Tyler Cowen in Stubborn Attachments: The real issue is that we don’t know whether our actions today will in fact give rise to a better future, even when it appears that they will. If you ponder these time travel conundrums enough, you’ll realize that the effects of our current actions are very hard to predict,
The more plausible scenarios seem to be 'accelerationist' ones.
I.e consider that Germany wins, take over much of Europe, and then gets overthrown by some anti-fascist European revolution which leads to all of Europe being some sort of relative utopia, free from the Cold War.
Or consider that the South wins, institutes some awful regime, and then the end of slavery comes from some anti-slaveholder cross-racial revolt, which institutes some near utopia in the U.S.
Thinking about this more there is an essential vagueness in the question. I could be asking something like: "If we could time travel and kill Hitler in way X that would save 1 billion utils but we'd have to reroll all subsequent years what's the chance that things would turn out better." This question just turns on the model we use for `rerolling' later years.
However, I think it's actually more common for people to understand the question as asking something like: how confident are you that (evaluated with respect to objective physical probability) that E(X| no nazis) > E(X| nazis) where X is sum of utils in years since 1940. This question doesn't depend at all on how variable the utils earned per year might be since that is averaged out in the expectation. It merely asks us to judge how confident we are in our assumption that on net the nazis were bad rather than say necessary to warn us off a much more destructive nuclear war that would have almost inevitably occurred otherwise.
All this depends greatly on the notion of probability involved. If we mean objective quantum probability it could well turn out that (surprisingly unbeknownst to us) there was a 90% chance of nuclear apocalypse given no nazis. If you have in mind some more subjective epistemic notion than it's not at all clear why this shouldn't be almost 1.
I'm not even sure what you mean by the variance here since we are asked about the probability of a simple binary outcome (the world is either better or worse as a result of the Nazi's influence).
As for the effect after 1000 years I think your wrong. Even being super generous and assuming each year produces X utils ~ N(0,1) i.i.d. and the nazis imposed a 50 util cost (no worse than switching from an average to a 20th percentile year) it's still a 95% chance that a randomly chosen no nazi scenario beats a randomly chosen nazi scenario. If we assume, more accurately, that there are downstream effects of the previous year, e.g., X_i ~ X_{i-1} * e^{N(0,1)} then it would actually increase the difference in expectations over time.
However, all of this presumes a very unnatural, IMO, interpretation of what we mean when we ask the question.
Yes, and 'we' in the post just means the humans who would be alive in the designated scenario. It's not that unusual to use words like 'we' to refer to a group even if it's individual members are likely to change. Think of a basketball player saying "We'll do much better next year if you trade me to Chicago for Michael Jordan." That makes sense even though that player won't be part of the 'we' next year if the trade is made.
So what if the world doesn't have goals/preferences. When I use that phrase (and I think when Hanson does too) I have in mind something like 'increases net utility/happiness' and surely you believe there is at least some vague sense in which I can compare overall degrees of suffering/pleasure.
People seem to be assuming without justification that it is better to make a less offensive statement than a more offensive statement. The alternative hypothesis is that people are better off if they contemplate ten offensive things before breakfast, perhaps because that would let them get through the rest of their day without having their buttons so easily pushed by others. In this case people writing offensive things would be providing a service to people doing this breakfast contemplation, or other people who would benefit from the breakfast contemplation but are unaware of this. I do not know which is true and would be interested to see evidence either way.
Robin! You frequently seem to run afoul of people who misinterpret your meanings and motivations. Why don't you take some time to do a little table setting so that people are less likely to pre-judge or mis-judge you?
Example. Sometimes I will tell a story and mention the race of a person in the story. I don't want to seem or be seen as the kind of person who specifies the race of every person in a story, so I'll make an aside--"I mention the race here because it becomes important later." This is a small investment to make, though unfortunate that it needs to be made.
I have friend who frequently wears soiled pants and complains that girls aren't interested in him. When I say, "Man, don't wear sweatpants! And if you do, make sure they are clean." He'll usually say, "Women should like me for who I am! Why does it matter what kind of pants I have?" I could argue that the women are making a good choice based upon the limited info they have. But whether or not they should "like him for him", he should simply do the things that can most foerseeably lead to good outcomes. And try to avoid easily foreseeable bad outcomes.
