algekalipso

• algekalipso 22 Nov 2019 3:43 UTC

  1 point

  in reply to: MichaelPlant's comment on: Log­a­r­ith­mic Scales of Plea­sure and Pain (@Effec­tive Altru­ism NYC)

  Thank you! I just left a re­ply to your com­ment. Here’s a sum­mary of the core claim:

  In this ac­count, the fact that peo­ple would nat­u­rally and spon­ta­neously use a log­a­r­ith­mic scale to re­port their level of pain is a sim­ple im­pli­ca­tion of the fact that you can only defini­tively tell that “the pain got worse” when it got 10% worse and not when it be­came 1 unit worse (which soon be­comes hard to no­tice when you talk about ex­pe­riences with e.g. 1000 pain units per sec­ond).

• algekalipso 22 Nov 2019 3:42 UTC

  1 point

  in reply to: MichaelPlant's comment on: Log­a­r­ith­mic Scales of Plea­sure and Pain: Rat­ing, Rank­ing, and Com­par­ing Peak Ex­pe­riences Suggest the Ex­is­tence of Long Tails for Bliss and Suffering

  Hey Michael,

  Thank for com­ment­ing. First of all I agree that the claims of (A) and (B) do need to be dis­t­in­guished, and I ad­mit I didn’t make that con­cep­tual dis­tinc­tion very clear in the ar­ti­cle. I agree that the most im­por­tant take­away from the piece is (B), and I also think that this alone is already enough to challenge EA’s pri­ori­ti­za­tion meth­ods (i.e. ul­tra-painful ex­pe­riences are com­pletely fly­ing un­der the radar from the point of view of QALYs and similar met­rics; re­duc­ing the in­ci­dence of cluster headaches, mi­graines, kid­ney stones, etc. could be an ex­tremely cost-effec­tive EA ob­jec­tive).

  With that said, I would claim that (1) the ar­ti­cle does ac­tu­ally provide ev­i­dence for (A), (2) tak­ing this se­ri­ously clar­ifies a lot of puz­zling facts about ex­pe­rience and how it is re­ported, and (3) un­der­stand­ing that pain and plea­sure fol­lows a long-tail (most likely a log-nor­mal dis­tri­bu­tion) gives us a new prin­ci­pled way to ap­proach cause pri­ori­ti­za­tion.

  I un­der­stand the fact that the pain scales of stings and cluster headaches are *by con­struc­tion* log­a­r­ith­mic. But you have to un­der­stand that such a scale would only ever be “filled to the top” if ex­pe­riences ac­tu­ally differed in in­ten­sity also by the same amount. The ar­ti­cle (and pre­sen­ta­tion, which I strongly recom­mend you watch) ex­plain that all of the fol­low­ing are con­sis­tent with the pain scales (as re­ported!) are ac­tu­ally log­a­r­ith­mic:

  (a) the char­ac­ter­is­tic dis­tri­bu­tion of neu­ral ac­tivity is log-nor­mal, and un­der the mod­est as­sump­tion that in­ten­sity of ex­pe­rience is roughly pro­por­tional (or at least polyno­mi­ally pro­por­tional) to in­ten­sity of ex­pe­rience, that en­tails the dis­tri­bu­tion of in­ten­sity is also log-nor­mal.

  (b) the above can be fur­ther un­der­stood as a kind of “neu­ronal weather” (see the “avalanches” metaphor in the video pre­sen­ta­tion)

  (c) the pre­dic­tions of the log-nor­mal world are held by the data, and in par­tic­u­lar:

  (c1) there are few cat­e­gories of ex­pe­riences that cap­ture most of the ex­tremely good and ex­tremely bad sensations

  (c2) there is con­sis­tency in the defer­ence judge­ments of the qual­ity of ex­pe­rience (as seen in the defer­ence graph), and importantly

  (c3) The ra­tio of “1st worst or best ex­pe­rience vs. 2nd worst or best ex­pe­rience” fits a log-nor­mal dis­tri­bu­tion and it does not fit a nor­mal dis­tri­bu­tion.

