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


Based on: the char­ac­ter­is­tic dis­tri­bu­tion of neu­ral ac­tivity, per­sonal ac­counts of in­tense plea­sure and pain, the way var­i­ous pain scales have been de­scribed by their cre­ators, and the re­sults of a pi­lot study we con­ducted which ranks, rates, and com­pares the he­do­nic qual­ity of ex­treme ex­pe­riences, we sug­gest that the best way to in­ter­pret plea­sure and pain scales is by think­ing of them as log­a­r­ith­mic com­pres­sions of what is truly a long-tail. The most in­tense pains are or­ders of mag­ni­tude more awful than mild pains (and sym­met­ri­cally for plea­sure).

This should in­form the way we pri­ori­tize al­tru­is­tic in­ter­ven­tions and plan for a bet­ter fu­ture. Since the bulk of suffer­ing is con­cen­trated in a small per­centage of ex­pe­riences, fo­cus­ing our efforts on pre­vent­ing cases of in­tense suffer­ing likely dom­i­nates most util­i­tar­ian calcu­la­tions.

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).


We­ber’s Law

We­ber’s Law de­scribes the re­la­tion­ship be­tween the phys­i­cal in­ten­sity of a stim­u­lus and the re­ported sub­jec­tive in­ten­sity of per­ceiv­ing it. For ex­am­ple, it de­scribes the re­la­tion­ship be­tween how loud a sound is and how loud it is per­ceived as. In the gen­eral case, We­ber’s Law in­di­cates that one needs to vary the stim­u­lus in­ten­sity by a mul­ti­plica­tive frac­tion (called “We­ber’s frac­tion”) in or­der to de­tect a just no­tice­able differ­ence. For ex­am­ple, if you can­not de­tect the differ­ences be­tween ob­jects weigh­ing 100 grams to 105 grams, then you will also not be able to de­tect the differ­ences be­tween ob­jects weigh­ing 200 grams to 210 grams (im­ply­ing the We­ber frac­tion for weight per­cep­tion is at least 5%). In the gen­eral case, the senses de­tect differ­ences log­a­r­ith­mi­cally.

There are two com­pel­ling sto­ries for in­ter­pret­ing this law:

In the first story, it is the low-level pro­cess­ing of the senses which do the log­a­r­ith­mic map­ping. The senses “com­press” the in­ten­sity of the stim­u­la­tion and send a “lin­earized” packet of in­for­ma­tion to one’s brain, which is then ren­dered lin­early in one’s ex­pe­rience.

In the sec­ond story, the senses, within the win­dow of adap­ta­tion, do a fine job of trans­lat­ing (some­what) faith­fully the ac­tual in­ten­sity of the stim­u­lus, which then gets ren­dered in our ex­pe­rience. Our in­abil­ity to de­tect small ab­solute differ­ences be­tween in­tense stim­uli is not be­cause we are not ren­der­ing such differ­ences, but be­cause We­ber’s law ap­plies to the very in­ten­sity of ex­pe­rience. In other words, the prop­er­ties of one’s ex­pe­rience could fol­low a long-tail dis­tri­bu­tion, but our abil­ity to ac­cu­rately point out differ­ences be­tween the prop­er­ties of ex­pe­riences is pro­por­tional to their in­ten­sity.

We claim that, at least for the case of valence (i.e the plea­sure-pain axis), the sec­ond story is much closer to the truth than the first. Ac­cord­ingly, this ar­ti­cle re­thinks the plea­sure-pain axis (also called the valence scale) by pro­vid­ing ev­i­dence, ar­gu­ments, and dat­a­points to sup­port the idea that how good or bad ex­pe­riences feel fol­lows a long-tail dis­tri­bu­tion.

As an in­tu­ition pump for what is to fol­low, we would like to high­light the em­piri­cal find­ing that brain ac­tivity fol­lows a long-tail dis­tri­bu­tion (see: Statis­ti­cal Analy­ses Sup­port Power Law Distri­bu­tions Found in Neu­ronal Avalanches, and Log­a­r­ith­mic Distri­bu­tions Prove that In­trin­sic Learn­ing is Heb­bian). The story where the “true valence scale” is a log­a­r­ith­mic com­pres­sion is en­tirely con­sis­tent with the em­piri­cal long-tails of neu­ral ac­tivity (in which “neu­ral avalanches” ac­count for a large frac­tion of over­all brain ac­tivity).

The con­crete line of ar­gu­ment we will pre­sent is based on the fol­low­ing:

  1. Phenomenolog­i­cal ac­counts of in­tense plea­sure and pain (w/​ ac­counts of phe­nom­e­nal time and space ex­pan­sion),

  2. The way in which pain scales are de­scribed by those who de­vel­oped them, and

  3. The an­a­lytic re­sults of a pi­lot study we con­ducted which in­ves­ti­gates how peo­ple rank, rate, and as­sign rel­a­tive pro­por­tions to their top 3 best and worst experiences

Why This Matters

Even if you are not a strict valence util­i­tar­ian, hav­ing the in­sight that the valence scale is long-tailed is still very im­por­tant. Most eth­i­cal sys­tems do give some weight to the pre­ven­tion of suffer­ing (in ad­di­tion to the cre­ation of sub­jec­tively valuable ex­pe­riences), even if that is not all they care about. If your eth­i­cal sys­tem weighted slightly the task of pre­vent­ing suffer­ing when be­liev­ing in a lin­ear valence scale, then learn­ing about the long-tailed na­ture of valence should in prin­ci­ple cause a ma­jor up­date. If in­deed the worst ex­pe­riences are ex­po­nen­tially more nega­tive than origi­nally be­lieved by one’s eth­i­cal sys­tem, which nonethe­less still cared about them, then af­ter learn­ing about the true valence scale the sys­tem would have to repri­ori­tize. We sug­gest that while it might be un­re­al­is­tic to have ev­ery eth­i­cal sys­tem re­fo­cus all of its en­er­gies on the pre­ven­tion of in­tense suffer­ing (and sub­se­quently on re­search­ing how to cre­ate in­tense bliss sus­tain­ably), we can nonethe­less ex­pect such sys­tems to raise this goal on their list of pri­ori­ties. In other words, while “end­ing all suffer­ing” will likely never be a part of most peo­ple’s eth­i­cal sys­tem, we hope that the data and ar­gu­ments here pre­sented at least per­suade them to add “…and pre­vent in­tense forms of suffer­ing” to the set of desider­ata.

In­deed, lack of aware­ness about the long-tails of bliss and suffer­ing may be the cause of an on­go­ing mas­sive moral catas­tro­phe (notes by Linch). If in­deed the de­gree of suffer­ing pre­sent in ex­pe­riences fol­lows a long-tail dis­tri­bu­tion, we would ex­pect the worst ex­pe­riences to dom­i­nate most util­i­tar­ian calcu­lus. The biggest bang for the buck in al­tru­is­tic in­ter­ven­tions would there­fore be those that are ca­pa­ble of di­rectly ad­dress­ing in­tense suffer­ing and gen­er­at­ing su­per-bliss.

Gen­eral Ideas

The Non-Lin­ear­ity of Plea­sure and Pain

Cap­tion: True long-tail plea­sure scale (warn­ing: psychedelics in­crease valence var­i­ance – the val­ues here are for “good/​lucky” trips and there is no guaran­tee e.g. LSD will feel good on a given oc­ca­sion). Also: Ma­nia is not always pleas­ant, but when it is, it can be su­per bliss­ful.
Cap­tion: True long-tail pain scale

As we’ve briefly dis­cussed in pre­vi­ous ar­ti­cles (1, 2, 3), there are many rea­sons to be­lieve that both plea­sure and pain can be felt along a spec­trum with val­ues that range over pos­si­bly or­ders of mag­ni­tude. Un­der­stand­ably, some­one who is cur­rently in a state of con­scious­ness around the hu­man me­dian of valence is likely to be skep­ti­cal of a claim like “the bliss you can achieve in med­i­ta­tion is liter­ally 100 times bet­ter than eat­ing your fa­vorite food or hav­ing sex.” In­tu­itively, we only have so much space in our ex­pe­rience to fit bliss, and when one is in a “nor­mal” or typ­i­cal state of mind for a hu­man, one is forced to imag­ine “ul­tra bliss­ful states” by ex­trap­o­lat­ing the el­e­ments of one’s cur­rent ex­pe­rience, which cer­tainly do not seem ca­pa­ble of be­ing much bet­ter than, say, 50% of the cur­rent level of plea­sure (or pain). The prob­lem here is that the very build­ing blocks of ex­pe­riences that en­able them to be ul­tra-high or ul­tra-low valence are them­selves nec­es­sary to imag­ine ac­cu­rately how they can be put to­gether. Talk­ing about ex­treme bliss to some­one who is an­he­do­nic is akin to talk­ing about the rich range of pos­si­ble color ex­pe­riences to some­one who is con­gen­i­tally fully col­or­blind (cf. “What Mary Didn’t Know“).

“Ok”, you may say, “you are just tel­ling me that plea­sure and pain can be or­ders of mag­ni­tude stronger than I can even con­ceive of. What do you base this on?”. The most straight­for­ward way to be con­vinced of this is to liter­ally ex­pe­rience such states. Alas, this would be deeply un­eth­i­cal when it comes to the nega­tive side, and it re­quires spe­cial ma­te­ri­als and pa­tience for the pos­i­tive side. In­stead, I will provide ev­i­dence from a va­ri­ety of meth­ods and con­di­tions.