You should to the same. If in the future if you choose to write a piece with something like the premise "Is it possible rape is good?" Maybe phrase it in a more tangential way. Or state clearly that you DON'T think rape is good. Even if you don't think that this needs to be stated, consider it an inconvenient investment in the possibility that you'll decrease the odds of being called "the creepy economist."
Here is my interpretation. I am 99.99...% certain that the North winning the civil war was the correct course of action for people at the time to maximize the probability of making the US better off overall. I agree that it is possible that 10%, say, of the alternate histories where the South won could be better for the US. But these are different questions, for which I can make different estimates.
I'm not sure why you're saying ">50%". I don't think anyone has suggested that it's greater than 50%. The closest option in the original poll was ">10%".
I haven't read all the other comments on this post, but I assume the way you get to >50% is to assume that history is so contingent on accidents that it is basically random. Then comparing two histories is like comparing two random numbers, and yes, the chance that A > B is 50%.
But to get to that point, you really have to assume a lot. Assuming that the Holocaust makes no difference ultimately is more of a leap than I'm willing to make. So if you ask the question to someone with a more common-sense view of history, then no, I don't think you hop in your time machine and make sure the Nazis win because hey, it's just a roll of the dice and the result might be better.
It would make an interesting (grim) SF story for a time traveler from the future Confederacy (son of a slave, perhaps) to go back and make sure the Union wins. Then he looks at the new history, sees WWII and the Holocaust and changes it back....
I don't think anyone is contending that the probability of better outcomes with a Nazi victory is *greater than 50%*. I think we all agree that it's probably better that the Nazis lost.
The question is about the amount of uncertainty involved (variance) and the behavior of those uncertanties over the relevant time spans. If the question also stipulated that the Sun was going to explode in 1946 destroying the entire solar system, then I think everyone would give an answer in excess of 90%.
But as time goes on, uncertainty compounds. If we are in a position to ask this question again in 1000 years, the only acceptable answer should be "very nearly 50%".
I think the lesson to be learned is that the moral lens is about the narratives we tell ourselves, while the actual world is a chaotic, tremendously complex system. And complex systems to not map well onto narratives.
But almost all of our predictions of physical events (both macroscopic and microscopic) are fundamentally averages of an infinite number of possible paths.
You say that it only makes sense if the better outcomes are *because* of the South won, as though that isn't plausible. And then immediately afterwards you suggest a plausible causal chain of events that results in better outcomes *because* the South won.
You also demonstrated that you understand the question can be interpreted in an altogether different way ("in the set of all possible worlds, what are the chances that the ones with a Nazi win are better than the current situation?"). You even went so far as to argue that, with this alternate interpretation, Robin's most controversial statement ("If you knew enough history you’d see >10% as the only reasonable answer ...") is justified ("Given the range of possible histories, any random criteria will select worlds some of which may be better than this one").
The question is, would you have any of these (apparently conflicted) objections if the original hypothetical didn't have the same kind of moral flavoring?
I'm not sure what he meant! But if he meant what I said, then yes, you are right and the question is confusing or even not phrased correctly.
That being said, what I proposed isn't meaningless as a question!
Your second paragraph, speaking of my interpretation, is entirely correct. The fact that the event is historically significant and emotionally laden does not chance the fact that maybe outcomes are still possible.
In fact, if I read Robin's post correctly, *I think* that's his point: even if something awful happened in the past, there is a non-trivial chance that things might still have turned as well or better than they actually have. That's interesting and unintuitive!
If you phrase the question as "...the world would have been better off if the Nazis had won WWII", you are talking about cause and effect.
You seem to think the question is "in the set of all possible worlds, what are the chances that the ones with a Nazi win are better than the current situation." As if you were asking "in the set of possible worlds, in the ones where I don't bother to reply, what chance of a world better than this one?"
Given the range of possible histories, any random criteria will select worlds some of which may be better than this one. But they aren't better because of that criteria. The worlds aren't better because the South won or the Nazis won.
Maybe if the South had won, the U.S. (North) would have been weaker militarily and never entered WWI, which would have had a different outcome and not produced WWII, and no Nazis. So what?
The question is meaningless unless you interpret it as "due to the South winning, there's a better set of outcomes." And I can't see that happening.