  For the above rea­sons, bring­ing up the Fech­ner-We­ber is not, I would claim, a red-her­ring. Rather, I think it ties to­gether the whole ar­gu­ment. Here is why:

  I un­der­stand that Fech­ner-We­ber’s law maps phys­i­cal in­ten­sity to sub­jec­tive in­ten­sity, and that valence is not ex­ter­nally driven a lot of the time. But you may have missed the ar­gu­ment I’m mak­ing here. And that is that in one in­ter­pre­ta­tion of the law, a pre-con­scious pro­cess does a log trans­form on the in­ten­sity of the in­put and that by the time we are aware of it, what we be­come aware of are the lin­ear differ­ences in our ex­pe­rience. In the al­ter­nate in­ter­pre­ta­tion of the law, which I pro­pose, the senses (within the win­dow of adap­ta­tion) trans­late the in­ten­sity of the in­put into an equiv­a­lent in­ten­sity of ex­pe­rience. And the rea­son *why* we can only de­tect mul­ti­plica­tive differ­ences in the in­put *is be­cause* we can only no­tice con­sciously mul­ti­plica­tive differ­ences in the in­ten­sity of ex­pe­rience. Do you see what I am say­ing? In this ac­count, the fact that peo­ple would nat­u­rally and spon­ta­neously use a log­a­r­ith­mic scale to re­port their level of pain is a sim­ple im­pli­ca­tion of the fact that you can only defini­tively tell that “the pain got worse” when it got 10% worse and not when it be­came 1 unit worse (which soon be­comes hard to no­tice when you talk about ex­pe­riences with e.g. 1000 pain units per sec­ond).

  In other words, the scales are log­a­r­ith­mic be­cause we can only no­tice with con­fi­dence mul­ti­plica­tive in­cre­ments in the in­ten­sity of ex­pe­rience. And while this is fine and does not seem to have strong im­pli­ca­tions on the lower end of the scale, it very quickly es­ca­lates, to the point where by the time you are in ⁷⁄₁₀ pain you live in a world with or­ders of mag­ni­tude more pain units per sec­ond than you did when you were in ²⁄₁₀ pain.

  Fi­nally, you re­ally need the log­a­r­ith­mic scales to make room for the ul­tra-in­tense lev­els of plea­sure and pain that I high­lighted in the “ex­is­tence of ex­tremes” sec­tion. If peo­ple re­ported their pain on a lin­ear scale, they would quickly run into the prob­lem that they can­not de­scribe even some­thing as painful as a bro­ken bone, let along some­thing like a cluster headache.

Log­a­r­ith­mic Scales of Plea­sure and Pain (@Effec­tive Altru­ism NYC)

• algekalipso 2 Oct 2019 6:47 UTC

  8 points

  on: Is pain just a sig­nal to en­list al­tru­ists?

  Thanks for writ­ing this.

  How would this model ex­plain Cluster Headaches? They are not par­tic­u­larly more in­ca­pac­i­tat­ing than mi­graines, yet they are (pos­si­bly liter­ally*) thou­sands of times more acutely painful than them. What is the role of this X1000 mul­ti­plier on phe­nom­e­nal pain in such cases? As far as I can tell, in the an­ces­tral en­vi­ron­ment no­body could have done any­thing to help you if you were hav­ing a Cluster Headache, and your chances of re­pro­duc­tion seem to be the same whether that pain was a thou­sand times less bad (which would still be VERY bad, but not in the level of ul­tra-Hel­lish pain). In par­tic­u­lar, other species are known the have Cluster Headaches too, such as cats. So per­haps we should cluster pains into two buck­ets—those that have so­cial sig­nifi­cance and those that don’t. I worry that this study will make peo­ple dis­miss ex­treme suffer­ing in non­hu­man an­i­mals, but that should only re­ally ap­ply to so­cially-use­ful pains. I sus­pect that there are many species-spe­cific ul­tra-painful ex­pe­riences that we will not dis­cover (and pri­ori­tize!) un­less we look for them.

