Thank you! I will get back to you. My honest reaction to this comment: "AAAAAHHHH damn, this should be so incredibly obvious!! where to even start?" But I recognize I'm deep into the inside view of log scales, phenomenology visualization, valence quantification, etc. What seems mind-numbingly foot-stompingly obvious to me might not be to others. Do I say: "PLEASE BE CREATIVE" or should I write a treaties expanding on every point? I'll do the latter. And I apologize for my candid reaction comment XD Seriously, I appreciarte the comment. It's just... ahhhhhh! 

[to be clear you raise very good points - I'm just trying to do the prosocial thing and communicate the large "inference gap" on this topic that currently exists - I'll do better! Will respond object level. Cheers!]

Thank you! This is a genuinely good question. (Note: I answered via voice and then edited the transcript below with Chat - can circle back if style is an issue, but this covers every point I discussed - if doing this is a problem for some reason I'm happy to write anew! The content is correct): 

Your question surfaces the key misunderstanding. The claim isn’t that we should fear drugs that help some people and hurt others. It’s that our measurement architecture is set up in a way that systematically misclassifies who is helped, who is harmed, and by how much, because the scales themselves flatten the underlying geometry of experience. Once you compress long-tailed intensities into a 1–10 box and then average them, you lose the structure that actually matters for real-world well-being.

In a world where symptoms behave linearly and add up nicely, a drug that helps half and hurts half is perfectly intelligible: you imagine two overlapping Gaussians, shrug, and say “worth a try.” But that isn’t the world we actually inhabit. If rumination goes from 6 to 4, the subjective win might be modest because you’re moving along a shallow part of the curve. If akathisia goes from 6 to 8, the subjective loss might be massive because you’ve crossed into a steep tail where each step carries exponential experiential weight. On the form, these are both “two-point changes.” In lived reality, they belong to different moral universes. This asymmetry in the tails means that “50% better, 50% worse” is not a neutral mixture; the average hides the fact that the extremes on one side can dominate the arithmetic.

I don't think it is abstract or merely theoretical, or too complex to do anything about. It has immediate practical consequences. Trials and regulators work with compressed reports, so the deepest harms appear as mild perturbations in the dataset. Drugs whose side-effect profiles involve steep-tail states like akathisia or mixed autonomic rebound look safer than they really are for a meaningful minority of users. Clinicians then inherit an evidence base where the worst experiential states have been squashed into “mild adverse events,” and that shapes expectations, heuristics, and prescribing norms. The problem is not clinician negligence — it’s that the underlying data they rely on has already thrown away the signal.

If we took the geometry seriously, we’d end up with a very different picture. High-variance drugs can be extraordinarily useful when we know how to identify responders and anti-responders. What we’re missing is the mapping. With better instruments, you’d get early detection of bad trajectories, N-of-1 response curves, and a more honest sense of which symptom profiles are compatible with which medications. The same drug could be life-changing for one subgroup and acutely harmful for another, and we could actually see that, instead of blending the two together into a 0.3σ effect size. This is less “anti-medication” and more “finally doing the epistemology correctly.”

Good clinical practice already tries to rely on patient narratives, but even that is downstream of the larger culture of interpretation we’ve built on top of flattened scales. When the scientific literature underweights the steepest affective states, everyone downstream learns to underweight them too. The patient who says “this made my inner restlessness unbearable” is intuitively competing with a literature that reports only “mild activation” for the same phenomenon. Countless examples of victims of this dynamic can be mentioned and documented.

The upshot is simple: this not trying to be pessimistic toward psychiatric meds. The core point is about epistemic clarity. The experiential landscape is long-tailed, clustered, and nonlinear; our measurement system is linear, additive, and tidy. When you force one onto the other, you get averages that obscure the very variation we need to guide good decisions. A better measurement pipeline wouldn’t make us more cautious or more reckless; it would make us more accurate. And accuracy is the only way to use high-variance interventions wisely — whether you’re trying to help one patient or setting policy for millions.

If the world ran fully on the arithmetic of symptom sheets, none of this would matter. But the world runs on compounding long-tail distributions of suffering and relief, and that geometry is strange, heavy-tailed, and morally lopsided. Our tools need to catch up.

