Dissolving the Thread of Personal Identity

(Back­ground: I got in­ter­ested in an­throp­ics about a week ago. It has tor­mented my wak­ing thoughts ever since in a cy­cle of “be con­fused, de­velop idea, work it out a bit, re­al­ize that it fails, re­peat” and it is se­ri­ously driv­ing me berserk by this point. While draw­ing a bunch of “thread of per­sonal con­ti­nu­ity” di­a­grams to try to flesh out my next idea, I sus­pected that it was a fairly non­sen­si­cal idea, came up with a thought ex­per­i­ment that showed it was definitely a non­sen­si­cal idea, re­al­ized I was try­ing to an­swer the ques­tion “Is there any mean­ingful sense in which I can ex­pect to wake up as my­self to­mor­row, rather than Brit­tany Spears?”, kept think­ing any­ways for about an hour, and even­tu­ally came up with this pos­si­ble re­duc­tion of per­sonal iden­tity over time. It differs some­what from Kaj So­tala’s. And I still have no idea what the hell to do about an­throp­ics, but I figured I should write up this in­ter­me­di­ate re­sult. It takes the form of a men­tal di­alogue with my­self, be­cause that’s what hap­pened.)

Doubt: Hang on, this whole no­tion of “thread of per­sonal con­ti­nu­ity” looks sort of fishy. Self, can you try to clar­ify what it is?

Self: Let’s see… I have a causal link to my past and fu­ture self, and this causal link is the thread of per­sonal iden­tity!

Cur­rent Me: Please no­tice Past Self’s use of the cached thought from “Time­less Iden­tity” even though it doesn’t fit.

Doubt: Causal links can’t pos­si­bly be the thread of per­sonal con­ti­nu­ity. Your state at time t+1 is not just caused by your state at time t, lots of events in your sur­round­ings also cause the t+1 state as well. A whole hell of a lot of stuff has a causal link to you. That can’t pos­si­bly be it. And when you die, al­ive you has a causal link to dead you.

Doubt: And an­other thing, per­sonal con­ti­nu­ity isn’t just an on-off thing. There’s a gra­di­ent to it.

Self: What do you mean?

Doubt: Let’s say you get frozen by cry­on­ics, and then re­vived a cen­tury later.

Self: Sure.

Doubt: Let’s say you know that you will be re­vived with ex­actly the same set of mem­o­ries, prefer­ences, thought pat­terns, etc, that you have cur­rently. As you are be­gin­ning the pro­cess, what is your sub­jec­tive cre­dence that you will wake up a cen­tury later?

Self: Fairly close to 1.

Doubt: Now, let’s say they could re­cover all the in­for­ma­tion from your brain ex­cept your ex­treme love for choco­late, so when your brain is re­stored, they patch in a generic av­er­age in­cli­na­tion for choco­late. What is your sub­jec­tive cre­dence that you will wake up a cen­tury later?

Self: Fairly close to 1.

Doubt: Let’s say that all your in­cli­na­tions and thought pat­terns and other stuff will be re­stored fully, but they can’t bring back mem­o­ries. You will wake up with to­tal am­ne­sia. What is your… you get the idea.

Self: Oh crap. I… I re­ally don’t know. 0.6??? But then again, this is the situ­a­tion that sev­eral real-life peo­ple have found them­selves in… Huh.

Doubt: For this one, in­cli­na­tions and thought pat­terns and many of your mem­o­ries are un­re­cov­er­able, so when your brain is re­stored, you only have a third of your mem­o­ries, a strong be­lief that you are the same per­son that was cry­op­re­served, and a com­pletely differ­ent set of… ev­ery­thing else ex­cept for the mem­o­ries and the be­lief in per­sonal con­ti­nu­ity. P(I wake up a cen­tury later)?

Self: Quite low. ~0.1.

Self: But I see your point. For that whole per­sonal iden­tity/​wak­ing up as your­self thing, it isn’t a bi­nary trait, it’s a slid­ing scale of be­lief that I’ll keep on ex­ist­ing which de­pends on the mag­ni­tude of the differ­ence be­tween my­self and the be­ing that wakes up. If up­load!me were fed through a lossy com­pres­sion al­gorithm and then re­con­structed, my de­gree of be­lief in con­tin­u­ing to ex­ist would de­pend on how lossy it was.