The more plausible scenarios seem to be 'accelerationist' ones.
I.e consider that Germany wins, take over much of Europe, and then gets overthrown by some anti-fascist European revolution which leads to all of Europe being some sort of relative utopia, free from the Cold War.
Or consider that the South wins, institutes some awful regime, and then the end of slavery comes from some anti-slaveholder cross-racial revolt, which institutes some near utopia in the U.S.
Thinking about this more there is an essential vagueness in the question. I could be asking something like: "If we could time travel and kill Hitler in way X that would save 1 billion utils but we'd have to reroll all subsequent years what's the chance that things would turn out better." This question just turns on the model we use for `rerolling' later years.
However, I think it's actually more common for people to understand the question as asking something like: how confident are you that (evaluated with respect to objective physical probability) that E(X| no nazis) > E(X| nazis) where X is sum of utils in years since 1940. This question doesn't depend at all on how variable the utils earned per year might be since that is averaged out in the expectation. It merely asks us to judge how confident we are in our assumption that on net the nazis were bad rather than say necessary to warn us off a much more destructive nuclear war that would have almost inevitably occurred otherwise.
All this depends greatly on the notion of probability involved. If we mean objective quantum probability it could well turn out that (surprisingly unbeknownst to us) there was a 90% chance of nuclear apocalypse given no nazis. If you have in mind some more subjective epistemic notion than it's not at all clear why this shouldn't be almost 1.
I'm not even sure what you mean by the variance here since we are asked about the probability of a simple binary outcome (the world is either better or worse as a result of the Nazi's influence).
As for the effect after 1000 years I think your wrong. Even being super generous and assuming each year produces X utils ~ N(0,1) i.i.d. and the nazis imposed a 50 util cost (no worse than switching from an average to a 20th percentile year) it's still a 95% chance that a randomly chosen no nazi scenario beats a randomly chosen nazi scenario. If we assume, more accurately, that there are downstream effects of the previous year, e.g., X_i ~ X_{i-1} * e^{N(0,1)} then it would actually increase the difference in expectations over time.
However, all of this presumes a very unnatural, IMO, interpretation of what we mean when we ask the question.
Yes, and 'we' in the post just means the humans who would be alive in the designated scenario. It's not that unusual to use words like 'we' to refer to a group even if it's individual members are likely to change. Think of a basketball player saying "We'll do much better next year if you trade me to Chicago for Michael Jordan." That makes sense even though that player won't be part of the 'we' next year if the trade is made.
So what if the world doesn't have goals/preferences. When I use that phrase (and I think when Hanson does too) I have in mind something like 'increases net utility/happiness' and surely you believe there is at least some vague sense in which I can compare overall degrees of suffering/pleasure.
People seem to be assuming without justification that it is better to make a less offensive statement than a more offensive statement. The alternative hypothesis is that people are better off if they contemplate ten offensive things before breakfast, perhaps because that would let them get through the rest of their day without having their buttons so easily pushed by others. In this case people writing offensive things would be providing a service to people doing this breakfast contemplation, or other people who would benefit from the breakfast contemplation but are unaware of this. I do not know which is true and would be interested to see evidence either way.
A substantial fraction of misinterpretation may be done on purpose, in order to discredit me, because I'm seen as being politically right.
Robin! You frequently seem to run afoul of people who misinterpret your meanings and motivations. Why don't you take some time to do a little table setting so that people are less likely to pre-judge or mis-judge you?
Example. Sometimes I will tell a story and mention the race of a person in the story. I don't want to seem or be seen as the kind of person who specifies the race of every person in a story, so I'll make an aside--"I mention the race here because it becomes important later." This is a small investment to make, though unfortunate that it needs to be made.
I have friend who frequently wears soiled pants and complains that girls aren't interested in him. When I say, "Man, don't wear sweatpants! And if you do, make sure they are clean." He'll usually say, "Women should like me for who I am! Why does it matter what kind of pants I have?" I could argue that the women are making a good choice based upon the limited info they have. But whether or not they should "like him for him", he should simply do the things that can most foerseeably lead to good outcomes. And try to avoid easily foreseeable bad outcomes.