Indeed, we're talking about 3mg vaporized for the typical sufferer. I've interviewed people who indeed described it as "from 10/10 pain to 1/10 or 0/10 withon 10 seconds" and they're not exagerating. Percent who responds this way? Likely 70% or so. For the rest they need higher doses.

I think DMT is also only psychologically risky above Magic Eye level, for which you need to take quite a bit more than the cluster headache range. But STILL, the worst DMT bad trip is still orders of magnitude better than a cluster headache. Everyone I've talked to who uses DMT for clusters says they wished they had tried it earlier. From experience, even high dose DMT is physically not as unpleasant as breaking a bone. Like, burning sensations and feelings of intense pressure. Unpleasant, but not that bad, really. 

Note, MAOI might actually deter against DMT's effect on clusters. It works best when no other medication or drug in your system. 

Very good points. QRI has a lot to say about all of these points, so I won't repeat myself too much. I'll address a couple points, but note there is a lot more to say that you can dig into in the provided links:

I think that pain is a particular manifestation of negative valence. It can trigger positive valence indirectly via e.g. energizing the system and then triggering neural annealing - nice waves of euphoria that are secondary and after the fact which are the reinforcing bits. Pain sans this secondary element is just unpleasant and bad, albeit perhaps not as bad as what you get when you mix both emotional and low level sensory negative valence together. 

Importantly: mixed valence certainly complicates the picture, but above a certain intensity of sensation pain overwhelms whetever else is going on.

Shinzen Young's concept of equanimity as not resisting sensations is a critical modifier on the valence of experience. That said, I'd say this modifies valence of the whole experience, and doesn't necessarily take care of the low-level sensory negative valence. Stil, the bulk of a person's valence under normal circumstances might be the result of how much they resist their current experience (even if pleasant otherwise). Extreme levels of equanimity, I'm convinced, can drastically lower the negative valence effects of pain. 

This, however, has a limit. Even highly attained meditators (the "Buddha" included) would describe extreme pain as still a cause of suffering. Daniel Ingram, for example, can tolerate broken bones and all kinds of very intense painful sensations without them turning into suffering, so to speak. But when he has a kidney stone then that overwhelms the system and even his meditation attainments aren't enough to counter-balance it. He still suffers intensely with kidney stones when they happen.

Lastly, most intense forms of pain tend to have a strong emotional component by default. Cluster headaches, for example, typically come with a powerful sense of doom along with the pain. It's just part of the package it seems.

https://qualiacomputing.com/2021/04/04/buddhist-annealing-wireheading-done-right-with-the-seven-factors-of-awakening/

https://forum.effectivealtruism.org/posts/bvtAXefTDQgHxc9BR/just-look-at-the-thing-how-the-science-of-consciousness

https://qri.org/blog/symmetry-theory-of-valence-2020

https://qualiacomputing.com/2019/09/30/harmonic-society-3-4-art-as-state-space-exploration-and-energy-parameter-modulation/

Thank you for this fascinating post. I'll share here what I posted on Twitter too:

I have many reasons why I don't think we should care about non-conscious agency, and here are some of them:

1) That which lacks frame invariance cannot be truly real. Algorithms are not real. They look real from the point of view of (frame invariant) experiences that *interpret* them. Thus, there is no real sense in which an algorithm can have goals - they only look like it from our (integrated) point of view. It's useful for us, pragmatically, to model them that way. But that's different from them actually existing in any intrinsic substantial way.

2) The phenomenal texture of valence is deeply intertwined with conscious agency when such agency matters. The very sense of urgency that drives our efforts to reduce our suffering has a *shape* with intrinsic causal effects. This shape and its causal effects only ever cash out as such in other bound experiences. So the very _meaning_ of agency, at least in so far as moral intuitions are concerned, is inherently tied to its sentient implementation.

3) Values are not actually about states of world, and that is because states of the world aside from moments of experience don't really exist. Or at least we have no reason to believe they exist. As you increase the internal coherence of one's understanding of conscious agency, it becomes, little by little, clear that the underlying *referent* of our desires were phenomenal states all along, albeit with levels of indirection and shortcuts.