Per­sonal Accounts

I’ve been lucky to not have ex­pe­rienced ma­jor pain in my life so far (the worst be­ing, per­haps, de­pres­sion dur­ing my teens). I have, how­ever, had two key ex­pe­riences that gave me some time to in­tro­spect on the non-lin­ear na­ture of pain. The first one comes from when I ac­ci­den­tally cut a su­per-spicy pep­per and touched it with my bare hands (the batch of pep­pers I was cut­ting were mild, but a su­per-hot one snuck into the pro­duce box). After a few min­utes of cut­ting the pep­pers, I no­ticed that a burn­ing heat be­gan to in­ten­sify in my hands. This was the start of ex­pe­rienc­ing “hot pep­per hands” for a full 8 hours (see other peo­ple’s ex­pe­riences: 1, 2, 3). The first two to three hours of this or­deal were the worst, where I ex­pe­rienced what I rated as a per­sis­tent 410 pain in­ter­spersed with brief mo­ments of 510 pain. The cu­ri­ous thing was that the 510 pain mo­ments were clearly dis­cernible as qual­i­ta­tively differ­ent. It was as if the very nu­mer­ous pin­pricks and burn­ing sen­sa­tions all over my hands were in a some­what di­s­or­ga­nized state most of the time, but when­ever they man­aged to build-up for long enough, they would start click­ing with each other (pre­sum­ably via phase-lock­ing), giv­ing rise to res­o­nant waves of pain that felt both more en­er­getic, and more aver­sive on the whole. In a way, this jump from what I rated as 410 to 510 was qual­i­ta­tive as well as quan­ti­ta­tive, and it gave me some idea of how some­thing that is already bad can be­come even worse.

My sec­ond ex­pe­rience in­volves a mild joint in­jury I ex­pe­rienced while play­ing Bub­ble Soc­cer (a very fun sport no doubt, and a com­mon cor­po­rate treat for Sili­con Valley cog­no­tari­ats, but ac­cord­ing to my doc­tor it is also a fre­quent source of in­juries among pro­gram­mers). Be­fore do­ing phys­i­cal ther­apy to treat this prob­lem (which mostly took care of it), I re­mem­ber spend­ing hours in­tro­spect­ing on the qual­ity of the pain in or­der to un­der­stand it bet­ter. It wasn’t par­tic­u­larly bad, but it was con­stant (I rated it as 210 most of the time). What stuck with me was how its con­stant pres­ence would slowly in­crease the stress of my en­tire ex­pe­rience over time. I com­pared the ex­pe­rience to hav­ing an un­com­fortable knot stuck in your body. If I had a lot of men­tal and emo­tional slack early in the day, I could eas­ily take the stress pro­duced by the knot and “send it el­se­where” in my body. But since the source of the stress was con­stant, even­tu­ally I would run out of space, and the knot would start mak­ing sec­ondary knots around it­self, and it was in those mo­ments where I would rate the pain at a 310. This would only go away if I rested and some­how “re­set” the amount of cog­ni­tive and emo­tional slack I had available.

The point of these two sto­ries is to high­light the ob­ser­va­tion that there seem to be phase-changes be­tween lev­els of dis­com­fort. An anal­ogy I of­ten make is with the phe­nomenon of sec­ondary coils when you twist a rope. The stress in­duced by pain- at least in­tro­spec­tively speak­ing- is pushed to less stressed ar­eas of your mind. But this has a limit, which is un­til your whole world-simu­la­tion is stressed to the point that the source of stress starts cre­at­ing sec­ondary “stress coils” on top of the already stressed back­ground ex­pe­rience. This was a very in­ter­est­ing re­al­iza­tion to me, which put in a differ­ent light weird ex­pres­sions that chronic pain pa­tients use like “my pain now has a pain of its own” or “I can’t let the pain build up”.

Cap­tion: DNA coils and su­per-coils as a metaphor for pain phase-changes?

Con­scious­ness Expansion

What about more ex­treme ex­pe­riences? Here we should briefly men­tion psychedelic drugs, as they seem to be able to in­crease the en­ergy of one’s con­scious­ness (and in some sense “mul­ti­ply the amount of con­scious­ness“) in a way that grows non-lin­early as a func­tion of the dose. An LSD ex­pe­rience with 100 micro­grams may be “only” 50% more in­tense than nor­mal ev­ery­day life, but an LSD ex­pe­rience with 200 micro­grams is felt as 2-3X as in­tense, while 300 micro­grams may in­crease the in­ten­sity of ex­pe­rience by per­haps 10X (rel­a­tive to nor­mal). Usu­ally peo­ple say that high-dose psychedelic states are in­de­scrib­ably more real and vivid than nor­mal ev­ery­day life. And then there are com­pounds like 5-MeO-DMT, which peo­ple of­ten de­scribe as be­ing in “a com­pletely differ­ent cat­e­gory”, as it gives rise to what many de­scribe as “in­finite con­scious­ness”. Ob­vi­ously there is no such thing as an ex­pe­rience with in­finite con­scious­ness, and that judge­ment could be ex­plained in terms of the lack of “in­ter­nal bound­aries” of the state, which gives the im­pres­sion of in­finity (not un­like how the sur­face of a torus can seem in­finite from the point of view of a flat­lander). That said, I’ve asked ra­tio­nal and in­tel­li­gent peo­ple who have tried 5-MeO-DMT in non-spiritual set­tings what they think the in­ten­sity of their ex­pe­riences was, and they usu­ally say that a strong dose of 10mg or more gives rise to an in­ten­sity and “quan­tity” of con­scious­ness that is at least 100X as high as nor­mal ev­ery­day ex­pe­riences. There are many rea­sons to be skep­ti­cal of this, no doubt, but the re­ports should not be dis­missed out of hand.

Cap­tion: Se­condary knots and links as a metaphor for higher bliss

As with the above ex­am­ple, we can rea­son that one of the ways in which both pain and plea­sure can be pre­sent in *mul­ti­ples* of one’s nor­mal he­do­nic range is be­cause the amount of con­scious­ness crammed into a mo­ment of ex­pe­rience is not a con­stant. In other words, when some­one in a typ­i­cal state of con­scious­ness asks “if you say one can ex­pe­rience so much pain/​plea­sure, tell me, where would that fit in my ex­pe­rience? I don’t see much room for that to fit in here”, one can re­spond by say­ing that “in other states of con­scious­ness there is more (phe­nom­e­nal) time and space within each mo­ment of ex­pe­rience”. In­deed, at Qualia Com­put­ing we have as­sem­bled and in­ter­preted a large num­ber of ex­pe­riences of high-en­ergy states of con­scious­ness that in­di­cate that both phe­nom­e­nal time, and phe­nom­e­nal space, can dras­ti­cally ex­pand. To sum it up – you can fit so much plea­sure and pain in peak ex­pe­riences pre­cisely be­cause such ex­pe­riences make room for them.

Let us now illus­trate the point with some paradig­matic cases of very high and vey low valence:

Peak Plea­sure States: Jhanas and Tem­po­ral Lobe Seizures

On the plea­sure side, we have Bud­dhist med­i­ta­tors who ex­pe­rience med­i­ta­tive states of ab­sorp­tion (aka. “Jhanas”) as ex­tremely, and counter-in­tu­itively, bliss­ful:

The ex­pe­rience can in­clude some very pleas­ant phys­i­cal sen­sa­tions such as goose bumps on the body and the hair stand­ing up to more in­tense plea­sures which grow in in­ten­sity and ex­plode into a state of ec­stasy. If you have pain in your legs, knees, or other part of the body dur­ing med­i­ta­tion, the pain will ac­tu­ally dis­ap­pear while you are in the jhanas. The pleas­ant sen­sa­tions can be so strong to elimi­nate your painful sen­sa­tions. You en­ter the jhanas from the pleas­ant ex­pe­riences ex­plod­ing into a state of ec­stasy where you no longer “feel” any of your senses.
9 Jhanas, Dhamma Wiki

There are 8 (or 9, de­pend­ing on who you ask) “lev­els” of Jhanas, and the above is de­scribing only the 1st of them! The higher the Jhana, the more re­fined the bliss be­comes, and the more de­tached the state is from the com­mon refer­ents of our ev­ery­day hu­man ex­pe­rience. Ul­tra-bliss does not look at all like sen­sual plea­sure or ex­cite­ment, but more like in­for­ma­tion-the­o­ret­i­cally op­ti­mal con­figu­ra­tions of res­o­nant waves of con­scious­ness with lit­tle to no in­ten­tional con­tent (cf. se­man­ti­cally neu­tral en­ergy). I know this sounds weird, but it’s what is re­ported.

Cap­tion: “Stream­lines from the in­sula to the cor­tex” – the in­sula (in red) is an area of the brain in­ti­mately im­pli­cated in the su­per-bliss that some­times pre­cedes tem­po­ral lobe epilepsy (source)

Another ex­am­ple I will provide about ul­tra-bliss con­cerns tem­po­ral lobe epilepsy, which in a minor­ity of suffer­ers gives rise to ex­traor­di­nar­ily in­tense states of plea­sure, or pain, or both. Such ex­pe­riences can re­sult in Geschwind syn­drome, a con­di­tion char­ac­ter­ized by hy­per­graphia (writ­ing non-stop), hy­per-re­li­gios­ity, and a gen­er­ally in­ten­sified men­tal and emo­tional life. No doubt, any ex­pe­rience that hits the valence scale at one of its ex­tremes is usu­ally in­ter­preted as other-wor­ldly and para­nor­mal (which gives rise to the ques­tion of whether valence is a spiritual phe­nomenon or the other way around). Fa­mously, Dos­to­evsky seems to have ex­pe­rienced tem­po­ral lobe seizures, and this ul­ti­mately in­formed his wor­ld­view and liter­ary work in profound ways. Here is how he de­scribes them:

“A hap­piness un­think­able in the nor­mal state and uni­mag­in­able for any­one who hasn’t ex­pe­rienced it… I am then in perfect har­mony with my­self and the en­tire uni­verse.”
– From a let­ter to his friend Niko­lai Strakhov.
“I feel en­tirely in har­mony with my­self and the whole world, and this feel­ing is so strong and so delight­ful that for a few sec­onds of such bliss one would gladly give up 10 years of one’s life, if not one’s whole life. […] You all, healthy peo­ple, can’t imag­ine the hap­piness which we epilep­tics feel dur­ing the sec­ond be­fore our fit… I don’t know if this felic­ity lasts for sec­onds, hours or months, but be­lieve me, I would not ex­change it for all the joys that life may bring.”
– from the char­ac­ter Prince Myshkin in Dos­to­evsky’s novel, The Idiot, which he likely used to give a voice to his own ex­pe­riences.

