The Ethics of Timelooping

I am writing a story (well more thinking about and working on the world building (hopefully) for now) where my MC has the ability to rewind time to designated save points. Let's call him John. So the way this works is John has the ability at a given time t_x to set a save point (provided he has not exhausted all his save slots) and at a future time t_y, John can "rewind" to t_x. The way John experiences this "rewinding" is that at t_x John would have all his memories of the duration [t_x, t_y], so it's a form of mental time travel (this is probably the main form of time travel (at least on a multiverse wide scale) that would be covered in this story). How this time travel actually works has some profound ethical implications.

I do not have a canon model of how time travel works in my story (if I find out that my current model is incoherent, I would throw it out, otherwise I would likely live with the implications of it), so while I do have a favoured mental framework for thinking about time travel, there is not much reason to privilege my particular framework and treat it as anything other than just another model in the model space of how time looping abilities can work.

My world is pretty big with several universes (probably an infinite amount of them), but it is probably strictly smaller than the Tegmarkian multiverse. Different universes have different temporal interactions with our universe, but John's time rewind ideally (I'm no physicist (actually, you can probably assume I have no knowledge of physics beyond what little I retained from cramming for high school exams, but I do intend to keep a coherent model. It is integral to the story that the time rewind occurs everywhere, so if certain details of my world makes that impossible, I would probably shrink it as needed) rewinds time on a multiverse wide scale.

The interesting question for the purposes of this thread, is: what happens at t_{y+1}? After John jumps back in time, what happens to the rest of the world?

There are many different ways of thinking of this. The approach I'm currently using is to imagine timelines. Whenever a rewind occurs, the timeline from origin to t_x (the point of the rewind) is duplicated (or perhaps the duplicate already exists), and the world state at t_x is updated to reflect the new information that John has. Let's call the "source" timeline T_1, and the duplicate timeline T_1.1. If we work with this framework, then what happens to T_1 at t_y has very profound implications. The way I see it there's two basic outcomes:

1. T_1 continues operation normally.
   
   
2. T_1 is terminated at t_y.

If T_1.1 is terminated at t_y, then by rewinding time John is committing omnicide. He's literally killing an infinite number of people each time he rewinds creating infinite disutility (or the upper bound of disutility if you use a bounded utility function). Suppose John assigns a probability of p to the hypothesis "rewinding time terminates the source timeline". I'm not sure how John assigns his probabilities, but whatever prior he's using — as long as it's sensible — I would imagine that it wouldn't assign a probability below say 10-10 to the above hypothesis (probably several orders of magnitude more in fact). If John is doing an expected value calculation, then the possibility of omnicide should dominate his calculations (even with a bounded utility function) when evaluating whether or not to rewind time. In short, John seems to be getting Pascal mugged (only without an extant mugger).

I mentioned above that either T_1.1 is created from T_1 due to the rewind, or T_1.1 already exists. I think this also has interesting ethical implications of its own. For one, if the rewind causes the creation of a new timeline, then each rewind causes amounts of utility equivalent to the total utility history of the multiverse (and if that number is positive), then the moral thing for John to do is to rewind as many times as possible (if the number is negative John creates even more disutility). Interestingly, if the rewinds creates the new timelines, and the multiverse is net positive utility, then the positive utility created by the multiverse may outweigh the negative utility created by omnicide.

As an aside, even if hypothesis 1 was true, then John may still cause omnicide as there's a multiverse ending event (John believes his time rewinding ability is to avert it (he auto rewinds to a fixed save point at when he first awakened his powers if he dies or gets stuck in an infinite loop)), so if the time rewind clones the timeline, if all possible timelines don't already exist and rewinding creates a new timeline, then it seems to me that John rewinding time is omnicide.


