Does it matter if you don’t remember?
Does it matter if you experienced pain in the past, but you don’t remember? (And there are no other side-effects, etc etc). At one point in Accelerando, Charles Strauss describes children that routinely decapitate and disembowel each other, only to be repaired (bodily and memory-wise) by the friendly local AI. This struck me as awful, but I’m suspicious of my intuition. Note that here I’m assuming pain is a terminal “bad” factor in your utility function. You can substitute “pain” for whatever you think is bad. I think there are at least two questions here:
Is it bad for someone to be in pain if they will not remember it in the future? I think yes, because by assumption pain is a terminal “bad” node. Being relieved of future painful memories is good, but nowhere near good enough to fully compensate.
Is it bad to have experienced pain in the past, if you don’t remember it? Or, can your utility function coherently include facts about the past, even if they have no causal connection to the present? My intuition here says yes, but I’d be interested in others’ thoughts. To make this concrete, imaging that you have a choice between medium pain that you will remember, or extreme pain followed by memory erasure.
It seems like it is worse for someone to be tortured to death than to be put in a coma for the same duration as the torture and then killed. Death eliminates memories as readily as fancy erasure techniques. So it seems like pain-moments are bad, and while pain-memories are probably also bad, they do not constitute the whole of the badness of pain-moments.
On the subject of pain-moment badness, If pain becomes more common (even if it gets forgotten), then at any given instant I’m more likely to be in pain than otherwise, which I would consider worse, even if I forget it afterwards.
So torturing people then removing the memory is a bad thing, as a general rule.
I’m still a bit shaky on how that affects question 2.
Well said. It also bears noting that being tortured and then killed is (I think?) considered worse than merely being tortured. I realize that this is mostly because death has greater consequences than just erasing memories, but it makes me think—if you’re going to be tortured anyway, would you rather remember it?
I think I might. Some of that is probably an instinctive aversion to having my consciousness messed with, or the vague feeling that I have a right to be as aware of my own past as possible. But it also seems like having experienced serious pain could dampen everyday sorts of pain by comparison, which could come in handy. On the other hand, of course, PTSD. (I wonder how many PTSD sufferers would have their memories erased, if that were possible?)
This is something I’ve worried about before, after reading about surgeries performed with sufficient anesthetic to immobilize a person, but insufficient to dull the pain. It’s driven me to wonder whether there are surgeries which are experienced but not remembered?
One important distinction is the difference between not remembering and not having experienced. Using surgery as an example, I may not remember the agony I went through, but my brain will have changed from the experience in a not-insignificant way; being rebooted to a save-state would be entirely different.
In fact, I strongly suspect that I was fed opiates as a child during a hospital stay—though I don’t remember many details, and none of those that I do are suggestive of being drugged, I find opinions and shards of idea within me which are highly suggestive of this experience. I find it far too easy to envision what an opiate high would be like—along with the nausea—and in the future would probably take the pain over the side-effects of the high.
Intuition pump: If you don’t remember it, it’s like it happened to somebody else.
I don’t particularly want other people to feel pain even if I don’t know about it, so I suppose I’d object to “forgetful me” experiencing pain, too.
Let’s make the distinction starker: Dead beings don’t remember. So if one concludes that remembered pain is irrelevant one should conclude that there’s no distinction between killing someone and torturing them to death. That clashes strongly enough with my intuition that I’d rather go and claim that lack of memory doesn’t make it not matter.
You are going to die. When you do, you won’t remember anything. If all that matters is what you remember, nothing matters.
It may be worthwhile thinking about the flipside of this question. Is it good for someone to remember a pleasure that never happened?
Suppose some uploaded human sets themselves the goal of achieving a perfect score at some game. Is an FAI doing her a favor by adjusting her memory to indicate that the goal has already been achieved?
But I think it is even more worthwhile to puzzle over whether your second question even makes sense.
You are asking for a moral judgment of past events. But can we (should we?) do this? The whole point of making moral judgments regarding future events is to help to guide present or future decision making. Can there be any point to making judgments regarding past events? How does this pay the rent?