  
  

  *See: https://​​fo­rum.effec­tivealtru­ism.org/​​posts/​​gtGe8WkeFvqucYLAF/​​log­a­r­ith­mic-scales-of-plea­sure-and-pain-rat­ing-rank­ing-and

• algekalipso 12 Sep 2019 19:11 UTC

  3 points

  in reply to: Julia_Wise's comment on: Cluster Headache Fre­quency Fol­lows a Long-Tail Distribution

  That’s a good point, thank you. We should dis­t­in­guish be­tween life­time use and cur­rent use in fu­ture sur­veys. Per­haps even ask­ing whether “they worked the first time you used them” to see if peo­ple who cur­rently use them had a bet­ter re­ac­tion to their first try rel­a­tive to those who did try them at some point but do not cur­rently use them.

  I would add that other rea­sons why peo­ple might have used them in the past but don’t cur­rently in­clude “can’t ac­cess it now”, “too afraid of le­gal reper­cus­sions”, and “so­cial stigma”. While dis­con­tin­u­ing them due to side-effects and lack of effec­tive­ness can make them look more effec­tive than they are among the “use them” group, the other rea­sons for dis­con­tinu­a­tion do not have this effect. I don’t know what % of past users dis­con­tinued for which rea­son, and that seems like a good thing to find out.

• algekalipso 12 Sep 2019 19:06 UTC

  1 point

  in reply to: Khorton's comment on: Cluster Headache Fre­quency Fol­lows a Long-Tail Distribution

  Depends on con­text. In most cases the ‘we’ refers to my team and I at the Qualia Re­search In­sti­tute. For ex­am­ple: “Since a num­ber of in­ter­views we’ve con­ducted have shown that even sub-hal­lu­cino­genic doses of DMT can abort cluster headaches” refers to QRI (with other mem­bers of the re­search group hav­ing con­ducted such in­ter­views).

  
  

  I should note that the word is also used in the ‘di­dac­tic we’ sense a num­ber of times (as in “we will ex­plore the era of the dinosaurs to­gether” in a Na­tional Geo­graphic doc­u­men­tary).

• algekalipso 4 Sep 2019 17:54 UTC

  5 points

  in reply to: Tetraspace Grouping's comment on: Log­a­r­ith­mic Scales of Plea­sure and Pain: Rat­ing, Rank­ing, and Com­par­ing Peak Ex­pe­riences Suggest the Ex­is­tence of Long Tails for Bliss and Suffering

  Ac­cord­ing to “Right Con­cen­tra­tion: A Prac­ti­cal Guide to the Jhanas” by L. Bras­ing­ton and “The Mind Illu­mi­nated” by Cu­ladasa, it is fea­si­ble to achieve Jhana states within two years of ded­i­cated prac­tice. This en­tails a few hours of med­i­ta­tion a day and at­tend­ing at least one 9-day re­treat over the course of this time pe­riod. The books ex­plain in de­tail how to get there in a very prac­ti­cal and no-non­sense way.

  I per­son­ally have yet to in­vest that time into this task, but I know that one of the other core mem­bers of the Qualia Re­search In­sti­tute, Romeo Stevens, is now able to achieve Jhanas thanks to his med­i­ta­tion prac­tice. I do in­tend to do this in the near fu­ture.

  Also, we are look­ing into do­ing EEG and fMRI stud­ies on peo­ple who can en­ter those states as a means to test the CDNS ap­proach to valence quan­tifi­ca­tion, which is a core part of our re­search plan.

• algekalipso 4 Sep 2019 17:26 UTC

  5 points

  on: Log­a­r­ith­mic Scales of Plea­sure and Pain: Rat­ing, Rank­ing, and Com­par­ing Peak Ex­pe­riences Suggest the Ex­is­tence of Long Tails for Bliss and Suffering

  2019-09-04 Up­date: Since post­ing this I’ve learned about the Bradley-Terry model for ob­tain­ing la­tent traits based on sets of rank­ings (https://​​en.wikipe­dia.org/​​wiki/​​Bradley%E2%80%93Terry_model) and also that there are libraries to do this (e.g. https://​​pypi.org/​​pro­ject/​​choix/​​).
  
