Short form (might be enough to understand my post ): If my valuation/measurement of different hedonic states is far off the average valuation, should I update for regression to the mean?

Hey y'all,

First of all I don't want to start another epic-length discussion on NU/NLU/CU/PU. Reading through all the discussions I made this little theory though and want to know if I'm making any sense. I don't know myself how much I value suffering vs. happiness (and intense vs. light versions) yet and I'm not even sure if I'm in a position to judge their value. I'm not sure either if my theory means anything when we deny moral realism because of Occam's Razor.

So here it goes. Our moral valuation of different hedonic states is based entirely on our remembered perception. Thus, the only measure we have for positive and negative moral value is (to this date) perception. If our perception said that a pin prick was worse than burning in hell then that would be our moral valuation of the two. If our perception says that an orgasm doesn't make up for a pin prick then that is our relative moral valuation of the two. In the latter case it seems reasonable to consider yourself a negative leaning utilitarian (at first sight), because your perception is far off that of the average person. We have to keep in mind that our moral valuation is based on our subjective (remembered) perception. Our perception is our only measurement.

As someone who makes a measurement that is far off the average perception/measurement one ought to correct for regression to the mean though. One should not think that one's own perception that is far off the middle of, say, the bell curve of perceptions was better than the other measurements. Thus one should update accordingly. Unless one has reason to believe that the other measurements are systematically flawed. Examples of this are mad in Daniel Kahneman's work (see e.g. "The riddle of experience vs memory" on TED).

One may also add that the perceptions of non-human animals are not included in what most humans would consider average measurements of hedonic states. There are large amounts of suffering by particularly large amounts of individuals taking place in the wild. This is getting speculative but still, they may take a pin prick for an orgasm and exchange some short, intense form of suffering for a long, mild one. One would still have to update one's measurements for regression to the mean if one is an observer with perceptions far off the mean.

This seems particularly relevant if one makes a decision for others. Taking suffering in exchange for happiness or intense suffering in exchange for a prolonged period of light suffering would be examples of such decisions.

Amongst other things I'm really not sure whether one can "measure" hedonic states. But doesn't it make sense that one should update if one's own perceptions were "outliers"? If yes, this could have implications for how I value different perceived hedonic states.

As someone who makes a measurement that is far off the average perception/measurement one ought to correct for regression to the mean though. One should not think that one's own perception that is far off the middle of, say, the bell curve of perceptions was better than the other measurements.

I think you would need to defend this more. Are there examples of humans being off on their subjective perceptions and that people become more accurate as they correct toward the mean?

I don't know if I disagree with you, but one alternative argument is that our personal experience of pain is exceptionally vivid and informed. Also, there might be reason to expect that your feelings of pain are different from others. For example, I think I have less of a tolerance for heat than average and therefore suffer slightly more than other people when asked to, say, remove something from an oven.

I would really like to know how to quantify agony. Or similarly extreme suffering.

To quote an anecdote from another thread: A friend of mine had painful kidney stones and had to undergo a painful removal procedure while conscious. She described the pain as the worst she'd ever felt. I asked her how much money she would need to be offered to agree to endure 10 additional seconds of this pain, all else equal, and her answer was half a million euros. She actually endured the pain much longer than 10 seconds during the procedure.

Most normal (good and bad) hedonic states are quite quantifiable and fit on a range where any day is no more than 5-10 times as good or bad as a roughly average day. Nothing that doesn't fit on a -5 to +5 scale. So there are differences, but they aren't extreme. (I measured this daily for over a year for myself, just for fun, integrating quite a wide range of experiences)

But then I get evidence like the anecdote above and it seems to shatter the scale. If we can measure anything, how can we measure this? What's the most common scientific approach?

I think you would need to defend this more. Are there examples of humans being off on their subjective perceptions and that people become more accurate as they correct toward the mean?

I couldn't name any and yes, it's not a very robust theory. If people became objectively more accurate by correcting to the mean we wouldn't know though because we can't measure it. The study I describe below might help with this when making some assumptions.