If you alter your model of the world, you should normally spend some resources sanity checking it—by feeding it recorded data from your memory stores—to see if it makes correct predictions for the data you have on record.
For most non-trivial updates, that process will naturally involve predicting the actions of your past self—and the actions of others—and so will involve issues associated with morality and decision making.
This reminds me of the standard argument for infant circumcision: that a baby won’t remember it or be permanently damaged by it, and a baby is constantly crying in pain or discomfort anyway. The logic goes that it would be more hurtful to circumcise an older child or an adult, because he would be more aware of the pain, and remember the experience.
(The same claim would hold for, say, ear-piercing—should you pierce a baby girl’s ears or let her get pierced at thirteen?)
Of course, here the tradeoff is slightly different—it’s between pain that you forget, when you’re too young to have agency, and pain that you remember, when you’re old enough to have a chance to say “no.”
Assuming I want to have pierced ears, I would rather have had it done in my infancy, than have to exert willpower and endure pain now.
If you live in a culture where most males are circumcised, or most females have pierced ears, and most adults prefer to have the procedure done, then it’s kinder to a child to get the pain over with and forgotten, than to give him/her the opportunity to stress over it and remember it at an older age.
It’s also true, or so I’ve heard, that mothers forget the pain of childbirth. Some kind of opioids kick in and erase the memory. Otherwise women would never have more than one child. Assuming that I want more than one child, I’d choose extreme pain that I forget, rather than moderate pain that I remember, because it’ll require less willpower to have a second child.
I believe that pain is apt to result in habitual muscle tension, even if the pain isn’t consciously remembered.
This relates closely to the relative value we should give to experienced vs remembered happiness.
Your utility function shouldn’t bother to include the past at all. You can’t affect the past, so measuring it’s utility/disutility is a waste of resources. (Unless you are an “average happiness measured over all time” maximiser, or similar)
This is a question about the future, not about the past.
And I’d take the medium pain= more experiences, shouldn’t break my mind, no version of me has to be killed (reverting to a saved version is killing the divergence)
I’d say it’d be a bad thing, since it’d result in wasteful expenditures of resources by the AI, as well as maladaptive learning by the children as they grow up; what if they go somewhere outside the AI’s control?
My primary concerns with pain are that they a) coincide with permanent injury, and b) will result in some kind of post-traumatic nervous disorder in the future. Offer me the chance to experience some sort of pain without either of those and I’d probably accept out of sheer curiosity.
But your curiosity won’t be sated because you won’t remember it in the future.
And what if it were 100 years of agony, after which you would return to your present condition and remember nothing?
Opportunity cost: I could have spent those 100 years doing things a lot more pleasurable, and remembered those.
I wonder if some kind of TMS-like intervention could be used to give oneself temporary pain asymbolia. Then you wouldn’t even have to posit loss of memory to get the “no memory of suffering” element of your scenario.
If I couldn’t remember it, then in what meaningful way have I experienced it?
Alternatively put, how can you consequentialistically assess something which has no consequence?
In the sense that it actually happened to you, like, fo’ real.
Well that’s part of my question and confusion, really. But it does seem worth answering: do you disagree that torture for 100 years followed by memory reversal is indeed a bad thing?
Did it happen to me, though? What’s the difference between this hypothetical me who’s been tortured for a hundred subjective years before having his memory erased and the actual real me who (presumably) hasn’t? He doesn’t have any more experience of being tortured than I do.
I agree that it’s a “bad thing” in the sense that it’s an incredibly shitty thing to do to someone, but I don’t believe a scenario in which an action has no consequences can be distinguished from scenarios in which those actions haven’t taken place.
Given that actions tend to have consequences, I’m probably not going to fret too much over it.
If the human race comes to an end with no descendants and aliens never find the remains, can anything that anyone has done be said to have consequences if you start with the premise that qualia don’t matter unless they affect later experiences?