Dos­to­evsky is far from the only per­son re­port­ing these kinds of ex­pe­riences from epilepsy:

As Pi­card [a sci­en­tist in­ves­ti­gat­ing seizures] ca­joled her pa­tients to speak up about their ec­static seizures, she found that their sen­sa­tions could be char­ac­ter­ised us­ing three broad cat­e­gories of feel­ings (Epilepsy & Be­havi­our, vol 16, p 539). The first was height­ened self-aware­ness. For ex­am­ple, a 53-year-old fe­male teacher told Pi­card: “Dur­ing the seizure it is as if I were very, very con­scious, more aware, and the sen­sa­tions, ev­ery­thing seems big­ger, over­whelming me.” The sec­ond was a sense of phys­i­cal well-be­ing. A 37-year-old man de­scribed it as “a sen­sa­tion of velvet, as if I were sheltered from any­thing nega­tive”. The third was in­tense pos­i­tive emo­tions, best ar­tic­u­lated by a 64-year-old woman: “The im­mense joy that fills me is above phys­i­cal sen­sa­tions. It is a feel­ing of to­tal pres­ence, an ab­solute in­te­gra­tion of my­self, a feel­ing of un­be­liev­able har­mony of my whole body and my­self with life, with the world, with the ‘All’,” she said.
– from “Fits of Rap­ture”, New Scien­tist (Jan­uary 25, 2014) (source)

All in all, these ex­am­ples illus­trate the fact that bliss­ful states can be deeper, richer, more in­tense, more con­scious, and qual­i­ta­tively su­pe­rior to the nor­mal ev­ery­day range of hu­man emo­tion.

Now, how about the nega­tive side?

Log­a­r­ith­mic Pain Scales: St­ings, Pep­pers, and Cluster Headaches

“The differ­ence be­tween 6 and 10 on the pain scale is an ex­po­nen­tial differ­ence. Believe it or not.”
In­suffer­able In­differ­ence, by Neil E. Cle­ment (who ex­pe­riences chronic pain rang­ing be­tween 610 to 1010, de­pend­ing on the day)

Three pain-scale ex­am­ples that illus­trate the non-lin­ear­ity of pain are: (1) the Sch­midt sting pain in­dex, (2) the Scov­ille scale, and (3) the KIP scale:

(1) Justin O. Sch­midt stung him­self with over 80 species of in­sects of the Hy­menoptera or­der, and rated the en­su­ing pain on a 4-point-scale. About the scale, he had to say the fol­low­ing:

4:28 – Justin Sch­midt: The har­vester ant is what got the sting pain scale go­ing in the first place. I had been stung by hon­ey­bees, yel­low jack­ets, pa­per wasps, etc. the gar­den va­ri­ety stuff, that you get bit­ten by var­i­ous bee­tles and things. I went down to Ge­or­gia, which has the Eastern-most ex­ten­sion of the har­vester ant. I got stung and I said “Wooooow! This is DIFFERENT!” You know? I thought I knew ev­ery­thing there was about in­sect stings, I was just this dumb lit­tle kid. And I re­al­ized “Wait a minute! There is some­thing differ­ent go­ing on here”, and that’s what got me to do the com­par­a­tive anal­y­sis. Is this unique to har­vester ants? Or are there oth­ers that are like that. It turns out while the an­swer is, now we know much later – it’s unique! [unique type of pain].
7:09 – Justin Sch­midt: I didn’t re­ally want to go out and get stung for fun. I was this des­per­ate grad­u­ate stu­dent try­ing to get a the­sis, so I could get out and get a real job, and stop be­ing a stu­dent even­tu­ally. And I re­al­ized that, oh, we can mea­sure tox­i­c­ity, you know, the kil­ling power of some­thing, but we can’t mea­sure pain… ouch, that one hurts, and that one hurts, and ouch that one over there also hurts… but I can’t put that on a com­puter pro­gram and math­e­mat­i­cally an­a­lyze what it means for the pain of the in­sect. So I said, aha! We need a pain scale. A com­puter can an­a­lyze one, two, three, and four, but it can’t an­a­lyze “ouch!”. So I de­cided that I had to make a pain scale, with the har­vester ant (cut­ting to the chase) was a 3. Honey bees was a 2. And I kind of tell peo­ple that each num­ber is like 10 equiv­a­lent of the num­ber be­fore. So 10 honey bee stings are equal to 1 har­vester ant sting, and 10 har­vester ant stings would equal one bul­let ant sting.
11:50 – [In­ter­viewer]: When I fi­nally worked up the courage to [put the Taran­tula Hawk on my arm] and take this sting. The sting of that in­sect was elec­tric in na­ture. I’ve been shocked be­fore, by ac­ci­den­tally tak­ing a zap from an elec­tri­cal cord. This was that times 10. And it put me on the ground. My arm seized up from mus­cle con­trac­tion. And it was prob­a­bly the worst 5 min­utes of my life at that point.
Justin Sch­midt: Yeah, that’s ex­actly what I call elec­trify­ing. I say, imag­ine you are walk­ing along in Ari­zona, and there is a wind storm, and the power line above snaps the wire, and it hits you, of course that hasn’t hap­pened to me, but that’s what you imag­ine it feels like. Be­cause it’s ab­solutely elec­trify­ing, I call it de­bil­i­tat­ing be­cause you want to be ma­cho, “ah I’m tough, I can do this!” Now you can’t! So I tell peo­ple lay down and SCREAM! Right?
[In­ter­viewer]: That’s what I did! And Mark would be like, this fa­mous “Coy­ote, are you ok? Are you ok?”
Justin Sch­midt: No, I’m not ok!
[In­ter­viewer]: And it was very hard to try to com­pose my­self to be like, alright, de­scribe what is hap­pen­ing to your body right now. Be­cause your mind goes into this state that is like blank empti­ness. And all you can fo­cus on is the fact that there’s ra­di­at­ing pain com­ing out of your arm.
Justin Sch­midt: That’s why you scream, be­cause now you’re fo­cus­ing on some­thing else. In ad­di­tion to the pain, you’re fo­cus­ing on “AAAAAAHHHHH!!!” [screams loudly]. Takes a lit­tle bit of the juice off of the pain, so maybe you lower it down to a three for as long as you can yell. And I can yell for a pretty long time when I’m stung by a taran­tula hawk.

Ori­gin of STINGS!, in­ter­view of Justin O. Schmidt

If we take Justin’s word for it, a sting that scores a 4 on his pain scale is about 1,000 times more painful than a sting that scores a 1 on his scale. Ac­cord­ingly, Christo­pher Starr (who repli­cated the scale), stated that any sting that scores a 4 is “trau­mat­i­cally painful” (source). Fi­nally, since the scale is re­stricted to stings of in­sects of the Hy­menoptera or­der, it re­mains pos­si­ble that there are stings whose pain would be rated even higher than 4. A 5 on the sting pain in­dex might per­haps be ex­pe­rienced with the stings of the box jel­lyfish that pro­duces Irukandji syn­drome, and the bite of the gi­ant desert cen­tipede. Need­less to say, these are to be avoided.

Mov­ing on…

Cap­tion: (source)

(2) The Scov­ille scale mea­sures how spicy differ­ent chili pep­pers and hot sauces are. It is calcu­lated by dilut­ing the pep­per/​sauce in wa­ter un­til it is no longer pos­si­ble to de­tect any spice in it. The num­ber that is as­so­ci­ated with the pep­per or sauce is the ra­tio of wa­ter-to-sauce that makes it just barely pos­si­ble to taste the spice. Now, this is of course not it­self a pain scale. I would nonethe­less an­ti­ci­pate that tak­ing the log of the Scov­ille units of a dish might be a good ap­prox­i­ma­tion for the re­ported pain it de­liv­ers. In par­tic­u­lar, peo­ple note that there are sev­eral qual­i­ta­tive jumps in the type and na­ture of the pain one ex­pe­riences when eat­ing hot sauces of differ­ent strengths (e.g. “Fuck you Sean! […] That was a leap, Sean, that was a LEAP!” – Ken Jeong right af­ter get­ting to the 135,000 Scov­ille units sauce in the pain porn Youtube se­ries Hot Ones). Ama­zon re­views of ul­tra-hot sauces can be mined for phe­nomenolog­i­cal in­for­ma­tion con­cern­ing in­tense pain, and the gen­eral im­pres­sion one gets af­ter read­ing such re­views is that in­deed there is a sort of ex­po­nen­tial range of pos­si­ble pain val­ues:

I know it may be fun to triv­ial­ize this kind of pain, but differ­ent peo­ple re­act differ­ently to it (prob­a­bly fol­low­ing a long-tail too!). For some peo­ple who are very sen­si­tive to heat pain, very hot sauce can be le­gi­t­i­mately trau­ma­tiz­ing. Hence I ad­vise against hav­ing ul­tra-spicy sauces around your house. The nov­elty value is not worth the prob­a­bil­ity of a re­gret­table ac­ci­dent, as ex­em­plified in some of the Ama­zon re­views above (e.g. a house guest as­sum­ing that your “Da’Bomb – Beyond In­san­ity” bot­tle in the fridge can’t pos­si­bly be that hot… and end­ing up in the ER and with PTSD).