Hedonic Treader wrote:I would really like to know how to quantify agony. Or similarly extreme suffering.

It seems to me that we have some inaccurate inbuilt measuring system that allows us to distinguish light from extreme suffering. Similarly it lets us distinguish pleasure from pain and make comparisons. At that point it may get even more inaccurate. I find this question of measuring hedonic states quite interesting and there doesn't seem to be a simple solution.

I don't know enough about neuroscience but maybe different intensities of suffering and happiness are reflected in the brain? I presume that Peter's sensitivity to heat would also be reflected and hopefully measurable in his brain state then. I also assume that the same amount of brain activity in individuals with same brain size means roughly the same experience which is debatable.

In that case a study could be designed like this: You expose subjects to slighter and more extreme (mwahahaha) amounts of suffering and pleasure and measure their brain states. Then you ask them to say which amounts of suffering they would have traded for which amount of pleasure. This would allow you to make some nice bell curves and see if the results strongly vary. Subjects who would trade pleasure that corresponds with a certain brain state for an unusually large amount of suffering (and the other way round) may somehow undervalue suffering under the assumptions I made. The other way round they could be unusually "suffering averse", that is, although their brain states are of the same intensity their valuation of suffering is larger. This could be described as irrational and possibly based on preconceived beliefs about the importance of suffering or pleasure.
I actually get this a bit more every time I watch a video of animal suffering or hear an anecdote like Hedonic Treader's. While I'm very happy I also tend to care more about great bliss.

Similarly you could ask which amount of intense pleasure/suffering they would trade for a slighter amount of longer duration. This would only work if you found a way to correct for "duration scope neglect" which is very relevant for hedonic states.

Of course this makes some assumptions. Maybe the study faces more difficulties that I haven't thought of.

One outcome might be that our valuation of hedonic states is different from our actual experience. If our valuation is far off the mean it may have to be corrected.

But then I get evidence like the anecdote above and it seems to shatter the scale.

A big difficulty should be that pleasure of this scale is just not available in biological organisms. Maybe one could get closer with wireheading..

one alternative argument is that our personal experience of pain is exceptionally vivid and informed

True, that's why caring about animal suffering probably makes you more accurate. Many seem to be blind for it. One question for NLUs or NUs: Do you apply very different exchange rates for everyday situations like stubbing your toe, pin pricks and orgasms or is it more about issues that other people are less informed about? Things like kidney stone removal, burning alive, starving, some other things that animals experience.

Sören Mind wrote:A big difficulty should be that pleasure of this scale is just not available in biological organisms. Maybe one could get closer with wireheading..

Presumably, you can only wirehead what the system is already capable of representing. If the system naturally can't represent such pleasure, then you probably can't wirehead it.

However, it is evident human brains can already represent considerable pleasure. It is also probable that there will be technology that allows people to literally change their minds in specific ways. A probable scenario for this would be after a substrate change to digital.

If digital whole-brain emulations become real, there will be the technological option to copy them in a mature state, back them up, manipulate synapse strengths, grow new (emulated) neurons, possibly implement entirely new cortical areas. Not even skull size is a limit anymore then. There would probably be a cambrian radiation type event in mind space, starting with original human brains. Something similar could be done with AGI designed from scratch, or theoretically even with biological organisms or 'cyborgized' brains.

Then at least in theory, pleasures that cannot be represented now can be represented and optionally wireheaded.

Hedonic Treader wrote:I would really like to know how to quantify agony. Or similarly extreme suffering.

To quote an anecdote from another thread: A friend of mine had painful kidney stones and had to undergo a painful removal procedure while conscious. She described the pain as the worst she'd ever felt. I asked her how much money she would need to be offered to agree to endure 10 additional seconds of this pain, all else equal, and her answer was half a million euros. She actually endured the pain much longer than 10 seconds during the procedure...

But then I get evidence like the anecdote above and it seems to shatter the scale. If we can measure anything, how can we measure this? What's the most common scientific approach?