  Ad­di­tion­ally, I’ve learned about “ex­treme value the­ory”, which de­scribes the statis­ti­cal dis­tri­bu­tion of ex­treme val­ues (e.g. https://​​en.wikipe­dia.org/​​wiki/​​Gen­er­al­ized_ex­treme_value_dis­tri­bu­tion) and seen some ap­pli­ca­tions to other long-tail events (see: https://​​blog.givewell.org/​​2015/​​07/​​13/​​ge­o­mag­netic-storms-us­ing-ex­treme-value-the­ory-to-gauge-the-risk/​​).
  
  I will use those two new key statis­ti­cal ap­proaches to an­a­lyze this pi­lot dataset and also fu­ture iter­a­tions of this study (fo­cused more on peo­ple who’ve ex­pe­rienced ex­tremes of valence like cluster headaches or 5-MeO-DMT states). I am cur­rently busy work­ing on a num­ber of other pro­jects crit­i­cal for the Qualia Re­search In­sti­tute, so do­ing this is cur­rently on the back-burner (though of course I’m happy to hear if any­one is in­ter­ested in tak­ing on this challenge as a vol­un­teer pro­ject).
  
  Cheers!

• algekalipso 2 Sep 2019 1:29 UTC

  14 points

  in reply to: Milan_Griffes's comment on: Cause X Guide

  To zoom in on the “log­a­r­ith­mic scales of plea­sure and pain” an­gle (I’m the au­thor), I would say that this way of see­ing the world sug­gests that the bulk of suffer­ing is con­cen­trated on a small per­centage of ex­pe­riences. Thus, find­ing scale­able treat­ments spe­cially for ul­tra-painful con­di­tions could take care of a much larger per­cent of the world bur­den of suffer­ing than most peo­ple would in­tu­itively re­al­ize. I re­ally think this should be up in the list of con­sid­er­a­tions for Cause X. Speci­fi­cally:

  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.

  
  

  (see also the writeup of an event we hosted about pos­si­ble new EA Cause Xs)

• algekalipso 16 Aug 2019 20:06 UTC

  8 points

  on: Ask Me Any­thing!

  Do you think that the em­piri­cal find­ing that pain and suffer­ing are dis­tributed along a log­nor­mal dis­tri­bu­tion (cf. Log­a­r­ith­mic Scales of Plea­sure and Pain) has im­pli­ca­tions for how to pri­ori­tize causes? In par­tic­u­lar, what do you say about these ten­ta­tive im­pli­ca­tions:

  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).

  Thank you!!!

• algekalipso 16 Aug 2019 19:51 UTC

  2 points

  in reply to: SoerenMind's comment on: Log­a­r­ith­mic Scales of Plea­sure and Pain: Rat­ing, Rank­ing, and Com­par­ing Peak Ex­pe­riences Suggest the Ex­is­tence of Long Tails for Bliss and Suffering

  Hey So­eren!

  I would dis­agree for the fol­low­ing rea­son. For a group to con­tribute equally it needs to have both its av­er­age and its size be such that when you mul­ti­ply them you get the same value. While it is true that peo­ple at the 50% per­centile get ¹⁄₁₀ of the peo­ple at the 90% (and ~1/​50 of the 99%), these do not define groups. What we need to look at in­stead is the cu­mu­la­tive dis­tri­bu­tion func­tion:

  The bot­tom 50% ac­counts for 3.17% of incidents

  The bot­tom 90% ac­counts for 30% of incidents

  The bot­tom 95% ac­counts for 43% of incidents

  What I am get­ting at is that for a given per­centile, the con­tri­bu­tion from the group “this per­centile and lower” will be a lot smaller than the value at that per­centile mul­ti­plied by the frac­tion of the par­ti­ci­pants be­low that level. This is be­cause the dis­tri­bu­tion is very skewed, so for any given per­centile the val­ues be­low it quickly de­crease.

  Another way of look­ing at this is by as­sum­ing that each per­centile has a cor­re­spond­ing value (in the ex­am­ple “num­ber of CHs per year”) pro­por­tional to the rar­ity of that per­centile or above. For sim­plic­ity, let’s say we have a step func­tion where each time we di­vide the group by half we get twice the value for those right above the cut-off:

  0 to 50% have 1/​year

  50 to 75% have 2/​year

  75 to 87.5% have 4/​year

  and so on...