I should add that me­dia that is widely con­sumed about ex­treme hot sauce (e.g. the Hot Ones men­tioned above and nu­mer­ous stunt Youtube chan­nels) may seem fun on the sur­face, but what doesn’t make the cut and is left in the edit­ing room is prob­a­bly not very palat­able at all. From an in­ter­view: “Has any­one thrown up do­ing it?” (in­ter­viewer) – “Yeah, we’ve run the gamuts. We’ve had peo­ple spit in buck­ets, half-pass out, sleep in the green room af­ter­wards, etc.” (Sean Evans, Hot Ones host). T.J. Miller, when asked about what ad­vice he would give to the show while eat­ing ul­tra-spicy wings, re­sponded: “Don’t do this. Don’t do this again. End the show. Stop do­ing the show. That’s my ad­vice. This is very hot. This is painful. There’s a prob­lem here.”

Cap­tion: Trigem­i­nal Neu­ral­gia pain scale – a con­di­tion similarly painful to Cluster Headaches

(3) Fi­nally, we come to the “KIP scale”, which is used to rate Cluster Headaches, one of the most painful con­di­tions that peo­ple en­dure:

The KIP scale
KIP-0 No pain, life is beau­tiful
KIP-1 Very minor, shad­ows come and go. Life is still beau­tiful
KIP-2 More per­sis­tent shad­ows
KIP-3 Shad­ows are get­ting con­stant but can deal with it
KIP-4 Start­ing to get bad, want to be left alone
KIP-5 Still not a “pacer” but need space
KIP-6 Wake up grum­bling, curse a bit, but can get back to sleep with­out “danc­ing”
KIP-7 Wake up, sleep not an op­tion, take the beast for a walk and fi­nally fall into bed ex­haust­ed
KIP-8 Time to scream, yell, curse, head bang, rock, what­ever work­s
KIP-9 The “Why me?” syn­drome starts to set in­
KIP-10 Ma­jor pain, scream­ing, head bang­ing, ER trip. De­pressed. Suici­dal.
The du­ra­tion fac­tor is mul­ti­plied by the in­ten­sity fac­tor, which uses the KIP scale in an ex­po­nen­tial way – a KIP 10 is not just twice as bad as a KIP 5, it’s ten times as in­tense.
Source: Keep­ing Track, by Cluster Busters

As seen above, the KIP scale is ac­knowl­edged by its cre­ator and users to be log­a­r­ith­mic in na­ture.

In sum­mary: We see that plea­sure comes in var­i­ous grades and that peak ex­pe­riences such as those in­duced by psychedelics, med­i­ta­tion, and tem­po­ral lobe seizures seem to be or­ders of mag­ni­tude more en­er­getic and bet­ter than ev­ery­day sober states. Like­wise, we see that across sev­eral cat­e­gories of pain, peo­ple re­port be­ing sur­prised by the leaps in both qual­ity and in­ten­sity that are pos­si­ble. More so, at least in the case of the Sch­midt In­dex and the Kip Scale, the cre­ators of the scale were ex­plicit that it was a log­a­r­ith­mic map­ping of the ac­tual level of sen­sa­tion.

While we do not have enough ev­i­dence (and con­cep­tual clar­ity) to as­sert that the in­ten­sity of pain and plea­sure does grow ex­po­nen­tially, the in­for­ma­tion pre­sented so far does sug­gest that the valence of ex­pe­riences fol­lows a long-tail dis­tri­bu­tion.

Defer­ence-type Ap­proaches for Ex­pe­rience Ranking

The above con­sid­er­a­tions un­der­score the im­por­tance of com­ing up with a plea­sure-pain scale that tries to take into ac­count the non-lin­ear­ity and non-nor­mal­ity of valence rat­ings. One idea we came up with was a “defer­ence”-type ap­proach, where we ask open-ended ques­tions about peo­ple’s best and worst ex­pe­riences and have them rank them against each other. Although lo­cally the data would be very sparse, the idea was that there might be meth­ods to in­te­grate the col­lec­tive pat­terns of defer­ence into an ap­prox­i­mate scale. If ex­tended to pop­u­la­tions of peo­ple who are known to have ex­pe­rienced ex­tremes of valence, the ap­proach would even al­low us to unify the var­i­ous pain scales (Scov­ille, Sch­midt, KIP, etc.) and as­sign a kind of uni­ver­sal valence score to differ­ent cat­e­gories of pain and plea­sure.* That will be ver­sion 2.0. In the mean­time, we thought to try to get a rough pic­ture of the ex­treme joys and af­fec­tions of mem­bers of the gen­eral pub­lic, which is what this ar­ti­cle will fo­cus on.

Nor­mal World vs. Log­nor­mal World

There is a world we could call the “Nor­mal World”, where valence out­liers are rare and most types of ex­pe­riences af­fect peo­ple more or less similarly, dis­tributed along a Gaus­sian curve. Then there is an­other, very differ­ent world we could call the “long-tailed world” or if we want to make it sim­ple (ac­knowl­edg­ing un­cer­tainty) “Log­nor­mal World”, where al­most ev­ery valence dis­tri­bu­tion is a long-tail. So in the “Log­nor­mal World”, say, for plea­sure (and sym­met­ri­cally for pain), we would ex­pect to see a long-tail in the mean plea­sure of ex­pe­riences be­tween differ­ent cat­e­gories across all peo­ple, a long-tail in the amount of plea­sure within a given type of ex­pe­rience across peo­ple, a long-tail for the num­ber of times an in­di­vi­d­ual has had a cer­tain type of plea­sure, a long-tail in the in­ten­sity of the plea­sure ex­pe­rienced with a sin­gle cat­e­gory of ex­pe­rience within a sin­gle per­son, and so on. Do we live in the Nor­mal World or the Log­nor­mal World?

Pre­dic­tions of Log­nor­mal World

If we lived in the “Log­nor­mal World”, we would ex­pect:

  • 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.

  • 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%).

  • 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).

To test the first and sec­ond pre­dic­tion does not re­quire a lot of data, but the third does be­cause one needs to have enough com­par­i­sons to fill a lot of tri­ads. The sur­vey re­sults we will dis­cuss bel­low are con­gru­ent with the first and sec­ond pre­dic­tion. We did what we could with the data available to in­ves­ti­gate the third, and ten­ta­tively, it seems to hold up (with ideas like defer­ence net­work cen­tral­ity anal­y­sis, tri­adic anal­y­sis, and tour­na­ment-style ap­proaches).

Sur­vey Setup

The sur­vey asked the fol­low­ing ques­tions: cur­rent level of plea­sure, cur­rent level of pain, top 3 most plea­surable ex­pe­riences (in de­creas­ing or­der) along with plea­sure rat­ings for each of them and the age when they were ex­pe­rienced, and the same for the top 3 most painful ex­pe­riences. I speci­fi­cally did not provide a set of broad cat­e­gories (such as “phys­i­cal” or “emo­tional”) or a drop-down menu of pos­si­ble nar­row cat­e­gories (e.g. go­ing to the movies, aer­o­bic ex­er­cise, etc.). I wanted to see what peo­ple would say when the ques­tion was as open-ended as pos­si­ble.

I also in­cluded ques­tions aimed more di­rectly at prob­ing the long-tailed na­ture of valence: I asked par­ti­ci­pants to rate “how many times more pleas­ant was the #1 top ex­pe­rience rel­a­tive to the #2 top ex­pe­rience” (and #2 rel­a­tive to #3, and the same for the top most painful ex­pe­riences).

I also asked them to de­scribe in more de­tail the sin­gle most pleas­ant and un­pleas­ant ex­pe­riences, and added a box for com­ments at the end in or­der to see if any­one com­plained about the task (most peo­ple said “no com­ment”, many said they en­joyed the task, and one per­son said that it made them nos­talgic). I also asked about ba­sic de­mo­graph­ics (age and gen­der). Par­ti­ci­pants earned $1.75 for the task, which seems rea­son­able given the time it took to com­plete in most cases.

Me­chan­i­cal Turk: Par­ti­ci­pant Composition

The sur­vey was run on Me­chan­i­cal Turk. I re­quested “Masters” qual­ifi­ca­tions for 90110 of the sur­vey re­sponses. To be a Me­chan­i­cal Turk Master you need to have a good track record of task ap­proval, which I rea­soned would in­crease the rate of well thought-out high-qual­ity re­sponses (this turned out to be true). The par­ti­ci­pants’ gen­ders were fairly bal­anced (50/​97 women, 4797 men), and the av­er­age age was 36.5 years (sd = 10):

Bots (easy to spot – un­less ex­tremely so­phis­ti­cated)

While read­ing the re­sponses I dis­cov­ered that some of them seemed to be gen­er­ated by bots. Ap­par­ently task re­questers started notic­ing the pres­ence of bots a cou­ple of years ago. This is ob­vi­ously a prob­lem for aca­demics try­ing to use the ser­vice for their re­search, and for ma­chine learn­ing pro­fes­sion­als us­ing the ser­vice for data-tag­ging. Hav­ing been both of those things in the past, I can say that a few years ago I did not see any sub­mis­sions that looked sus­pi­ciously bot-like. Though low-qual­ity rushed re­sponses did seem to be rel­a­tively com­mon back then, I never ob­served bot-like re­sponses to open-ended ques­tions. Un­for­tu­nately this time I was able to spot sev­eral re­sponses clearly writ­ten by bots. For ex­am­ple, one re­spon­dent an­swered the ques­tion “write a brief es­say about your #1 best ex­pe­rience” with:

The rea­son these types of as­sign­ments are given so of­ten is that any­one can write about their own ex­pe­rience and it doesn’t re­quire any out­side re­sources or re­search. How­ever, even though any­one can tell a story about their life, that does not mean any­one can write a good es­say about that ex­pe­rience. As a pro­fes­sor and teacher for 30 years, I’ve read thou­sands of es­says and can tell you there is a dis­tinct differ­ence from tel­ling a story about your­self and writ­ing an ex­cel­lent per­sonal ex­pe­rience es­say. The differ­ence be­tween good and great:

And it ends that way, with a colon in­di­cat­ing that the re­spon­dent is about to ex­plain what the differ­ence be­tween good and great is. But it never does it. This an­swer, great, it is not.