4) Even if we were to believe that non-sentient agency (imo an oxymoron) is valuable, we would have also good reasons to believe it is in fact disvaluable. Intense wanting is unpleasant, and thus sufficiently self-reflective organisms try to figure out how to realize their values with as little desire as possible.

5) Open Individualism, Valence Realism, and Math can provide a far more coherent system of ethics than any other combo I'm aware of, and they certainly rule out non-conscious agency as part of what matters.

6) Blindsight is poorly understood. There's an interesting model of how it works where our body creates a kind of archipelago of moments of experience, in which there is a central hub and then many peripheral bound experiences competing to enter that hub. When we think that a non-conscious system in us "wants something", it might very well be because it indeed has valence that motivates it in a certain way. Some exotic states of consciousness hint at this architecture - desires that seem to "come from nowhere" are in fact already the result of complex networks of conscious subagents merging and blending and ultimately binding to the central hub.

------- And then we have pragmatic and political reasons, where the moment we open the floodgates of insentient agency mattering intrinsically, we risk truly becoming powerless very fast. Even if we cared about insentient agency, why should we care about insentient agency in potential? Their scaling capabilities, cunning, and capacity for deception might quickly flip the power balance in completely irreversible ways, not unlike creating sentient monsters with radically different values than humans.

Ultimately I think value is an empirical question, and we already know enough to be able to locate it in conscious valence. Team Consciousness must wise up to avoid threats from insentient agents and coordinate around these risks catalyzed by profound conceptual confusion.

Thank you Gavin (algekalipso here).

I think that the most important EA-relevant link for #1 would be this: Logarithmic Scales of Pleasure and Pain: Rating, Ranking, and Comparing Peak Experiences Suggest the Existence of Long Tails for Bliss and Suffering 

For a summary, see: Review of Log Scales.

In particular, I do think aspiring EAs should take this much more seriously:

An important pragmatic takeaway from this article is that if one is trying to select an effective career path, as a heuristic it would be good to take into account how one’s efforts would cash out in the prevention of extreme suffering (see: Hell-Index), rather than just QALYs and wellness indices that ignore the long-tail. Of particular note as promising Effective Altruist careers, we would highlight working directly to develop remedies for specific, extremely painful experiences. Finding scalable treatments for migraines, kidney stones, childbirth, cluster headaches, CRPS, and fibromyalgia may be extremely high-impact (cf. Treating Cluster Headaches and Migraines Using N,N-DMT and Other Tryptamines, Using Ibogaine to Create Friendlier Opioids, and Frequency Specific Microcurrent for Kidney-Stone Pain). More research efforts into identifying and quantifying intense suffering currently unaddressed would also be extremely helpful. Finally, if the positive valence scale also has a long-tail, focusing one’s career in developing bliss technologies may pay-off in surprisingly good ways (whereby you may stumble on methods to generate high-valence healing experiences which are orders of magnitude better than you thought were possible).

Best,

Andrés :)

This post significantly adds to the conversation in Effective Altruism about how pain is distributed. As explained in the review of Log Scales, understanding that intense pain follows a long-tail distributions significantly changes the effectiveness landscape for possible altruistic interventions. In particular, this analysis shows that finding the top 5% of people who suffer the most in a given medical condition and treating them as the priority will allow us to target a very large fraction of the total pain such a condition generates. In the case of cluster headaches, the distribution is extremely skewed: 5% of sufferers experience over 50% of all cluster headaches. 

More so, the survey also showed that the leading cause for why sufferers don't use tryptamines to treat their condition is the difficulty of acquiring them. Thus, changing the legal landscape via e.g. providing programs for the easy access to tryptamines to sufferers of migraines and cluster headaches might be a very cost-effective way of massively reducing suffering throughout the world.

Zooming out, perhaps the significance of this goes beyond cluster headaches in particular: it perhaps hints at a more significant paradigmatic change for analyzing the cost-effectiveness of interventions.

As explained in the review of Log Scales, cluster headaches are some of the most painful experiences people can have in life. If a $5 DMT Vape Pen produced at scale is all it takes to fully take care of the problem for people sufferers, this stands to be an Effective Altruist bargain.