  Here each group con­tributes equally (size * # of CH is the same for each group). Counter-in­tu­itively, this does not im­ply that ex­tremes ac­count for a small amount. On the con­trary, it im­plies that the av­er­age is in­finite (cf. St. Peters­burg para­dox): even though you will have that for any given per­centile, the av­er­age be­low it is always finite (e.g. be­tween 0 and 40% it’s 1/​year), the av­er­age (and to­tal con­tri­bu­tion) above that per­centile is always in­finite. In this ideal­ized case, it will always be the case that “the bulk is con­cen­trated on a tiny per­centile” (and in­deed you can make that per­centile as small as you want and still get in­finitely more above it than be­low it).

  The em­piri­cal dis­tri­bu­tion is not so skewed that we need to worry about in­finity. But we do need to worry about the 57% ac­counted for by the top 5%.

• algekalipso 15 Aug 2019 23:29 UTC

  1 point

  in reply to: SoerenMind's comment on: Log­a­r­ith­mic Scales of Plea­sure and Pain: Rat­ing, Rank­ing, and Com­par­ing Peak Ex­pe­riences Suggest the Ex­is­tence of Long Tails for Bliss and Suffering

  Hi! Thank you for elab­o­rat­ing on what your ques­tion is :)

  “Bulk” is in­deed a very am­bigu­ous term. Would you say 80% is “the bulk”? And 20% is “a small per­centage”? If so we would be in agree­ment. If not, it is more of a word­ing is­sue than a mat­ter of sub­stance, I think.

  Good catch that the num­bers I pro­vided would sug­gest a power law that just keeps go­ing (e.g. similar to St. Peters­burg para­dox?). If we use the Cluster Headache dataset, the num­bers are:

  50% per­centile ex­pe­riences 70 CH/​year

  80% per­centile ex­pe­riences 365 CH/​year

  90% per­centile ex­pe­riences 730 CH/​year

  98% per­centile ex­pe­riences 2190 CH/​year

  So at least in this case the 90th per­centile does get 10X the amount of the 50th per­centile. But the 98th and 99th per­centile is not as high as 100X, and more like 20 to 50x. So not quite the num­bers I used as an ex­am­ple, but also not too far off.

  I should also clar­ify that by fre­quency I mean the product of ‘how many peo­ple have it’, ‘how of­ten’ and ‘for how long’.

  Here is the main idea: In Log­nor­mal World, you would see a log­nor­mal dis­tri­bu­tion for “amount of suffer­ing per per­son”, “peak suffer­ing per per­son”, “how long suffer­ing above a cer­tain thresh­old lasts for each per­son”, etc.

  To illus­trate this point, let’s say that each per­son’s he­do­nic tone per each sec­ond of their life is dis­tributed along a log­nor­mal with an ex­po­nent that is a Gaus­sian with mean x and sd of y. We would then also have that x, across differ­ent peo­ple, is dis­tributed along a Gaus­sian with a mean of z and sd of t. Now, if you want to get the global dis­tri­bu­tion of suffer­ing per sec­ond across peo­ple, you would need to con­volve two Gaus­si­ans on the log­a­r­ith­mic pain scale (which rep­re­sent the ex­po­nents of the log­nor­mal dis­tri­bu­tions). Since con­volv­ing two Gaus­si­ans gives you an­other Gaus­sian, we would then have that the global dis­tri­bu­tion of suffer­ing per sec­ond is also a log­nor­mal dis­tri­bu­tion! So both at the in­di­vi­d­ual, and the global scale the log­nor­mal long-tails will be pre­sent. Now, for you to ap­pre­ci­ate the “bulk” of the suffer­ing, you would need to look at the in­di­vi­d­u­als who have the largest means for the nor­mal dis­tri­bu­tion in the ex­po­nent (x in this case). Hence why look­ing at one’s own in­di­vi­d­ual % of time in ex­treme pain does not provide a good idea of how much of it there is in the wild across peo­ple (es­pe­cially if one is close to the me­dian; i.e. a pretty happy per­son).