In most cases the differ­ence be­tween a gen­uine re­sponse and a bot re­sponse was very ob­vi­ous. That said, I erred on the side of cau­tion for fil­ter­ing bots and I got rid of an­swers even if they seemed just a lit­tle sus­pi­cious. This left me with 97 out of the 110 origi­nal re­sponses. The fol­low­ing anal­y­sis was con­ducted on those 97 re­sponses.


Since the re­sponses were open-ended I had to tag each of them with an ex­pe­rience cat­e­gory. To do this I read each re­sponse and iden­ti­fied the key theme in them and clas­sified them with a la­bel that was spe­cific enough to dis­t­in­guish it from nearby ex­pe­riences (e.g. differ­ent types of frac­tures), but not so spe­cific that we would never get more than one re­sponse per cat­e­gory (e.g. “break­ing the mid­dle finger in el­e­men­tary school”). In gen­eral, most re­sponses fell into very un­am­bigu­ous cat­e­gories (e.g. “When my father passed away” and “Watch­ing my father die and take his last few breaths.” were both clas­sified as “Father death”). About 10% of the re­sponses were rel­a­tively am­bigu­ous: it wasn’t clear what the source of the pain or plea­sure was. To deal with those re­sponses I used the la­bel “Un­speci­fied”. When some de­tail was pre­sent but am­bi­guity re­mained, such as when a broad type of pain or plea­sure was men­tioned but not the spe­cific source I tagged it as “Un­speci­fied X” where X was a broad cat­e­gory. For ex­am­ple, one per­son said that “bro­ken bones” was the most painful ex­pe­rience they’ve had, which I la­beled as “Un­speci­fied frac­ture”.


I should pref­ace the fol­low­ing by say­ing that we are very aware of the lack of sci­en­tific rigor in this sur­vey; it re­mains a pi­lot ex­plo­ra­tory work. We didn’t spec­ify the time-scale for the ex­pe­riences (e.g. are we ask­ing about the best minute of your life or the best month of your life?) or whether we were re­quest­ing in­stances of phys­i­cal or psy­cholog­i­cal pain/​plea­sures. De­spite this lack of con­straints it was in­ter­est­ing to see very strong com­mon­al­ities among peo­ple’s re­sponses:

Ap­pear­ance Base Rates

There were 77 and 124 cat­e­gories of plea­sure and pain iden­ti­fied, re­spec­tively. On the whole it seemed like there was a higher di­ver­sity of ways to suffer than of ways to ex­pe­rience in­tense bliss. Sum­mon­ing the spirit of Tols­toy: “Happy fam­i­lies are all al­ike; ev­ery un­happy fam­ily is un­happy in its own way.”

Here are the raw counts for each cat­e­gory with at least two ap­pear­ances:

Cap­tion: Best ex­pe­riences ap­pear­ances (with at least two re­ports)
Cap­tion: Worst ex­pe­rience ap­pear­ances (with at least two re­ports)

For those who want to see the full list of num­ber of ap­pear­ances for each ex­pe­rience men­tioned see the bot­tom of the ar­ti­cle (I also clar­ify some of the more con­fus­ing la­bels there too)**.

A sim­ple way to try to in­cor­po­rate the in­for­ma­tion about the rank­ing is to weight ex­pe­riences rated as top #1 with 3 points, those as top #2 with 2 points, and those as the top #3 with 1 point. If you do this, the ex­pe­riences scores are:

Cap­tion: Weighted ap­pear­ances of best ex­pe­riences (#1 – 3 points, #2 – 2 points, #3 – 1 point)
Cap­tion: Weighted ap­pear­ances of worst ex­pe­riences (#1 – 3 points, #2 – 2 points, #3 – 1 point)

Aver­age ratings

Given the rel­a­tively small sam­ple size, I will only re­port the mean rat­ing for pain and plea­sure (out of 10) for cat­e­gories of ex­pe­rience for which there were 6 or more re­spon­dents:

For pain:

  1. Father death (n = 19): mean 8.53, sd 2.3

  2. Child­birth (n = 16): mean 7.94, sd 2.16

  3. Grand­mother death (n = 13): mean 8.12, sd 2.5

  4. Mother death (n = 11): mean 9.4, sd 0.62

  5. Car ac­ci­dent (n = 9): mean 8.42, sd 1.52

  6. Kid­ney stone (n = 9): mean 5.97, sd 3.17

  7. Mi­graine (n = 9): mean 5.36, sd 3.11

  8. Ro­man­tic breakup (n = 9): mean 7.11, sd 1.52

  9. Bro­ken arm (n = 6): mean 8.28, sd 0.88

  10. Bro­ken leg (n = 6): mean 7.33, sd 2.02

  11. Work failure (n = 6): mean 5.88, sd 3.57

(Note: the very high var­i­ance for kid­ney stones and mi­graine is partly ex­plained by the pres­ence of some very low re­sponses, with val­ues as low as 1.1/​10 – per­haps mis­re­ported, or per­haps illus­trat­ing the ex­treme di­ver­sity of ex­pe­riences of mi­graines and kid­ney stones).

And for plea­sure:

  1. Fal­ling in love (n = 42): mean 8.68, sd 1.74

  2. Chil­dren born (n = 41): mean 9.19, sd 1.64

  3. Mar­riage (n = 21): mean 8.7, sd 1.25

  4. Sex (n = 19): mean 8.72, sd 1.45

  5. Col­lege grad­u­a­tion (n = 13): mean 7.73, sd 1.4

  6. Or­gasm (n = 11): mean 8.24, sd 1.63

  7. Al­co­hol (n = 8): mean 6.84, sd 1.59

  8. Va­ca­tion (n = 6): mean 9.12, sd 0.73

  9. Get­ting job (n = 6): mean 7.22, sd 1.47

  10. Per­sonal fa­vorite sports win (n = 6): mean 8.17, sd 1.23

Defer­ence Graph of Top Experiences

We will now fi­nally get to the more ex­plo­ra­tory and fun/​in­ter­est­ing anal­y­sis, at least in that it will gen­er­ate a cool way of vi­su­al­iz­ing what causes peo­ple great joy and pain. Namely, the idea of us­ing peo­ple’s rank­ings in or­der to pop­u­late a global scale across peo­ple and show it in the form of a graph of defer­ences. While the sci­en­tific liter­a­ture has some stud­ies that com­pare pain across differ­ent cat­e­gories (e.g. 1, 2, 3) I was not able to find any dataset that in­cluded ac­tual rank­ings across a va­ri­ety of cat­e­gories. Hence why it was so ap­peal­ing to vi­su­al­ize this.

The sim­plest way of graph­ing ex­pe­rience defer­ences is to as­sign a node to each ex­pe­rience cat­e­gory and add an edge be­tween ex­pe­riences with defer­ence re­la­tion­ships with a weight pro­por­tional to the num­ber of di­rected defer­ences. For ex­am­ple, if 4 peo­ple have said that A was bet­ter than B, and 3 peo­ple have said that B was bet­ter than A, then there will be an edge from A to B with a weight of 4 and an edge from B to A with a weight of 3. Ad­di­tion­ally, we can then run a graph cen­tral­ity al­gorithm such as PageRank to see where the “defer­ences end up pool­ing”.

The images be­low do this: the PageRank of the graph is rep­re­sented with the color gra­di­ent (darker shades of green/​red rep­re­sent­ing higher PageRank val­ues for good/​bad ex­pe­riences). In ad­di­tion, the graphs also rep­re­sent the num­ber of ap­pear­ances in the dataset for each cat­e­gory with the size of each node:

Cap­tion: Best ex­pe­riences defer­ences – edge thick­ness based on num­ber of defer­ences, node size based on num­ber of ap­pear­ances, and color scheme based on PageRank
Cap­tion: Worst ex­pe­riences defer­ences – edge thick­ness based on num­ber of defer­ences, node size based on num­ber of ap­pear­ances, and color scheme based on PageRank

The main prob­lem with the ap­proach above is that it dou­ble (triple?) counts ex­pe­riences that are very com­mon. Say that, for ex­am­ple, tak­ing 5-MeO-DMT pro­duces a con­sis­tently higher-valence feel­ing rel­a­tive to hav­ing sex. If we only have a cou­ple of peo­ple who re­port both 5-MeO-DMT and sex as their top ex­pe­riences, the edge from sex to 5-MeO-DMT will be very weak, and the PageRank al­gorithm will un­der­es­ti­mate the value of 5-MeO-DMT.

