Revisiting torture vs. dust specks

cousin_it

In line with my fine tradition of beating old horses, in this post I'll try to summarize some arguments that people proposed in the ancient puzzle of Torture vs. Dust Specks and add some of my own. Not intended as an endorsement of either side. (I do have a preferred side, but don't know exactly why.)

The people saying one dust speck is "zero disutility" or "incommensurable utilities" are being naive. Just pick the smallest amount of suffering that in your opinion is non-zero or commensurable with the torture and restart.
Escalation argument: go from dust specks to torture in small steps, slightly increasing the suffering and massively decreasing the number of people at each step. If each individual change increases utility, so does the final result.
Fluctuation argument: the probability that the universe randomly subjects you to the torture scenario is considerably higher than 1/3^^^3 anyway, so choose torture without worries even if you're in the affected set. (This doesn't assume the least convenient possible world, so fails.)
Proximity argument: don't ask me to value strangers equally to friends and relatives. If each additional person matters 1% less than the previous one, then even an infinite number of people getting dust specks in their eyes adds up to a finite and not especially large amount of suffering. (This assumption negates the escalation argument once you do the math.)
Real-world analogy: we don't decide to pay one penny each to collectively save one starving African child, so choose torture. (This is resolved by the proximity argument.)
Observer splitting: if you split into 3^^^3 people tomorrow, would you prefer all of you to get dust specks, or one of you to be tortured for 50 years? (This neutralizes the proximity argument, but the escalation argument also becomes non-obvious.)

Oh what a tangle. I guess Eliezer is too altruistic to give up torture no matter what we throw at him; others will adopt excuses to choose specks; still others will stay gut-convinced but logically puzzled, like me. The right answer, or the right theory to guide you to the answer, no longer seems so inevitable and mathematically certain.

Edit: I submitted this post to LW by mistake, then deleted it which turned out to be the real mistake. Seeing the folks merrily discussing away in the comments long after the deletion, I tried to undelete the post somehow, but nothing worked. All right; let this be a sekrit area. A shame, really, because I just thought of a scenario that might have given even Eliezer cause for self-doubt:

Observer splitting with a twist: instead of you, one of your loved ones will be split into 3^^^3 people tomorrow. Torture a single branch for 50 years, or give every branch a dust speck?

The people saying one dust speck is "zero disutility" or "incommensurable utilities" are being naive. Just pick the smallest amount of suffering that in your opinion is non-zero or commensurable with the torture and restart.
Escalation argument: go from dust specks to torture in small steps, slightly increasing the suffering and massively decreasing the number of people at each step. If each individual change increases utility, so does the final result.
Fluctuation argument: the probability that the universe randomly subjects you to the torture scenario is considerably higher than 1/3^^^3 anyway, so choose torture without worries even if you're in the affected set. (This doesn't assume the least convenient possible world, so fails.)
Proximity argument: don't ask me to value strangers equally to friends and relatives. If each additional person matters 1% less than the previous one, then even an infinite number of people getting dust specks in their eyes adds up to a finite and not especially large amount of suffering. (This assumption negates the escalation argument once you do the math.)
Real-world analogy: we don't decide to pay one penny each to collectively save one starving African child, so choose torture. (This is resolved by the proximity argument.)
Observer splitting: if you split into 3^^^3 people tomorrow, would you prefer all of you to get dust specks, or one of you to be tortured for 50 years? (This neutralizes the proximity argument, but the escalation argument also becomes non-obvious.)

Observer splitting with a twist: instead of you, one of your loved ones will be split into 3^^^3 people tomorrow. Torture a single branch for 50 years, or give every branch a dust speck?

There's another argument I think you might have missed:

Utilitarism is about being optimal. Instinctive morality is about being failsafe.

Implicit in all decisions is a nonzero possibility that you are wrong. Once you take that into account, having some "hard" rules like not agreeing to torture here (or in other dilemmas), not pushing the fat guy on the tracks in the trolley problem, etc, can save you from making horrible mistakes at the cost of slightly suboptimal decisions. Which is, incidentally, how I would want a friendly AI to decide as well - losing a bit in the average case to prevent a really horrible worst case.

That rule alone would, of course, make you vulnerable to Pascal's Mugging. I think the way to go here is to have some threshold at which you round very low (or very high) probabilities off to zero (or one) when the difference is small against the probability of you being wrong. Not only will this protect you against getting your decisions hacked, it will also stop you from wasting computing power on improbable outcomes. This seems to be the reason why Pascal's Mugging usually fails on humans.

Both of these are necessary patches because we operate on opaque, faulty and potentially hostile hardware. One without the other is vulnerable to hacks and catastrophic failure modes, but both taken together are a pretty strong base for decisions that, so far, have served us humans pretty well. In two rules:

1) Ignore outcomes to which you assign a lower probability than to you being wrong/mistaken about the situation. 2) Ignore decisions with horrible worst case scenarios if there are options with a less horrible worst case and still acceptable acceptable average case.

When both of these apply to the same thing, or this process eliminates all options, you have a dilemma. Try to reduce your uncertainty about 1) and start looking for other options in 2). If that is impossible, shut up and do it anyway.

10

Revisiting torture vs. dust specks

10

10

10

Revisiting torture vs. dust specks

10

10