• algekalipso 15 Aug 2019 18:46 UTC

  3 points

  in reply to: Lukas_Gloor's comment on: Log­a­r­ith­mic Scales of Plea­sure and Pain: Rat­ing, Rank­ing, and Com­par­ing Peak Ex­pe­riences Suggest the Ex­is­tence of Long Tails for Bliss and Suffering

  Ad­ding to what Lu­cas men­tioned (how we are mo­ti­vated in part by long­ing/​ad­dic­tion for strong re­wards): Suffer­ing and nega­tive re­in­force­ment are cor­re­lated but are by no means the same thing. In the case of ex­treme suffer­ing, there seems to be a point where the pain has already maxed out in terms of nega­tive re­in­force­ment ca­pac­ity, and any­thing above it is just sense­less suffer­ing. Cluster headaches would not cause any less be­hav­ioral sup­pres­sion if they were 10 or even 100 times less painful. Like­wise, our abil­ity to rea­son about pain and plea­sure is limited by our state-de­pen­dent abil­ity to imag­ine it. As I ar­gued in the ar­ti­cle, our abil­ity to imag­ine any pain or plea­sure that goes be­yond our abil­ity to ex­trap­o­late with the qualia ac­cessible to us at the mo­ment is very limited.

  The bliss of 5-MeO-DMT or epilep­tic tem­po­ral lobe seizures is as Dos­to­evsky said “a hap­piness un­think­able in the nor­mal state and uni­mag­in­able for any­one who hasn’t ex­pe­rienced it”. Like­wise for ex­treme pain. So you wouldn’t be able to in­fer that these states ex­ist (and are much more preva­lent than one in­tu­itively be­lieves) merely from ob­serv­ing the pat­terns of re­in­force­ment from a third-per­son point of view.

• algekalipso 15 Aug 2019 18:41 UTC

  4 points

  in reply to: SoerenMind's comment on: Log­a­r­ith­mic Scales of Plea­sure and Pain: Rat­ing, Rank­ing, and Com­par­ing Peak Ex­pe­riences Suggest the Ex­is­tence of Long Tails for Bliss and Suffering

  The short an­swer is—ex­treme pain is vastly more com­mon than is gen­er­ally be­lieved. Statis­tics such as 20% of peo­ple in the USA ex­pe­rience chronic pain, with 8% ex­pe­rienc­ing high-im­pact chronic pain (in­terferes with most as­pects of life). If in­deed we live in Log­nor­mal World, we can ex­pect that the me­dian per­son will prob­a­bly have rel­a­tively low ac­quain­tance with ex­treme suffer­ing (un­til old age), but that the peo­ple in the top 10% of suffer­ers will have 10X the amount, and peo­ple in the 99% will have 100X the amount. If we take a per­son-neu­tral point of view (i.e. Empty or Open In­di­vi­d­u­al­ism) and care about “mo­ments of ex­pe­rience” it does not re­ally mat­ter who gets to ex­pe­rience it, at least not morally. There are no diminish­ing re­turns per per­son when it comes to the nega­tive value of suffer­ing (once adap­ta­tion has been taken into ac­count).

  
  

  As with other long-tails, it may seem hard to be­lieve that ”...the bulk of suffer­ing is con­cen­trated in a small per­centage of ex­pe­riences...”. But so it is hard to imag­ine that there are billion­aires out there if all one knows about is the in­come of one’s fam­ily and small cir­cle of friends. Million­aires are rare, but not that rare (about 3% of the pop­u­la­tion), and we have that in the case of in­come the bulk of cap­i­tal is con­cen­trated in a small per­cent of peo­ple (e.g. ~20% of the pop­u­la­tion con­trol­ling ~80% of the wealth, and the top 1% con­trol­ling ~45% of it).

  
  

  Like­wise, the re­search pre­sented here would sug­gest that in the case of suffer­ing there are “suffer­ing billion­aires” out there, and that they ac­count for a much larger % of to­tal suffer­ing than we in­tu­itively would imag­ine.