In or­der to avoid the dou­ble count­ing effect of com­monly-re­ported peak ex­pe­riences we can in­stead add edge weights on the ba­sis of the pro­por­tion with which an ex­pe­rience defers to the other. Let’s say that f(a, b) means “num­ber of times that b is re­ported as higher than a”. Then the pro­por­tion would be f(a, b) /​ (f(a, b) + f(b, a)). Now, this in­tro­duces an­other prob­lem, which is that pairs of ex­pe­riences that ap­pear to­gether very in­fre­quently might get a very high pro­por­tion score due to a low sam­ple size. In or­der to pre­vent this we use Laplace smooth­ing and mod­ify the equa­tion to (f(a, b) + 1) /​ (f(a, b) + f(b, a) + 2). Fi­nally, we trans­form this pro­por­tion score from the range of 0 to 1 to the range of −1 to 1 by mul­ti­ply­ing by 2 and sub­tract­ing one. We call this a “re­bal­anced smoothed pro­por­tion” w(a, b):

I should note that this is not based on any rigor­ous math. The equa­tion is based on my in­tu­ition for what I would ex­pect to see in such a graph, namely a sort of con­fi­dence-weighted strength of di­rec­tion­al­ity, but I do not guaran­tee that this is a prin­ci­pled way of do­ing so (did I men­tion this is a pi­lot small-scale low-bud­get ‘to a first ap­prox­i­ma­tion’ study?). I think that, nonethe­less, do­ing this is still an im­prove­ment upon merely us­ing the raw defer­ence counts as the edge weights. To vi­su­al­ize what w(a, b) looks like I graphed its val­ues for a and b in the range of 0 to 20 (liter­ally typ­ing the equa­tion into the google search bar):

Cap­tion: Re­bal­anced smoothed pro­por­tion equa­tion w(a, b)

To pop­u­late the graph I only use the pos­i­tive edge weights so that we can run the PageRank al­gorithm on it. This now looks a lot more rea­son­able and in­for­ma­tive as a defer­ence graph than the pre­vi­ous at­tempts:

Cap­tion: Best ex­pe­riences defer­ence graph: Edge weights based on the re­bal­anced smoothed pro­por­tions, size of nodes is pro­por­tional to num­ber of ap­pear­ances in the dataset, and the color tracks the PageRank of the graph. Edge color based on source node.
Cap­tion: Worst ex­pe­riences defer­ence graph: Edge weights based on the re­bal­anced smoothed pro­por­tions, size of nodes is pro­por­tional to num­ber of ap­pear­ances in the dataset, and the color tracks the PageRank of the graph. Edge color based on source node.

By tak­ing the PageRank of these graphs (calcu­lated with Net­workX) we ar­rive at the fol­low­ing global rank­ings:

Cap­tion: PageRank of the graph of best ex­pe­riences with edge weights com­puted with the re­bal­anced smoothed pro­por­tion equation
Cap­tion: PageRank of the graph of worst ex­pe­riences with edge weights com­puted with the re­bal­anced smoothed pro­por­tion equation

In­tu­itively this rank­ing seems more al­igned with what I’ve heard be­fore, but I will with­hold judge­ment on it un­til we have much more data.

Tri­adic Analysis

With a more pop­u­lated defer­ence graph we can an­a­lyze in de­tail the de­gree to which tri­ads (i.e. sets of three ex­pe­riences such that each of the three pos­si­ble defer­ences are pre­sent in the graph) show tran­si­tivity (cf. Balance vs. Sta­tus The­ory).

In par­tic­u­lar, we should com­pare the prevalence of these two tri­ads:

Cap­tion: Left: 030T, Right: 030C (source)

The tri­ads above are 030T, which is tran­si­tive, and 030C, which is a loop. The higher the de­gree of agree­ment be­tween peo­ple and the higher the prob­a­bil­ity of the ex­is­tence of an un­der­ly­ing shared scale, we would ex­pect to see more tri­ads of the type 030T rel­a­tive to 030C. That said, a sim­ple ra­tio is not enough, since the ex­pected pro­por­tion be­tween these two tri­ads can be an ar­ti­fact of the way the graph is con­structed and/​or its gen­eral shape (and hence the im­por­tance of com­par­ing against ran­dom­ized graphs that pre­serve as many other statis­ti­cal fea­tures as pos­si­ble). With our graph, we no­ticed that the very way in which the edges were in­tro­duced gen­er­ated an ar­ti­fact of a very strong differ­ence be­tween these two types of tri­ads:

In the case of pain there are 105 ‘030T’, and 3 ‘030C’. And for the plea­sure ques­tions there were 98 ‘030T’, and 9 ‘030C’. That said, many of these tri­ads are the ar­ti­fact of tak­ing into ac­count the top three ex­pe­riences, which already gen­er­ates a tran­si­tive triad by de­fault when n = 1 for that par­tic­u­lar triad of ex­pe­riences. To avoid this ar­ti­fact, we filtered the graph by only adding edges when a pair of ex­pe­riences ap­peared at least twice (and dis­count­ing the edges where w(a, b) = 0). With this ad­just­ment we got 2 ‘030T’, and 1 ‘030C’ for the pain ques­tions, and 1 ‘030T’, and 0 ‘030C’ for the plea­sure ques­tion. Clearly there is not enough data to mean­ingfully con­duct this type of anal­y­sis. If we ex­tend the study and get a larger sam­ple size, this anal­y­sis might be much more in­for­ma­tive.

La­tent Trait Ratings

A fi­nal ap­proach I tried for de­riv­ing a global rank­ing of ex­pe­riences was to as­sume a la­tent pa­ram­e­ter for pain or plea­sure of differ­ent ex­pe­riences and treat­ing the rank­ings as the tour­na­ment re­sults of par­ti­ci­pants with skill equal to this la­tent trait. So when some­one says that an ex­pe­rience of sex was bet­ter than an ex­pe­rience of get­ting a new bike we imag­ine that “sex” had a match with “get­ting bike” and that “sex” won that match. If we do this, then we can im­port any of the many tour­na­ment al­gorithms that ex­ist (such as the Elo rat­ing sys­tem) in or­der to ap­prox­i­mate the la­tent “skill” trait of each ex­pe­rience (ex­cept that here it is the “skill” to cause you plea­sure or pain, rather than any kind of gam­ing abil­ity).

In­ter­est­ingly, this strat­egy has also been used in other ar­eas out­side of ac­tual tour­na­ments, such as de­riv­ing uni­ver­sity rank­ings based on the choices made by stu­dents ad­mit­ted to more than one col­lege (see: Re­vealed Prefer­ence Rank­ings of US Col­leges and Univer­si­ties).

I should men­tion that the fact that we are ask­ing about peak ex­pe­riences likely vi­o­lates some of the as­sump­tions of these al­gorithms, since the fact that a match takes place is already in­for­ma­tion that both ex­pe­riences made it into the top 3. That said, if the pat­terns of defer­ence are very strong, this might not rep­re­sent a prob­lem.

To come up with this tour­na­ment-style rank­ing I de­cided to go for a state-of-the-art al­gorithm. The one that I was able to find and use was Microsoft Re­search’s al­gorithm called TrueSkill (which is em­ployed to rank play­ers in Xbox LIVE). Ac­cord­ing to their doc­u­men­ta­tion, to ar­rive at a con­ser­va­tive “leader­board” that bal­ances the es­ti­mated “true skill” and the un­cer­tainty around it, they recom­mend rank­ing by the ex­pected skill level minus three times the stan­dard er­ror around this es­ti­mate. If we do this, we ar­rive at the fol­low­ing ex­pe­rience “leader­boards”:

Cap­tion: Con­ser­va­tive TrueSkill scores for best ex­pe­riences (mu – 3*sigma)
Cap­tion: Con­ser­va­tive TrueSkill scores for worst ex­pe­riences (mu – 3*sigma)

Long-tails in Re­sponses to “How Many Times Bet­ter/​Worse” Question

The sur­vey in­cluded four ques­tions aimed at com­par­ing the rel­a­tive he­do­nic val­ues of peak ex­pe­riences: “Rel­a­tive to the 1st most pleas­ant ex­pe­rience, how many times bet­ter was the 2nd most pleas­ant ex­pe­rience?” (This was one, the other three were the per­mu­ta­tions of also ask­ing about 2nd vs. 3rd and about the bad ex­pe­riences):

(Note: I’ll ig­nore the re­sponses to the com­par­i­son be­tween the 2nd and 3rd worst pains be­cause I messed up the ques­tion -I for­got to sub­sti­tute “bet­ter” for “worse”).

I would un­der­stand the skep­ti­cism about these graphs. But at the same time, I don’t think it is ab­surd that for many peo­ple the worst ex­pe­rience they’ve had is in­deed 10 or 100 times worse than the sec­ond worst. For ex­am­ple, some­one who has en­dured a bad Cluster Headache will gen­er­ally say that the pain of it is tens or hun­dreds of times worse than any other kind of pain they have had (say, break­ing a bone or hav­ing skin burns).