I don’t think they can outside of the scope of being able to experience them.
Here’s a good one: Omega appears before you and offers you an arbitrarily huge amount of money. He then tells you that earlier in the day he predicted whether or not you’d accept his offer. If you did accept, he tortured you horribly for a subjective hundred years and then erased your memory of the event. If you didn’t accept, he didn’t do anything.
Do you accept his offer?
Yes, because I could buy a lot of expected utility with an arbitrarily huge amount of money.
But if the offer were for a million dollars, then I would not accept it.
Upvoted because I like the analogy; it made me think about why I believe what I do about this point.
I would accept the offer, but I don’t find that inconsistent, for two reasons. First, the original post was asking whether the memory erasure makes up for the pain, which we’re still debating. You’re asking if the memory erasure AND an arbitrarily fat wad of cash make up for the pain, which may be less delicately balanced.
The better point is that I am not only the endpoint of my experiences. At some point, every moment in my life between the present one and death will be “now.” Considering a choice about potential pain which might occur in the future, I know that I will have to live through each of those moments, as they pass, and they will be very unpleasant. There may also be a lot more of them than there are happy, rich, pain-forgotten moments afterwards. That’s the kind of choice we’ve been discussing elsewhere in the thread, and that’s why I care about experiences I will have in the future even if they are forgotten later.
However, when Omega presents me with that choice, I am not choosing whether I will live through those moments. It’s a fait accompli. Regardless of my choice, I will not be aware of having been tortured; the only difference is whether I walk away with the wad of cash.
… assuming that having been tortured for 100 years really will have no lasting effects just because I don’t remember it. I’m making that assumption because I think it’s what you intend the problem imply; if this came up in real life I’d ask some questions about it before choosing. (If I were not allowed to do so, I would say no to be on the safe side.)
This is an interesting question but it seems to conflate this question a bit too much with all the very difficult to resolve decision theory questions about counterfactual issues. This seems at least to me to hinder intuition rather than acting as a good intuition pump.
No—just imagining sketchily what I would go through and remembering that I did that to myself would have a cost.
And I can’t help wondering how thorough the memory erasure would be.
Also, although this doesn’t bear on the larger question, I think that knowing while I was being tortured that the I who was being tortured was going to be eliminated and wouldn’t get any of the money would make it worse.
Well, even from a strict consequentialist position, it involves a century’s worth of opportunity costs, which is no small thing. Also, every step of the way until the very end, it involves a non-zero chance of not being able to fully reverse the effects, unless you want to invoke an unrealizable condition like knowing for sure that you’ll be able to apply memory reversal when the time comes.
It happened to a person who shares a past with me.
If there are no consequences then me, and the tortured person, are both futures of “pre-torture-me”, but we’re not really the same person (at least, I’m not the same person zhe would have been by the end of the 100 years)
I’ve done things in the past I don’t remember doing, but I don’t think it’s a good use of the word “experience” to say that I didn’t experience those things. What if my parents tell me about something I did as a kid, which even after being told I don’t recall, but I believe them? (This has happened, of course; introductions to their friends, for example, or places we’ve been.) On which side of the alleged experience/nonexperience line does that fall? For that matter, why are we assuming that something which is not remembered has no consequences? Something could have happened to me when I was young which influenced my tastes or choices but which I still later didn’t remember.
It’s a troublesome word. I don’t feel I have experience of an event I don’t remember, but I presumably have experienced it if I was a participant. I suppose I think a defining characteristic of experience is being able to draw upon it, which I can’t do if I don’t remember it.
Hmm. If it can be used in both ways, perhaps we can be more productive/clear if we use another one.
I don’t think I have another word for it in either context. “Participatory experience” and “retrospective experience”, maybe?
I think I might be having some conceptual trouble with this one, since an event I have no memory of is an experience I haven’t experienced.
Exactly. I might say they’re “events which have occurred” and “events one can recall,” which itself is distinct from “events is aware of,” given other observers who recall them. I suggest we taboo “experience” altogether.