• algekalipso 15 Aug 2019 7:59 UTC

  2 points

  in reply to: SoerenMind's comment on: Log­a­r­ith­mic Scales of Plea­sure and Pain: Rat­ing, Rank­ing, and Com­par­ing Peak Ex­pe­riences Suggest the Ex­is­tence of Long Tails for Bliss and Suffering

  Thank you!

  The ar­ti­cle does fo­cus on the long-tail of in­ten­sity and qual­ity of both plea­sure and pain rather than fre­quency. That said, it dis­cusses the Log­nor­mal World as a gen­eral prin­ci­ple, which would also pre­dict that the fre­quency of pain or plea­sure would fol­low a long-tail in ad­di­tion to their in­ten­sity and qual­ity.

  This is backed up by the pre­vi­ous ar­ti­cle “Cluster Headache Fre­quency Fol­lows a Long-Tail Distri­bu­tion”, where we an­a­lyzed a sur­vey about Cluster Headache fre­quency among suffer­ers, and showed it fol­lowed a long-tail (with statis­tics like “The bot­tom 80% ac­counts for 17% of in­ci­dents and the bot­tom 90% ac­counts for 30% of in­ci­dents”, and val­ues rang­ing from 1 Cluster Headache a year all the way to more than 1,000). We should col­lect data on e.g. kid­ney stone, mi­graine, etc. fre­quency per in­di­vi­d­ual to see if they also fol­low a long-tail. Given the gen­eral pat­tern, we sus­pect they prob­a­bly do.

Log­a­r­ith­mic Scales of Plea­sure and Pain: Rat­ing, Rank­ing, and Com­par­ing Peak Ex­pe­riences Suggest the Ex­is­tence of Long Tails for Bliss and Suffering

• algekalipso 9 Aug 2019 19:36 UTC

  3 points

  in reply to: Greg_Colbourn's comment on: Cluster Headache Fre­quency Fol­lows a Long-Tail Distribution

  It’s great that it works for some peo­ple, some of the time. In ab­solute terms, it is a mas­sive good, so it should be pro­moted more. Prag­mat­i­cally it might make sense to em­pha­size it right now given the low prob­a­bil­ity that DMT will be ap­proved as a treat­ment in the next few years, so un­til then Em­gal­ity should be dis­cussed more. That said, yes, in terms of % re­lief it still is in a com­pletely differ­ent class than DMT. That is, it tends to re­duce in­ci­dence rather than get rid of them, and it is only ap­proved for epi­sodic (rather than chronic) CHs, which ac­count for a rel­a­tively small % of the num­ber of CHs ex­pe­rienced, as de­scribed in this ar­ti­cle (due to the long-tail).

• algekalipso 7 Aug 2019 20:07 UTC

  3 points

  in reply to: Ozzie Gooen's comment on: Cluster Headache Fre­quency Fol­lows a Long-Tail Distribution

  In the ar­ti­cle speci­fi­cally about N,N-DMT as a pos­si­ble treat­ment for CHs, Quintin added a rough QALY calcu­la­tion (I should add that any QALY es­ti­mate con­cern­ing CHs and other ul­tra-painful di­s­or­ders will typ­i­cally severely un­der­es­ti­mate the value of the in­ter­ven­tions, given the log­a­r­ith­mic na­ture of pain scales):

  
  

  While we be­lieve that tra­di­tional met­rics such as the QALY do not ac­cu­rately cap­ture the suffer­ing caused by a cluster headache (see up­com­ing post on the true pain/​plea­sure scale), a rough QALY calcu­la­tion would be as fol­lows (fo­cus­ing on chronic cluster headache suffer­ers rather than av­er­age, since they com­pro­mise up to 83% of to­tal headaches[3]):

  Face­book AD cam­paign:

  1. An es­ti­mated 370,000 Amer­i­cans suffer from cluster headaches, 68% of whom are on Face­book[4] (=251,000). About 15% of these suffer from chronic cluster headaches (=37,740). Ac­cord­ing to Sprout So­cial, the av­er­age es­ti­mated cost per click of an ad cam­paign is $1.72. As­sum­ing ¹⁄₁₀ who click are cluster headache suffer­ers, to reach all chronic suffer­ers would take (bal­l­park) $650,000.