In the future, I would love to see more analysis of this sort. Namely, analysis that look at particular highly painful conditions (the "pain points of humanity", as it were), and identify tractable, cost-effective solutions to them. Given the work in this area so far, I expect this to generate dozens of interventions that, in aggregate, might take care of perhaps even the majority of dolors experienced by people.

Most people who know about drugs tend to have an intuitive model of drug tolerance where "what goes up must come down". In this piece, the author shows that this intuitive model is wrong, for drug tolerance can be reversed pharmacologically. This seems extremely important in the context of pain relief: for people who simply have no option but to take opioids to treat their chronic pain, anti-tolerance would be a game-changer. I sincerely believe this will be a paradigm shift in the world of pain management, with a clear before-and-after cultural shift around it. But before that, a lot of foundational research needs to take place. That's the stage we are at.

We anticipate and hope that the field of anti-tolerance drugs  soon materializes in an academically credible way. Given how common chronic pain is, we would all benefit from its fruits in the future.

I would like to suggest that Logarithmic Scales of Pleasure and Pain (“Log Scales” from here on out) presents a novel, meaningful, and non-trivial contribution to the field of Effective Altruism. It is novel because even though the terribleness of extreme suffering has been discussed multiple times before, such discussions have not presented a method or conceptual scheme with which to compare extreme suffering relative to less extreme varieties. It is meaningful because it articulates the essence of an intuition of an aspect of life that deeply matters to most people, even if they cannot easily put it into words. And it is non-trivial because the inference that pain (and pleasure) scales are better understood as logarithmic in nature does require one to consider the problem from multiple points of view at once that are rarely, if ever, brought together (e.g. combining deference analysis, descriptions of pain scales by their creators, latent-trait analysis, psychophysics, and so on). 

Fundamentally, we could characterize this article as a conceptual reframe that changes how one assesses magnitudes of suffering in the world. To really grasp the significance of this reframe, let’s look back into how Effective Altruism itself was an incredibly powerful conceptual reframe that did something similar. In particular, a core insight that establishes the raison d'etre of Effective Altruism is that the good that you can do in the world with a given set of resources varies enormously depending on how you choose to allocate it: by most criteria that you may choose (whether it’s QALYs or people saved from homelessness), the cost-effectiveness of causes seem to follow much more closely (at least qualitatively) a long-tail rather than a normal distribution (see: Which world problems are the most pressing to solve? by Benjamin Todd). In turn, this strongly suggests that investigating carefully how to invest one’s altruistic efforts is likely to pay off in very large ways: choosing a random charity versus a top 1% charity will lead to benefits whose scale differs by orders of magnitude.

Log Scales suggests that pain and pleasure themselves follow a long-tail distribution. In what way, exactly? Well, to a first approximation, across the entire board! The article (and perhaps more eloquently the subsequent video presentation at the NYC EA Meetup on the same topic) argues that when it comes to the distribution of the intensity of hedonic states, we are likely to find long-tails almost any way we choose to slice or dice the data. This is analogous to, for example, how all of the following quantities follow long-tail distributions: avalanches per country, avalanches per mountain, amount of snow in mountains, number of avalanche-producing mountains per country, size of avalanches, number of avalanches per day, etc. Likewise, in the case of the distribution of pain, the arguments presented suggest we will find that all of the following distributions are long-tails: average pain level per medical condition, number of intensely painful episodes per person per year, intensity of pain per painful episode, total pain per person during one’s life, etc. Thus, that such a small percentage of cluster headache patients accounts for the majority of episodes per year would be expected (see: Cluster Headache Frequency Follows a Long-Tail Distribution), and along with it, the intensity of such episodes themselves would likely follow a long-tail distribution.

This would all be natural, indeed, if we consider neurological phenomena such as pain to be akin to weather phenomena. Log Scales allows us to conceptualize the state of a nervous system and what it gives rise to as akin to how various weather conditions give rise to natural disasters: a number of factors multiply each other resulting in relatively rare, but surprisingly powerful, black swan events. Nervous systems such as those of people suffering from CRPS, fibromyalgia, and cluster headaches are like the Swiss Alps of neurological weather conditions… uniquely suited for ridiculously large avalanches of suffering.

