping a switch or in some other way, and works as follows.  If the Combiner is given two bits that are equal, i.e., both zero or both one, then it immediately outputs the same bit value.  If the two bits are not equal, the machine checks its setting.  If the setting is X, then the bit that came in from a is put out.  If the setting is Y, then the bit that came in from b is put out.

            The following two claims seem to hold given psychological identity theory:

(1)  If the Combiner setting is X, then A but not B is identical with the post-operative person with brain c. 

(2)  If the Combiner setting is Y, then B but not A is identical with the post-operative person with brain c. 

These claims hold for any two persons A and B whose mental states supervene on brain states in such a way that the brain states can be described with N bits.

            Now consider a special case.  In this special case, the brains a and b are qualitatively identical but numerically distinct.  Maybe A and B are twins, one of whom grew up on Earth and the other on Twin-Earth, or maybe they are the result of a previous fission.  In any case, however this may have happened, A and B have qualitatively the same brain state at the beginning of the procedure while being distinct persons.

            I now claim that who survives the operation and who is identical with the c-brain person after the operation does not depend on whether the Combiner setting is X or Y.  The reason for this is simple.  Because the brains a and b have the same state at the beginning of the procedure, the bits read from each are always the same, and the Combiner simply outputs the common value.  The Combiner never checks which setting it has when it gets a pair of equal inputs.  We could imagine the Combiner’s setting being set not by flipping a switch, but by putting the inscription “X” or “Y” in a sealed envelope, with the Combiner having manipulators and sensors which would open the envelope as soon as it gets two bits that differ, so it could find out how to process them.  But the envelope is never in fact opened when the brains a and b are in the same state.  And whether a person survives and inhabits a given brain surely does not depend on what had been written in a sealed envelope whose contents in fact have no causal influence on anything outside the envelope. 

            If this is correct, then in the case at hand, the identity facts for A and B are the same regardless of the Combiner setting.  But according to (1) and (2), the identity facts for A and B differ depending on the value of the Combiner setting.  Since (1) and (2) follow from the psychological continuity theory of personal identity, it follows that the psychological continuity theory is false.

3. Objections

            I. Tu quoque.  This is a puzzling case, and like many such it will be puzzling for every theory of personal identity. 

            This objection is simply off-base.  There is nothing problematic about this case on body-identity, brain-identity, animalist or dualist views.  The first three of these views imply that (1) and (2) are both false, regardless of whether the states of a and b are the same or not.  Dualism implies that the cases are underdescribed because we failed to say what happens to the immaterial soul(??Swinburne).

            II. No-branching requirements.  Psychological identity theorists typically (??refs) make an exception for cases of branching, and say that psychological continuity is sufficient for identity absent branching.  Branching can take the form of fission or fusion.  Here we have fusion, the objection goes, and hence (1) and (2) are false.

            But we do not have fusion here.  If the Combiner has setting X, then it is a machine that copies the state of a into c, with the state of b being counterfactually irrelevant, while if it has setting Y, then it is a machine that copies the state of b into c.  In neither setting is the Combiner a machine that fuses the states.  Now it is true that in the case where a and b are qualitatively identical, the person with brain c ends up being qualitatively just like A and just like B.  But, nonetheless, brain c is a copy of a, not a copy of b, given setting X, and is a copy of b, not a copy of a, given setting Y, assuming “copy of” implies counterfactual dependence.  There is no fusion.

            III. Causal dependence.  Suppose a and b do have qualitatively the same state.  Then given setting X, the output of the Combiner has no counterfactual dependence on b.  But nonetheless it has a causal dependence on b in the sense that the machine functions differently depending on the state of b.  If the bit coming from b were not the same as the bit coming from a, then the Combiner would have to have looked at the setting, and decided in light of it.  Thus the causal functioning of the Combiner with setting X does depend on the state of b, despite the lack of counterfactual dependence of the output on b.  And this is enough to ensure that there is psychological continuity between B and the person with brain c.  Given that there clearly also is psychological continuity between A and the person with brain c, it follows that we do have a case of fusion—both A and B are psychologically continuous with the possessor of c—and hence (1) is false, the objection insists.  By the same token, (2) is false, and hence there is no problem here for the psychological continuity theorist.