Sidestepping the talk of timelines for a bit, it seems to me that no matter how you dice it timelooping/rewinding has profound ethical implications. After the rewind happens, it's possible to treat the period of time [t_x, t_y] as a mere simulation of the future which the MC now has knowledge about (perhaps John's power merely simulates the multiverse (and informs him of the results of said simulation)). This brings up the interesting question of whether simulated copies of an agent have moral worth. If 1,000,000 copies of Jane Smith exist and I terminate 100,000 of them have I created disutility? If one believes that simulated copies don't have moral worth separate from the original, then I'd like to ask you this question: if someone told you that you were a simulation and decided to terminate you, would you consider it murder? Alternatively, if we found out that this universe was a simulation (and that there was a real "earth" somewhere) should we consider the termination of the said simulation omnicide? In short, do you believe in mindcrime(scraped on github (cause Arbital isn't reliable))?


If one privileges some "real" timeline, I'd like to point out that if the simulation goes on for long enough, the simulated agents may sufficiently diverge from their "real" copies that they're no longer the same moral persons. Alternatively, some entirely new agents may be created in the simulation that would not exist in the "real world".


I find it interesting that literally every other action John takes seems negligible in the face of the astronomical amounts of (dis)utility created by his time rewinding ability. I also wonder how someone who believes(believed?) in "shut up and multiply" would react to that realisation.

 

Er, I have to say that this just kind of seems like creating problems from comparatively thin reasons.

The possibility of time traveling meaning that you destroy all realities, seems like a massive jump to me. There does exist 'time travel' like that (For example a time loop in "Mother of Learning" functions not by going back in time but by destroying the world and recreating it as it was at the time they 'looped' back to), but outside those circumstances regarding it as the destruction of reality seems pretty out there to me. Different points in time exist, and while there can be ethical considerations about what going back to one of them means, I really don't see that as equal to literally blowing up the multiverse.

For the MC potentially coming to think of one of the rewinds as 'simulation' and whether killing someone who is potentially a simulation counts as murder - of course it does. And if the MC seriously considers excusing his murders with something like that he's an asshole. To him all of that seems to still function like reality, so even if there were 'nothing under the hood' for the people he interacts with, him simply deciding that it's ok to kill them because he decided it probably works like that would be monstrous.

The potential detachment created by seeing someone rewind and act out the same actions again and again, and a MC reacting by slowly sliding towards basically considering them less then human, would be an interesting thing to explore. But him simply deciding, these people are probably just simulations and so I can easily switch of my empathy and kill them without problem would say fairly bad things about the character.


It's fiction, so it will obviously play by the rules you decide on and this could result in some interesting consideration, but to me this is too removed from my conception of time to be properly appealing - at least as you have written it out here.

 

It seems strange to me that one would frame this in terms of ending universes and creating new timelines when this power could be much more simply conceived as replacing the state of the universe at T_y with the state of the universe at T_x (with the exception that John's mind proceeds continuously from its state at T_y).

The question of ending worlds is a distraction. The questions of simulations and alternate timelines are distractions. Even the idea of "time travel," to begin with, is a distraction. John's power is reality alteration, with a universal scope, limited to states of reality that have previously occurred.


Edit: Just noticed the poll. My answer would be that time looping only has ethical implications in the sense that it enhances one's ability to do the most good. Since I don't think that's the sense in which you mean the question, I'll be answering "no."

 

Even so I know plenty of people who would consider being modified with out their consent ethically wrong so this doesn't quite answer the question. Let's say you were an unemployed, unhappy, lazy slob but by complete chance met the girl of your dreams and turn you life around. You get a job, have some kids and become truly happy. Would you be ok with some jack ass erasing all your memory's, erasing your kid, and reverting you back to the unhappy slob you were to start with?

 

The OP is pretty clearly a utilitarian, so it's relevant for me to point out that all the guys who used to be happily employed family men married to the girls of their dreams and then, by complete chance, lost everything and everyone they loved have now also been reverted to their previous states.

The math works out very differently for a scenario where you're killing universes vs. a scenario where you're adjusting things to another state of affairs; in the latter case, you start asking questions like whether the newly restored state hold more net utility than the state that was overwritten, while in the former case, you start adding up all the negative utility you accrued by killing a universe so you can compare it to some other value.

I don't think it makes any sense to approach a problem like this as though it involved omnicide --that's all. Any other point you might imagine I was trying to make either isn't a point I was trying to make or isn't a point I'm interested in defending in this thread.