Log Scales are not just of academic interest. In the context of Effective Altruism, they are a powerful generator for identifying new important, neglected, and tractable cause areas to focus on. For instance, DMT for cluster headaches, microdose ibogaine for augmentation of painkillers in sufferers of chronic pain, and chanca piedra for kidney stones (writeup in progres) are all what we believe to be highly promising interventions (of the significant, neglected, and tractable variety) that might arguably reduce suffering in enormous ways and that would not have been highlighted as EA-worthy were it not for Log Scales. (See also: Get-Out-Of-Hell-Free Necklace). On a personal note, I’ve received numerous thank you notes by sufferers of extreme pain for this research. But the work has barely begun: with Log Scales as a lens, we are poised to tackle the world’s reserves of suffering with laser-focus, assured in the knowledge that preventing a small fraction of all painful conditions is all that we need to abolish the bulk of experiential suffering.

But does Log Scales make accurate claims? Does it carve reality at the joints? How do we know?

The core arguments presented were based on (a) the characteristic distribution of neural activity, (b) phenomenological accounts of extreme pleasure and pain, (c) the way in which the creators of pain scales have explicitly described their meaning, and (d) the results of a statistical analysis of a pilot study we conducted where people ranked, rated, and assigned relative proportions to their most extreme experiences. We further framed this in terms of comparing qualitative predictions from what we called the ​​Normal World vs. Lognormal World. In particular, we stated that: “If we lived in the ‘Lognormal World’, we would expect: (1) That people will typically say that their top #1 best/worst experience is not only a bit better/worse than their #2 experience, but a lot better/worse. Like, perhaps, even multiple times better/worse. (2) That there will be a long-tail in the number of appearances of different categories (i.e. that a large amount, such as 80%, of top experiences will belong to the same narrow set of categories, and that there will be many different kinds of experiences capturing the remaining 20%). And (3) that for most pairs of experiences x and y, people who have had both instances of x and y, will usually agree about which one is better/worse. We call such a relationship a ‘deference’. More so, we would expect to see that deference, in general, will be transitive (a > b and b > c implying that a > c).” And then we went ahead and showed that the data was vastly more consistent with Lognormal World than Normal World. I think it holds up.

An additional argument that since has been effective at explaining the paradigm to newcomers has been in terms of exploring the very meaning of Just-Noticeable Differences (JNDs) in the context of the intensity of aspects of one’s experience. Indeed, for (b), the depths of intensity of experience simply make no sense if we were to take a “Just-Noticeable Pinprick” as the unit of measurement and expect a multiple of it to work as the measuring rod between pain levels in the 1-10 pain scale. The upper ends of pain are just so bright, so immensely violent, so as to leave lesser pains as mere rounding errors. But if on each step of a JND of pain intensity we multiply the feeling by a constant, sooner or later (as Zvi might put it) “the rice grains on the chessboard suddenly get fully out of hand” and we enter hellish territory (for a helpful visual aid of this concept: start at 6:06 of our talk at the 2020 EAGxVirtual Unconference on this topic).

From my point of view, we can now justifiably work under the assumption that the qualitative picture painted by Log Scales is roughly correct. It is the more precise quantitative analysis which is a work in progress that ought to be iterated over in the coming years. This will entail broadening the range of people interviewed, developing better techniques to precisely capture and parametrize phenomenology (e.g. see our tool to measure visual tracers), use more appropriate and principled statistical methods (e.g. see the comment about the Bradley-Terry model and extreme value theory), experimental work in psychophysics labs, neuroimaging research of peak experiences, and the search for cost-effective pragmatic solutions to deal with the worst suffering. I believe that future research in this area will show conclusively the qualitative claims, and perhaps there will be strong consilience on the more precise quantitative claims (but in the absence of a true Qualiascope, the quantitative claims will continue to have a non-negligible margin of error).

Ok, you may say, but if I disagree about the importance of preventing pain, and I care more about e.g. human flourishing, why should I care about this? Here I would like to briefly address a key point that people in the EA sphere have raised in light of our work. The core complaint, if we choose to see it that way, is that one must be a valence utilitarian in order to care about this analysis. That only if you think of ethics in terms of classical Benthamite pain-minimization and pleasure-maximization should we be so keen on mapping the true distribution of valence across the globe. 

But is that really so?