You should to the same. If in the future if you choose to write a piece with something like the premise "Is it possible rape is good?" Maybe phrase it in a more tangential way. Or state clearly that you DON'T think rape is good. Even if you don't think that this needs to be stated, consider it an inconvenient investment in the possibility that you'll decrease the odds of being called "the creepy economist."
Here is my interpretation. I am 99.99...% certain that the North winning the civil war was the correct course of action for people at the time to maximize the probability of making the US better off overall. I agree that it is possible that 10%, say, of the alternate histories where the South won could be better for the US. But these are different questions, for which I can make different estimates.
I'm not sure why you're saying ">50%". I don't think anyone has suggested that it's greater than 50%. The closest option in the original poll was ">10%".
Any chance you misread Robin's original post?
People often talk nonsense.
I haven't read all the other comments on this post, but I assume the way you get to >50% is to assume that history is so contingent on accidents that it is basically random. Then comparing two histories is like comparing two random numbers, and yes, the chance that A > B is 50%.
But to get to that point, you really have to assume a lot. Assuming that the Holocaust makes no difference ultimately is more of a leap than I'm willing to make. So if you ask the question to someone with a more common-sense view of history, then no, I don't think you hop in your time machine and make sure the Nazis win because hey, it's just a roll of the dice and the result might be better.
It would make an interesting (grim) SF story for a time traveler from the future Confederacy (son of a slave, perhaps) to go back and make sure the Union wins. Then he looks at the new history, sees WWII and the Holocaust and changes it back....
I don't think anyone is contending that the probability of better outcomes with a Nazi victory is *greater than 50%*. I think we all agree that it's probably better that the Nazis lost.
The question is about the amount of uncertainty involved (variance) and the behavior of those uncertanties over the relevant time spans. If the question also stipulated that the Sun was going to explode in 1946 destroying the entire solar system, then I think everyone would give an answer in excess of 90%.
But as time goes on, uncertainty compounds. If we are in a position to ask this question again in 1000 years, the only acceptable answer should be "very nearly 50%".
I think the lesson to be learned is that the moral lens is about the narratives we tell ourselves, while the actual world is a chaotic, tremendously complex system. And complex systems to not map well onto narratives.
But almost all of our predictions of physical events (both macroscopic and microscopic) are fundamentally averages of an infinite number of possible paths.
You say that it only makes sense if the better outcomes are *because* of the South won, as though that isn't plausible. And then immediately afterwards you suggest a plausible causal chain of events that results in better outcomes *because* the South won.
You also demonstrated that you understand the question can be interpreted in an altogether different way ("in the set of all possible worlds, what are the chances that the ones with a Nazi win are better than the current situation?"). You even went so far as to argue that, with this alternate interpretation, Robin's most controversial statement ("If you knew enough history you’d see >10% as the only reasonable answer ...") is justified ("Given the range of possible histories, any random criteria will select worlds some of which may be better than this one").
The question is, would you have any of these (apparently conflicted) objections if the original hypothetical didn't have the same kind of moral flavoring?
I'm not sure what he meant! But if he meant what I said, then yes, you are right and the question is confusing or even not phrased correctly.
That being said, what I proposed isn't meaningless as a question!
Your second paragraph, speaking of my interpretation, is entirely correct. The fact that the event is historically significant and emotionally laden does not chance the fact that maybe outcomes are still possible.
In fact, if I read Robin's post correctly, *I think* that's his point: even if something awful happened in the past, there is a non-trivial chance that things might still have turned as well or better than they actually have. That's interesting and unintuitive!
If you phrase the question as "...the world would have been better off if the Nazis had won WWII", you are talking about cause and effect.
You seem to think the question is "in the set of all possible worlds, what are the chances that the ones with a Nazi win are better than the current situation." As if you were asking "in the set of possible worlds, in the ones where I don't bother to reply, what chance of a world better than this one?"
Given the range of possible histories, any random criteria will select worlds some of which may be better than this one. But they aren't better because of that criteria. The worlds aren't better because the South won or the Nazis won.
Maybe if the South had won, the U.S. (North) would have been weaker militarily and never entered WWI, which would have had a different outcome and not produced WWII, and no Nazis. So what?
The question is meaningless unless you interpret it as "due to the South winning, there's a better set of outcomes." And I can't see that happening.