Doubt: Now you re­al­ize that the “thread of sub­jec­tive ex­pe­rience” doesn’t ac­tu­ally ex­ist. There are just ob­server-mo­ments. What would it even mean for some­thing to have a “thread of sub­jec­tive ex­pe­rience”?

Self: (Taps into in­tu­ition) What about that big rock over there? For­get “sub­jec­tive”, that rock has a “thread of ex­is­tence”. That rock will still be the same rock if it is moved 3 feet to the left, that rock will still be the same rock if a piece of it is chipped off, that rock will still be the same rock if it gets cov­ered in moss, but that rock will cease to be a rock if a nuke goes off, turn­ing it into rock va­por! I don’t know what the hell the “thread of ex­is­tence” is, but I know it has to work like that rock!!

Doubt: So you’re say­ing that per­sonal iden­tity over time works like the Ship of Th­e­seus?

Self: Ex­actly! We’ve got a fuzzy cat­e­gory, like “this ship” or “this rock” or “me”, and there’s stuff that we know falls in the cat­e­gory, stuff that we know doesn’t fall in the cat­e­gory, and stuff for which we aren’t sure whether it falls in the cat­e­gory! And the thing changes over time, and as long as it stays within cer­tain bounds, we will still lump it into the same cat­e­gory.

Doubt: Huh. So this “thread of ex­is­tence” comes from the hu­man ten­dency to as­sign things into fuzzy cat­e­gories. So when a per­son goes to sleep at night, they know that in the morn­ing, some­body ex­tremely similar to them­selves will be wak­ing up, and that some­body falls into the fuzzy cluster that the per­son fal­ling asleep la­bels “I”. As some­body con­tinues through life, they know that two min­utes from now, there will be a per­son that is similar enough to fall into the “I” cluster.

Doubt: But there’s still a prob­lem. 30year­future!me will prob­a­bly be differ­ent enough from pre­sent!me to fall out­side the “I” cat­e­gory. If I went to sleep, and I knew that 30year­future!me woke up, I’d con­sider that to be tan­ta­mount to death. The two of us would share only a frac­tion of our mem­o­ries, and he would prob­a­bly have a differ­ent set of prefer­ences, val­ues, and thought pat­terns. How does this whole thing work when ver­sions of your­self fur­ther out than a few years from your pre­sent self don’t fall in the “I” cluster in thingspace?

Self: That’s not too hard. The “I” cluster shifts over time as well. If you com­pare me at time t and me at time t+1, they would both fall within the “I” cluster at time t, but the “I” cluster of time t+1 is differ­ent enough to ac­com­mo­date “me” at time t+2. It’s like this rock.

Doubt: Not the rock again.

Self: Quiet. If you had this rock, and 100year­future!this­rock side by side, they would prob­a­bly not be rec­og­niz­able as the same rock, but there is a con­tin­u­ous se­ries of in­ter­me­di­ates lead­ing from one to the other, each of which would be rec­og­niz­able as the same rock as its im­me­di­ate an­ces­tors and de­scen­dants.

Self: If there is a con­tin­u­ous se­ries of in­ter­me­di­ates that doesn’t hap­pen too fast, lead­ing from me to some­thing very non­hu­man, I will an­ti­ci­pate even­tu­ally ex­pe­rienc­ing what the non­hu­man thing does, while if there is a dis­con­tin­u­ous jump, I won’t an­ti­ci­pate ex­pe­rienc­ing any­thing at all.

Doubt: Huh.

Self: So that’s where the feel­ing of the “thread of per­sonal iden­tity” comes from. We have a fuzzy cat­e­gory la­beled “I”, an­ti­ci­pate ex­pe­rienc­ing the sorts of things that prob­a­ble fu­ture be­ings who fall in that cat­e­gory will ex­pe­rience, and in ev­ery­day life, there aren’t fast jumps to spots out­side of the “I” cat­e­gory, so it feels like you’ve stayed in the same cat­e­gory the whole time.