The above dis­tri­bu­tions sug­gest a long-tail for the he­do­nic qual­ity of ex­pe­riences: say that the he­do­nic qual­ity of each day is dis­tributed along a log-nor­mal dis­tri­bu­tion. A 45 year old has ex­pe­rienced roughly 17,000 days. Let’s say that such a per­son’s ex­pe­rience of pain each day is sam­pled from a log-nor­mal dis­tri­bu­tion with a Gaus­sian ex­po­nent with a mean of 10 and a stan­dard de­vi­a­tion of 5. If we take 100 such peo­ple, and for each of them we take the sin­gle worst and the sec­ond worst days of their lives, and then take the ra­tio be­tween them, we will have a dis­tri­bu­tion like this (simu­lated in R):

If you smooth the em­piri­cal curves above you would get a dis­tri­bu­tion that looks like these simu­la­tions. You re­ally need a long-tail to be able to get re­sults like “for 25% of the par­ti­ci­pants the sin­gle worst ex­pe­rience was at least 4 times as bad as the 2nd worst ex­pe­rience.” Com­pare that to the sort of pat­tern that you get if the dis­tri­bu­tion was nor­mal rather than log-nor­mal:

As you can see (zoom­ing in on the y-axis), the ra­tios sim­ply do not reach very high val­ues. With the nor­mal dis­tri­bu­tion simu­lated here, we see that the high­est ra­tio we achieve is around 1.3, as op­posed to the em­piri­cal ra­tios of 10+.*** If you are in­clined to be­lieve the sur­vey re­sponses- or at least as­sign some level of cred­i­bil­ity to the re­sponses in the 90th-per­centile and be­low-, the data is much more con­sis­tent with a long-tail dis­tri­bu­tion for he­do­nic val­ues rel­a­tive to a nor­mal dis­tri­bu­tion.


Key Plea­sures Surfaced

Birth of children

I have heard a num­ber of moth­ers and father say that hav­ing kids was the best thing that ever hap­pened to them. The sur­vey showed this was a very strong pat­tern, es­pe­cially among women. In par­tic­u­lar, a lot of the re­ports deal with the very mo­ment in which they held their first baby in their arms for the first time. Some quotes to illus­trate this pat­tern:

The best ex­pe­rience of my life was when my first child was born. I was un­sure how I would feel or what to ex­pect, but the mo­ment I first heard her cry I fell in love with her in­stantly. I felt like sud­denly there was an­other per­son in this world that I cared about and loved more than my­self. I felt a sud­den urge to pro­tect her from all the bad in the world. When I first saw her face it was the most beau­tiful thing I had ever seen. It is al­most an in­de­scrib­able feel­ing. I felt like I un­der­stood the pur­pose and mean­ing of life at that mo­ment. I didn’t know it was pos­si­ble to feel the way I felt when I saw her. I was the hap­piest I have ever been in my en­tire life. That mo­ment is some­thing that I will cher­ish for­ever. The only other time I have ever felt that way was with the sub­se­quent births of my other two chil­dren. It was al­most a eu­phoric feel­ing. It was an in­tense calm and con­tent­ment.
I was young and had a difficult preg­nancy with my first born. I was scared be­cause they had to do an emer­gency c-sec­tion be­cause her health and mine were at risk. I had an­ti­ci­pated and thought about how the mo­ment would be when I fi­nally got to hold my first child and re­al­ize that I was a mother. It was un­be­liev­ably emo­tional and I don’t think any­thing in the world could top the amount of plea­sure and joy I had when I got to see and hold her for the first time.
I was 29 when my son was born. It was amaz­ing. I never thought I would be a father. Watch­ing him come into the world was eas­ily the best day of my life. I did not re­al­ize that I could love some­one or some­thing so much. It was at about 3am in the morn­ing so I was re­ally tired. But it was won­der­ful nonethe­less.
I ab­solutely loved when my child was born. It was a wave of emo­tions that I haven’t felt by any­thing be­fore. It was ex­cit­ing and scary and beau­tiful all in one.

No luck for anti-na­tal­ists… the su­per-strong drug-like effects of hav­ing chil­dren will pre­sum­ably con­tinue to mo­ti­vate most hu­mans to re­pro­duce no mat­ter how strong the eth­i­cal case against do­ing so may be. Com­ing soon: a drug that makes you feel like “you just had 10,000 chil­dren”.

Fal­ling in Love

The cat­e­gory of “fal­ling in love” was also a very com­mon top ex­pe­rience. I should note that the ex­pe­riences re­ported were not merely those of “hav­ing a crush”, but rather, they typ­i­cally in­volved un­usu­ally for­tu­nate cir­cum­stances. For in­stance, a woman re­ported be­ing friends with her crush for 7 years. She thought that he was not in­ter­ested in her, and so she never dared to con­fess her love for him… un­til one day, out of the blue, he con­fessed his love for her. Other ex­pe­riences of fal­ling in in­volve chance en­coun­ters with child­hood friends that led to movie-de­serv­ing ro­man­tic es­capades, for­bid­den love situ­a­tions, and cases where the per­son was con­vinced the lover was out of his or her league.


The terms “travel” and “va­ca­tion” may sound rel­a­tively frivolous in light of some of the other plea­sures listed. That said, these were not just any kind of travel or va­ca­tion. The ex­pe­riences de­scribed do seem rather ex­traor­di­nary and life-chang­ing. For ex­am­ple, talk­ing about back-pack­ing alone in France for a month, bik­ing across the US with your best friend, or a long trip in South East Asia with your sibling that goes much bet­ter than planned.


It is sig­nifi­cant that out of 97 peo­ple four of them listed MDMA as one of the most pleas­ant ex­pe­riences of their lives. This is salient given the rel­a­tively low base rate of us­age of this drug (some sur­veys say­ing about 12%, which is prob­a­bly not too far off from the base rate for Me­chan­i­cal Turk work­ers us­ing MDMA). This means that a high per­centage of peo­ple who have tried MDMA will rate it as as one of their top ex­pe­riences, thus im­ply­ing that this drug pro­duces ex­pe­riences sam­pled from an ab­surdly long-tailed high-valence dis­tri­bu­tion. This un­der­scores the civ­i­liza­tional sig­nifi­cance of in­vent­ing a method to ex­pe­rience MDMA-like states of con­scious­ness in a sus­tain­able fash­ion (cf. Cool­ing It Down To Par­ty­ing It Up).

Like­wise, the ap­pear­ance of LSD and psilo­cy­bin is sig­nifi­cant for the same rea­son. That said, mea­sures of the sig­nifi­cance of psychedelic ex­pe­riences in psychedelic stud­ies have shown that a high per­centage of those who ex­pe­rience such states rate them among their top most mean­ingful ex­pe­riences.

Games of Chance Earnings

Four par­ti­ci­pants men­tioned earn­ings in games of chance. Th­ese cases in­volved earn­ing amounts rang­ing from $2,000 all the way to a truck (which was im­me­di­ately sold for money). What I find sig­nifi­cant about this is that these ex­pe­riences are at times ranked above “col­lege grad­u­a­tion” and other clas­si­cally mean­ingful life mo­ments. This brings about a crazy util­i­tar­ian idea: if in­deed ed­u­ca­tion is as use­less as many peo­ple in the in­tel­lec­tual elite are say­ing these days (ex. The Case Against Ed­u­ca­tion) we might as well stop sub­si­diz­ing higher ed­u­ca­tion and in­stead make peo­ple par­ti­ci­pate in opt-out games of chance rigged in their fa­vor. Sub­sti­tute the Depart­ment of Ed­u­ca­tion for a Depart­ment of Lucky Mo­ments and give peo­ple mean­ingful life ex­pe­riences at a frac­tion of the cost.

Key Pains Surfaced

Kid­ney Stones and Migraines

The fact that these two med­i­cal is­sues were sur­faced is, I think, ex­tremely sig­nifi­cant. This is be­cause the life­time in­ci­dence of kid­ney stones is about 10% (~13% for men, 7% for women) and for mi­graines it is around 13% (9% for men, 18% for women). In the sur­vey we saw 993 peo­ple men­tion­ing kid­ney stones, and the same num­ber of peo­ple men­tion­ing mi­graines. In other words, there is rea­son to be­lieve that a large frac­tion of the peo­ple who have had ei­ther of these con­di­tions will rate them as one of their top 3 most painful ex­pe­riences. This fact alone un­der­scores the mas­sive util­i­tar­ian benefit that would come from be­ing able to re­duce the in­ci­dence of these two med­i­cal prob­lems (luck­ily, we have some good re­search leads for ad­dress­ing these prob­lems at a large scale and in a cost-effec­tive way: DMT for mi­graines, and fre­quency spe­cific microcur­rent for kid­ney stones)


Child­birth was men­tioned 16 times, mean­ing that roughly 30% of women rate it as one of their three most painful ex­pe­riences. While many peo­ple may look at this and sim­ply nod their heads while say­ing “well, that’s just life”, here at Qualia Com­put­ing we do not con­done that kind of defeatism and de­spi­ca­ble lack of com­pas­sion. As it turns out, there are fas­ci­nat­ing re­search leads to ad­dress the pain of child­birth. In par­tic­u­lar, Jo Cameron, a 70 year old ve­gan schoolteacher, de­scribed her child­birth by say­ing that it “felt like a tickle”. She hap­pens to have a mu­ta­tion in the FAAH gene, which is usu­ally in charge of break­ing down anan­damine (a neu­ro­trans­mit­ter im­pli­cated in pain sen­si­tivity and he­do­nic tone). As we’ve ar­gued be­fore, ev­ery child is a com­plete ge­netic ex­per­i­ment. In the fu­ture, we may as well try to at least make ed­u­cated guesses about our chil­dren’s genes as­so­ci­ated with low mood, anx­iety, and pain sen­si­tivity. In defi­ance of com­mon sense (and the Bible) the fu­ture of child­birth could in­deed be one de­void of in­tense pain.

Car accidents

Car ac­ci­dents are ex­tremely com­mon (the base rate is so high that by the age of 40 or so we can al­most as­sume that most peo­ple have been in at least one car ac­ci­dent, pos­si­bly mul­ti­ple). More so, it seems likely that the health-dam­ag­ing effects of car ac­ci­dents, by their na­ture, fol­low a long-tail dis­tri­bu­tion. The high base rate of peo­ple men­tion­ing car ac­ci­dents in their top 3 most painful ex­pe­riences un­der­scores the im­por­tance of stream­lin­ing the pro­cess of tran­si­tion­ing into the era of self-driv­ing cars.