            But the mere fact that b causally affects the functioning of the Combiner is of little importance.  After all, so do many other things, such as atmospheric changes in the vicinity of the Combiner, vibrations in the table on which it is set, and brain-waves from the person who turns the Combiner on.  As long as these things do not affect the output values, they are irrelevant to us here.

            Let us, however, consider the claim that there can be a causal connection between the state of b and the output of the Combiner in setting X.  Suppose that the nth bit in the description of b is 1, and so is the nth bit in the description of a.  One might then argue that the nth bit of the description of b’s being 1 together with the nth bit of the description of a’s being 1 causes the Combiner to output a 1.  Thus the states of both a and b are partial causes of the nth bit of c being 1.  If, further, a and b have qualitatively the same brain state, as in the case I am interested in, then this is true for every value of n, and so the state of a is a partial cause of the state of c and so is the state of b.

Hence, if being a partial cause is enough for quasi-memory-type connections, then A and B are each psychologically connected with the resulting person who has brain c, and so by no-branching considerations neither A nor B is identical with the c-brain person.  On the other hand, if being a partial cause is not enough, then neither A nor B is connected or continuous with the resulting person who has brain c.   In either case, it seems, the psychological continuity theorist can deny each of (1) and (2).

            However this overstates the importance of being a partial cause.  Let us suppose that we have a machine that simply copies bits from brain a to brain c while destroying brain a, and does not read b at all, but the machine is powered by the electrical activity of brain b.  Then the states of brain b, insofar as they affect the electrical activity of b, do indeed enter into the causal story.  But surely this is not enough to make the person with brain c count as connected with B.  In the case where the Combiner is powered by the electrical activity of b, there is no counterfactual dependence of the Combiner’s output on the details of b’s brain state, and exactly the same thing is true in the case of the Combiner with setting X.  Likewise, the fact that b is nearby may jiggle some of the electrical potentials in an a-to-c copy machine in a way that is insignificant vis-à-vis the outputs of the machine, but this surely is not relevant to questions of identity, again because of the lack of counterfactual dependence.

            It seems that counterfactual dependence is what matters for the causal aspect of quasi-memories, and this is indeed how it should be.  Someone who correctly quasi-remembers something also knows it.  But knowledge through memory or quasi-memory involves precisely that kind of counterfactual dependence: this dependence is what we need in order to ensure that it is not a coincidence that the person has the apparent memory or quasi-memory.  If the setting is X, then were the states of a different, those of c would be correspondingly different, but no similar relation holds between the states of b and those of c.  This seems to be the right counterfactual dependence condition for quasi-memories and quasi-intentions.  Hence, (1) holds if psychological continuity is the right theory of personal identity.  By the same reasoning, so does (2).

            There is also a general response that can be made to objections that proceed by insisting that in the special case where a and b are qualitatively identical, (1) and (2) are both false and in fact neither A nor B is identical with the possessor of c after the operation.  Since (1) and (2) hold in all other cases given psychological continuity, we get the absurdity that, say, with the Combiner set to X, person A would survive the operation if state of B’s were slightly unlike that of A’s, whereas A would not survive if B had exactly the same qualitative brain state, even though the output of the machine would be qualitatively exactly the same whether B was exactly or almost exactly like A, and would have exactly the same counterfactual dependencies. 

            IV. Biting the bullet.  The responses to the above objections might teach one that what matters for the causal aspect of psychological continuity is counterfactual dependence.  But now the road is open to denying the claim that the Combiner setting is irrelevant to identity facts in the case where a and b are qualitatively identical.  Even in that case, whether A is identical with the resulting person with brain c or whether it is B that is identical with that person depends precisely on the setting of the Combiner, on the present objection.  The Combiner setting never enters into the causal story, but it is nonetheless is counterfactually relevant, and this is what we need.