Doubt: You’ll have to un­pack “an­ti­ci­pate ex­pe­rienc­ing the sorts of things that prob­a­ble fu­ture be­ings who fall in that cat­e­gory will ex­pe­rience”. Why?

Self: Flip­pant an­swer: If we didn’t work that way, evolu­tion would have kil­led us a long time ago. Ac­tual an­swer: Me at time t+1 ex­pe­riences the same sorts of things as me at time t an­ti­ci­pated, so when me at time t+1 an­ti­ci­pates that me at time t+2 will ex­pe­rience some­thing, it will prob­a­bly hap­pen. Look­ing back­wards, an­ti­ci­pa­tions of past selves fre­quently match up with the ex­pe­riences of slightly-less-past selves, so look­ing for­wards, the an­ti­ci­pa­tions of my cur­rent self are likely to match up with the ex­pe­riences of the fu­ture be­ing who falls in the “I” cat­e­gory.

Doubt: Makes sense.

Self: You’ll no­tice that this also de­fuses the an­thropic trilemma (for hu­mans, at least). There is a 1 in a billion chance of the quan­tum ran­dom num­ber gen­er­a­tor gen­er­at­ing the win­ning lot­tery ticket. But then a trillion copies are made, but you at time (right af­ter the gen­er­a­tor re­turned the win­ning num­ber) has a trillion ex­pected near-fu­ture be­ings who fall within the “I” cat­e­gory, so the 1 in a billion prob­a­bil­ity is split up a trillion ways among all of them. P(loser) is about 1, P(spe­cific win­ner clone) is 1 in a quin­til­lion. All the spe­cific win­ner clones are then merged, and since a trillion differ­ent hy­pothe­ses each with a 1 in a quin­til­lion prob­a­bil­ity all pre­dict the same se­ries of ob­served fu­ture events from time(right af­ter you merge) on­wards, P(se­ries of ex­pe­riences fol­low­ing from win­ning the quan­tum lot­tery) is 1 in a billion.

Doubt: Doesn’t this im­ply that an­thropic prob­a­bil­ities de­pend on how big a bound­ary the mind draws around stuff it con­sid­ers “I”?

Self: Yes. Let’s say we make 2 copies of a mind, and a third “copy” pro­duced by run­ning the mind through a lossy com­pres­sion al­gorithm, and un­com­press­ing it. A blue screen will be shown to one of the perfect mind copies (which may try to de­stroy it). A mind that con­sid­ered the crappy copy to fall in the “I” cat­e­gory would pre­dict a 13 chance of see­ing the blue screen, while a mind that only con­sid­ers near-perfect copies of it­self as “I” would pre­dict a 12 chance of see­ing the blue screen, be­cause the mind with the broad defi­ni­tion of “I” se­ri­ously con­sid­ers the pos­si­bil­ity of wak­ing up as the crappy copy, while the mind with the nar­row defi­ni­tion of “I” doesn’t.

Doubt: This seems to ren­der prob­a­bil­ity use­less.

Self: It means that prob­a­bil­ities of the form (I will ob­serve X) are mind-de­pen­dent. Differ­ent minds given the same data will dis­agree on the prob­a­bil­ity of that state­ment, be­cause they have differ­ent refer­ence classes for the word “I”. Prob­a­bil­ities of the form (re­al­ity works like X)… to be hon­est, I don’t know. An­throp­ics is still ex­tremely ag­gra­vat­ing. I haven’t figured out the hu­man ver­sion of an­throp­ics (us­ing the per­sonal con­ti­nu­ity no­tion) yet, I es­pe­cially haven’t figured out how it’s go­ing to work if you have a AI which doesn’t as­sign ver­sions of it­self to a fuzzy cat­e­gory la­beled “I”, and I’m dis­trust­ful of how UDT seems like it’s op­ti­miz­ing over the en­tire tegmark 4 mul­ti­verse when there’s a chance that our re­al­ity is the only one there is, in which case it seems like you’d need prob­a­bil­ities of the form (re­al­ity works like X) and some way to up­date far away from the Boltz­mann Brain hy­poth­e­sis. This above sec­tion may be con­fused or flat-out wrong.