Looks more like problems with response to abstract hypothetical situations, without the opportunity for extensive empirical feedback. Consider these results from the classic Cognitive Reflection Test paper (http://test-nexus.som.yale.edu/cci/site ... Making.pdf):

In the group that performed poorly on the cognitive reflection test, only 40% of men and 25% of women said they would take a 15% chance of $1,000,000 over a sure $500. In the high-scoring group, 80% of men and 38% of women said they would take the 15% chance of a million dollars.

In a comparison of a 100% chance of $1,000 and a 90% chance of $5,000, only 59% of men and 46% of women in the low-CRT group, and 81% of men and 59% of women in the high-CRT group, said they would take the risky option.

Given empirical feedback, actual belief in the payoffs, a chance to learn by imitation (if such gambles were regularly available all one's role models in wealth would have taken such risks), and better understanding almost everyone would take these slightly risky but huge EV gambles. But one can get nonsense answers this way.

michal123 wrote:If there is no arbitration, then you have infinite regression.

Haha such complicated words can only come from great minds

I think one fundamental question is, which do you think is more likely:

1. Your terminal exchange rate is closer to the mean exchange rate than your belief of your terminal of exchange rate.

Or,

2. Your terminal exchange rate is closer to your belief of your terminal of exchange rate than the mean exchange rate.


If 1 is true, regress, if 2, don't?

Ruairi wrote:I think one fundamental question is, which do you think is more likely:

1. Your terminal exchange rate is closer to the mean exchange rate than your belief of your terminal of exchange rate.

Or,

2. Your terminal exchange rate is closer to your belief of your terminal of exchange rate than the mean exchange rate.


If 1 is true, regress, if 2, don't?

Thanks I think this describes pretty well why I find this confusing from a moral anti-realist perspective. Hits it on the spot.

I would add that what I wrote in the original post is not a very accurate respresentation of what I think now anymore. My best guess is that morally its a question of which computations you care about how much. And that empirically one could be off in the way you describe in point 1), for example when you over- or underestimate more or less extreme hedonic states. That could be corrected on the basis of research we'll hopefully have in the future or perhaps even by regression to the mean. But whether 'the mean' is a reliable place to regress to is unclear so I'm not sure if what I wrote has any practical relevance.

Btw I'd be interested to hear more anecdotes or preferably evidence about how painful extreme suffering really is compared to say the day to day suffering of humans, farmed and wild animals.

One of Kahneman's studies concluded that we underestimate the relevance of duration, so that could be a reason to update one's intensity-duration exchange rate.

Sören Mind wrote:I would add that what I wrote in the original post is not a very accurate respresentation of what I think now anymore. My best guess is that morally its a question of which computations you care about how much.

Yeah, my main quibble with the post was going to be that "perception" is not quite the right way to think about it, because it's not like measuring a physical quantity where there's a right answer that people are trying to converge on; rather, what you think is the answer is the answer, at least in the moment of experiencing it. That said, whether you assess beforehand the same thing you feel in the moment is an important question, but in general, organisms' preferences are not stable over time. Bad enough torture and you'll forgo anything to make it stop; great enough pleasure and you may accept anything to make it continue (though I'm not sure the latter is possible, or at least as common, for humans).

Sören Mind wrote:One question for NLUs or NUs: Do you apply very different exchange rates for everyday situations like stubbing your toe, pin pricks and orgasms or is it more about issues that other people are less informed about? Things like kidney stone removal, burning alive, starving, some other things that animals experience.

Only for the extremes. I have rarely in my life experienced really severe suffering.

I don't think starving is that bad until maybe the end when you actually die. Of course, cumulatively it's not fun, but given that I weight severe pains as much worse than moderate ones, it takes a lot of starvation to add up to a few seconds of a different way of dying.

Sören Mind wrote:Btw I'd be interested to hear more anecdotes or preferably evidence about how painful extreme suffering really is compared to say the day to day suffering of humans, farmed and wild animals.

Me too!