Three key points stand out: First, that imperfect metrics that are proxies for aspects of what you care about (even when not all that you care about) can nonetheless be important. Second, that if you cared a little about suffering already, then the post-hoc discovery that suffering is actually that freaking skewed really ought to be a major update. And third, there really are reasons other than valence maximization as a terminal goal to care about extreme suffering: suffering is antithetical to flourishing since it has long-term sequelae. More so, even if confined to non-utilitarian ethical theories, one can make the case that there is something especially terrible about letting one’s fellow humans (and non-humans) suffer so intensely without doing anything about it. And perhaps especially so if stopping such horrors turn out to be rather easy. 

Let’s tackle each in turn.

(1) Perhaps here we should bring a simple analogy: GDP. Admittedly, there are very few conceptions of the good in which it makes sense for GDP to be the metric to maximize. But there are also few conceptions of the good where you should disregard it altogether. You can certainly be skeptical of the degree to which GDP captures all that is meaningful, but in nearly all views of economic flourishing, GDP will likely have a non-zero weight. Especially if we find that, e.g. some interventions we can do to the economy would cause a 99.9% reduction in a country’s GDP, one should probably not ignore that information (even if the value one assigns to GDP is relatively small compared to what other economists and social scientists assign it). Likewise for extreme suffering. There might be only a few conceptions of the good where that is the only thing we ought to work on. But avoiding hellish states is a rather universally desired state for oneself. Why not take it at least somewhat into account?

In truth, this is not something that classical questions in Effective Altruism pre-Log Scales could overcome either. For instance, as far as I am aware, in practice QALYs are used more as a guide than as an absolute; their value within EA comes from the fact that in practice interventions are orders of magnitude different when it comes to their cost-effectiveness when assessed with QALYs. So even though the vast majority of EAs are not QALY absolutists, the differences in QALYs saved between interventions are large enough that as an approximate guide, the metric still generates huge amounts of consilience.

(2) In turn, the post-hoc finding that hellish states are much, much worse than one would intuitively believe really should at least rebalance one’s priorities somewhat. Is there really no amount of suffering that would do so? Unless one has a utility function akin to a ReLu activation function, going far enough down into the depths of hell ought to count for something. And…

(3) Speaking candidly, in order to fully articulate the true significance of this finding will take us to philosophically polemical territory: philosophy of personal identity where person-affecting views will see the situation quite differently than person-moment-affecting views, philosophy of mind where the ontological status of pleasure and pain are questioned, and intricate questions that arise at the interface between the views of virtue ethicists, deontologists, negative and classical utilitarians. Of course a negative utilitarian who believes in Empty Individualism and Eternalism at the same time will likely be especially horrified by this information. But I would suggest that there are good reasons to care no matter how antithetical one’s view is to philosophical hedonism. 

In particular, I would argue that deontologists and virtue ethicists should still take note. The cultivation of virtue requires a minimum of wellbeing in order to maintain motivation to live. And perhaps deontologists might find extreme suffering particularly egregious from the point of view of “things so horrible that ought not to be''. Really, the people we interviewed for the cluster headache writeup told us that experiencing such levels of hellish suffering causes profound psychological transformations (e.g. one interviewee told us that experiencing the higher end of pain in a cluster headache feels like a profound “spiritual violation” from which you may never recover - a feeling most certainly aggravated by the seeming indifference of people at large about their plight). Virtue ethicists and deontologists might as well recognize this cause area as work that is unconscionable not to perform, regardless of arguments based on precise mathematical optimization for the prevention of negative valence. 

And finally, in all seriousness, as the cognitive intelligentsia of the world begins to see clearly the nature of the distribution of pleasure and pain, we can expect there to be a big social benefit to being the one who destroys hell. Right now there isn’t a huge social reward to be obtained by working on this cause, but I predict this will change. And, pragmatically, it is sensible to present this cause in a motivating rather than depressing light: indeed, let’s give honor, glory, and endless admiration to whoever makes tangible progress in tearing hell down. And to all of the millionaires and billionaires reading this: this could be you! You could be the one who took on the mantle of preventing all future cluster headaches, established the field of anti-tolerance drugs for severe chronic pain, or got rid of kidney stones (and you did it before it was cool!). Let’s get to work!