  2. As­sum­ing about 30% of those who view the ad will pur­sue the treat­ment (rough es­ti­mate-those who put 2 or less on sur­vey re­sults for ques­tions of le­gal­ity, difficulty to ac­quire etc.) and that in 68% of cases it cured or nearly cured their clusters (based on sur­vey re­sults), then the re­sult­ing in­crease in QALYs would be (37,740 peo­ple * 0.3 * 0.68) * [0.760 (QALY co­effi­cient) * 1 QALY – ( −0.429 (QALY co­effi­cient)* (0.47QALY)) [5]] = $650,000/​7, 404QALY = $87.70 per QALY.

  3. Th­ese ads could also be tar­geted to users in coun­tries where psilo­cy­bin and DMTare le­gal for use recre­ation­ally, in­creas­ing con­ver­sion rate. Fur­ther tar­get­ing could be done on Face­book groups (and other so­cial me­dia groups) which are as­so­ci­ated with cluster headache treat­ment.

• algekalipso 4 Aug 2019 8:50 UTC

  7 points

  in reply to: Larks's comment on: Cluster Headache Fre­quency Fol­lows a Long-Tail Distribution

  I com­pletely agree that the mem­bers of a cluster headache sub­red­dit or face­book group are not nec­es­sar­ily rep­re­sen­ta­tive, and in fact quite likely not rep­re­sen­ta­tive at all.

  I think that the con­clu­sion that the dis­tri­bu­tion fol­lows a long-tail re­gard­less is still ac­cu­rate. I rea­son this based on the fol­low­ing point: even if the prob­a­bil­ity of par­ti­ci­pat­ing in the sur­vey in­creased ex­po­nen­tially as a func­tion of the num­ber of times one ex­pe­riences CHs per year (or sig­moid at the limit), you would nonethe­less not be able to make a Gaus­sian dis­tri­bu­tion look like a log-nor­mal. The rea­son is that the rate at which a Gaus­sian de­creases is pro­por­tional to the in­verse *squared* of the dis­tance from the mean. So we would still get a net de­crease at an ex­po­nen­tial rate, which does not pro­duce a long-tail (just a some­what more bulky tail that still ta­pers off rather quickly). For it to ex­hibit a long-tail, the prob­a­bil­ity of par­ti­ci­pat­ing in the sur­vey as a func­tion of the num­ber of CHs per year would have to grow *dou­bly* ex­po­nen­tially, at which point we re­ally run out very quickly of pos­si­ble par­ti­ci­pants.

  That said, I do agree that there is likely an over-es­ti­ma­tion of the fre­quency, but I would ar­gue due to the above rea­sons that such over-es­ti­ma­tion can’t ac­count for the long-tail.

• algekalipso 3 Aug 2019 23:21 UTC

  11 points

  in reply to: HaukeHillebrandt's comment on: Cluster Headache Fre­quency Fol­lows a Long-Tail Distribution

  Thank you. The sur­vey said that 68% of suffer­ers who have used psychedelics found gave them a rat­ing of 4 or 5, where 5 means “They have com­pletely elimi­nated the cluster headaches”. I would cer­tainly stand by the claim that “there are mil­lions of peo­ple suffer­ing need­lessly from this con­di­tion who could be nearly-in­stantly cured with some­thing as sim­ple as grow­ing and eat­ing some magic mush­rooms.” We’ve in­ter­viewed peo­ple for whom sub-hal­lu­cino­genic doses of DMT and psilo­cy­bin took a ¹⁰⁄₁₀ pain CH all the way to a ¹⁄₁₀ or ⁰⁄₁₀. And in the case of DMT, due to its method of ad­minis­tra­tion, this takes place within sec­onds (more than one but less than 10). Even if this works only for, say, 20% of the suffer­ers, it is still mil­lions in ab­solute terms.

  Thanks for say­ing: “Gen­er­ally I re­ally find this re­search agenda in­ter­est­ing”. My ex­pe­rience has been that few peo­ple take se­ri­ously the long-tails of plea­sure and pain. This is pre­cisely the sort of miss­ing piece of in­for­ma­tion that can add an en­tire new wing to EA.