            This view is absurd for the reasons already noted.  The contents of sealed envelopes which do not actually enter into causal relations with anybody in the story do not determine facts about personal identity in that story (of course the contents might be evidence for a fact about personal identity if, say, they are a statement by an expert on some relevant question, but that is not what we are talking about).  We can vary the example in many ways.  Instead of a sealed envelope, we can imagine that the Combiner when first faced with a mismatch would ask a third party, Fred.  Then who, if anybody, is identical with whom would depend on facts about Fred’s disposition to answer a hypothetical question.  This is absurd.

V. Survival, not identity.  Parfit distinguishes survival from identity, and makes survival the important feature.  Perhaps, then, I can say that in the case where a and b are qualitatively identical, both A and B survive, regardless of the Combiner setting.  The case of qualitative identity is, then, different from that where a and b are not qualitatively identical, but there is principled ground for the difference: it is because the brain of A has the same data as the brain of B that both can survive through one person with brain c.

But if the Combiner setting is, say, X, then it is not clear what makes it be the case that B survives through the person with brain c.  The mere fact that c happens to have the same mental states as B is insufficient, since that fails to connect these mental states—it is just a coincidence that the possessor of c has the same mental states as B, since the counterfactual dependence is on the states of A.  If I were sentenced to death and found out that there is a Twin Earth where there is a twin qualitatively identical with me, but the histories of Earth and Twin Earth are about to diverge so that the twin would survive, this would not assuage my egocentric concerns, though it would make me glad that somebody would fulfill some of my plans, even if he would not fulfill them because they were mine.

4. Another case

            We may now note that there is a further counterexample to any psychological identity theorist who requires an appropriate counterfactual dependence of quasi-memories on the quasi-remembered events.  In this case, we have only two brains a and b with qualitatively identical states—there is no separate output brain c.  A scanning machine examines each bit of data in a, and if that bit matches the corresponding bit in b, it leaves it be in a, but otherwise it imposes the data from b onto the data in a.  All the while, b is being destroyed.  Thus there is counterfactual dependence of the post-operative state of brain a on the state of b before the operation, but no counterfactual dependence of the post-operative state of brain a on its own pre-operative state. 

            Now if A survives after the operation, it is clear that a is A’s brain, since there is no other brain there for A to have after the operation.  But the quasi-memories in a after the operation do not have the needed counterfactual dependence on the events that had befallen A before the operation.  Hence, if the psychological theory of personal identity holds and if counterfactual dependence of quasi-memories on quasi-remembered events is needed, it follows that A can no longer exist after the operation.  But this is absurd, because the only thing that actually happened to A during the operation was that A’s brain was scanned.  All the other changes were merely counterfactual—brain a was put in danger of having the states of b imposed on it, but in fact no such imposition happened.  The intuition that A would not be affected by such an operation is very similar to the intuition in Frankfurt-style arguments that one’s freedom is not affected by the mere fact of a neurosurgeon’s standing watching one’s brain and being ready to intervene in counterfactual situations.

Things are tough for the psychological continuity theorist: I have just given an argument for why she should deny the necessity of counterfactual dependence.  But at the same time she should accept the necessity of counterfactual dependence because causal dependence is not enough.  This is strong evidence against psychological identity theories of personal identity.

5. Conclusions

            The possibility of qualitatively identical persons provides strong evidence against typical psychological continuity theories.  The reason for this appears to me to be that psychological connectedness places too great an emphasis on qualitative identity rather than on non-qualitative conditions, such as persistence of chunks of matter.  Consequently there is probably a whole slew of counterexamples to psychological continuity theories that can be generated by using qualitatively identical persons.  For instance, imagine scanning two such persons, destroying the brains, and saving the scans on a hard drive for future restoration.  However, suppose that the computer has clever data compression software that detects identical bit-sequences, and replaces them with a single sequence plus a note that the sequence is to be duplicated.  Thus, the backups of the two persons got replaced with a backup of one, plus a note that there was another file identical with that one.  Could one then recreate the two original persons from such a compressed backup?  If so, which person would continue which?  I leave it to the imagination of the reader to generate other cases.

            Are cases of qualitative identity a problem for other theories of identity?  Maybe.  But at least the present problem is not, as we have seen in the response to the first objection.