Death of Father and Mother

This one does not come as a sur­prise, but what may stand out is the rel­a­tively higher fre­quency of men­tions of “death of father” rel­a­tive to “death of mother”. I think this is an ar­ti­fact of the longevity differ­ence be­tween men and women. This is in agree­ment with the ob­served effect of age: about 15% vs. 25% of peo­ple un­der and over 40 had men­tioned the death of their father, as op­posed to a differ­ence of 5% vs. 25% for death of mother. The rea­son why the father might be over-rep­re­sented might sim­ply be due to the lower life ex­pec­tancy of men rel­a­tive to women, and hence the father, on av­er­age, dy­ing ear­lier. Thus, it be­ing re­ported more fre­quently by a younger pop­u­la­tion.

Fu­ture Direc­tions for Method­olog­i­cal Ap­proaches:

Graph­i­cal Models with Log-Nor­mal Priors

After try­ing so many an­a­lytic an­gles on this dataset, what else is there to do? I think that as a proof of con­cept the anal­y­sis pre­sented here is pretty well-rounded. If the Qualia Re­search In­sti­tute does well in the fund­ing de­part­ment, we can ex­pect to ex­tend this pi­lot study into a more com­pre­hen­sive anal­y­sis of the plea­sure-pain axis both in the gen­eral pop­u­la­tion and among pop­u­la­tions who we know have en­dured or en­joyed ex­tremes of valence (such as cluster headache suffer­ers or peo­ple who have tried 5-MeO-DMT).

In terms of statis­ti­cal mod­els, an ad­e­quate amount of data would en­able us to start us­ing prob­a­bil­is­tic graph­i­cal mod­els to de­ter­mine the most likely long-tail dis­tri­bu­tions for all of the key pa­ram­e­ters of plea­sure and pain. For in­stance, we might want to de­velop a model similar to Item Re­sponse The­ory where:

  1. Each par­ti­ci­pant sam­ples ex­pe­riences from a dis­tri­bu­tion.

  2. Each ex­pe­rience cat­e­gory gen­er­ates sam­ples with an em­piri­cally-de­ter­mined base rate prob­a­bil­ity (e.g. chances that it hap­pens in a given year), along with a la­tent he­do­nic value dis­tri­bu­tion.

  3. A “dis­crim­i­na­tion func­tion” f(a, b) that gives the prob­a­bil­ity that ex­pe­rience of he­do­nic value a is rated as more pleas­ant (or painful) rel­a­tive an ex­pe­rience with a he­do­nic value of b.

  4. And a gen­er­a­tive model that es­ti­mates the like­li­hood of ob­serv­ing ex­pe­riences as the top 3 (or top x) based on the pa­ram­e­ters pro­vided.

In brief, with an ap­proach like the above we can po­ten­tially test the model fit for differ­ent dis­tri­bu­tion types of he­do­nic val­ues per ex­pe­rience. In par­tic­u­lar, we would be able to de­ter­mine if the model fit is bet­ter if the ex­pe­riences are drawn from a Gaus­sian vs. a log-nor­mal (or other long-tailed) dis­tri­bu­tion.

Fi­nally, it might be fruit­ful to ex­plic­itly ask about whether par­ti­ci­pants have had cer­tain ex­pe­riences in or­der to cal­ibrate their rat­ings, or even have them try a bat­tery of stan­dard­ized pain/​plea­sure-in­duc­ing stim­uli (cap­saicin ex­tract, elec­troshocks, stings, mas­sage, or­gasm, etc.). We could also find the way to com­bine (a) the nu­mer­i­cal rat­ings, (2) the rank­ing in­for­ma­tion, and (3) the “how many times bet­ter/​worse” re­sponses into a sin­gle model. And for best re­sults, re­strict the anal­y­sis to very re­cent ex­pe­riences in or­der to re­duce re­call bi­ases.

Clos­ing Thoughts on the Valence Scale

To sum­ma­rize, I be­lieve that the case for a long-tail ac­count of the plea­sure-pain axis is very defen­si­ble. This pic­ture is sup­ported by:

  1. The long-tailed na­ture of neu­ronal cas­cades,

  2. The phe­nomenolog­i­cal ac­counts of in­tense plea­sure and pain (w/​ phe­nomenolog­i­cal ac­counts of time and space ex­pan­sion),

  3. The way in which pain scales are con­structed by those who de­vel­oped them, and

  4. The an­a­lytic re­sults of the pi­lot study we con­ducted and pre­sented here.

In turn, these re­sults give rise to a new in­ter­pre­ta­tion of psy­chophys­i­cal ob­ser­va­tions such as We­ber’s Law. Namely, that Just No­tice­able Differ­ences may cor­re­spond to ge­o­met­ric differ­ences in qualia, not only in sen­sory stim­uli. That is, that the ex­po­nen­tial na­ture of many cases where We­ber’s Law ap­pears are not merely the re­sult of a log­a­r­ith­mic com­pres­sion on the pat­terns of stim­u­la­tion at the “sur­face” of our sense or­gans. Rather, the ob­ser­va­tions pre­sented here sug­gest that these long-tails deal di­rectly with the­qual­ity and in­ten­sity of con­scious ex­pe­rience it­self.

Ad­di­tional Material

Di­men­sion­al­ity of Plea­sure and Pain

Pain and plea­sure may have an in­trin­sic “di­men­sion­al­ity”. Without elab­o­rat­ing, we will merely state that a gen­er­a­tive defi­ni­tion for the “di­men­sion­al­ity of an ex­pe­rience” is the high­est “vir­tual di­men­sion” im­plied by the pat­terns of cor­re­la­tion be­tween de­grees of free­dom. The hot pep­per hands ac­count I re­lated sug­gested a kind of di­men­sional phase tran­si­tion be­tween 410 and 510 pain, where the pat­terns of a cer­tain type (4/​10 “sparks” of pain) would some­times syn­chro­nize and gen­er­ate a new type of higher-di­men­sional sen­sa­tion (5/​10 “soli­tons” of pain). To illus­trate this idea fur­ther:

First, in Hot Ones, Ku­mail Nan­ji­ani de­scribes sev­eral “leaps” in the spici­ness of the wings, first at around 30,000 Scov­ille (“this new ghost that ap­pears and only here starts to visit you”), and sec­ond at around 130k Scov­ille (para­phras­ing: “like how NES to Su­per Nin­tendo felt like a big jump, but then Su­per Nin­tendo to N64 was an even big­ger leap” – “Now we are play­ing in the big leagues moth­er­fucker! This is fuck­ing real!”). This hints at a change in di­men­sion­al­ity, too.

And sec­ond, Shinzen Young‘s ad­vice about deal­ing with pain in­volves not re­sist­ing it. He dis­cusses how suffer­ing is gen­er­ated by the co­or­di­na­tion be­tween emo­tional, cog­ni­tive, and phys­i­cal men­tal for­ma­tions. If you can keep each of these men­tal for­ma­tions hap­pen­ing in­de­pen­dently and don’t al­low their co­or­di­nated forms, you will avoid some of what makes the ex­pe­rience bad. This also sug­gests that higher-di­men­sional pain is qual­i­ta­tively worse. Prag­mat­i­cally, train­ing to do this may make sense for the time be­ing, since we are still some years away from sus­tain­able pain-re­lief for ev­ery­one.

Mixed States

We have yet to dis­cuss in de­tail how mixed states come into play for a log-nor­mal valence scale. The Sym­me­try The­ory of Valence would sug­gest that most states are neu­tral in na­ture and that only pro­cesses that re­duce en­tropy lo­cally such as neu­ral an­neal­ing would pro­duce highly-valenced states. In par­tic­u­lar, we would see that high-valence states have very nega­tive valence states nearby in con­figu­ra­tion space; if you take a very good high-en­ergy state and dis­tort it in a ran­dom di­rec­tion it will likely feel very un­pleas­ant. The points in be­tween would be mixed valence, which ac­count for the ma­jor­ity of ex­pe­riences in the wild.

Qualia Formalism

Qualia For­mal­ism posits that for any given sys­tem that sus­tains ex­pe­riences, there is a math­e­mat­i­cal ob­ject such that the math­e­mat­i­cal fea­tures of that ob­ject are iso­mor­phic to the sys­tem’s phe­nomenol­ogy. In turn, Valence Struc­tural­ism posits that the he­do­nic na­ture of ex­pe­rience is en­coded in a math­e­mat­i­cal fea­ture of this ob­ject. It is eas­ier to find some­thing real if you posit that it ex­ists (rather than try to ex­plain it away). We have sug­gested in the past that valence can be ex­plained in terms of the math­e­mat­i­cal prop­erty of sym­me­try, which cashes out in the form of neu­ral dis­so­nance and con­so­nance.

In con­trast to elimi­na­tivist, illu­sion­ist, and non-for­mal ap­proaches to con­scious­ness, at QRIwe sim­ply start by as­sum­ing that ex­pe­rience has a deep ground truth struc­ture and we see where we can go from there. Although we cur­rently lack the con­cep­tual schemes, sci­ence, and vo­cab­u­lary needed to talk in pre­cise terms about differ­ent de­grees of plea­sure and pain (though we are try­ing!), that is not a good rea­son to dis­miss the first-per­son claims and in­di­rect pieces of ev­i­dence con­cern­ing the true amounts of var­i­ous kinds of qualia bound in each mo­ment of ex­pe­rience. If valence does turn out to in­trin­si­cally be a math­e­mat­i­cal fea­ture of our ex­pe­rience, then both its qual­ity and quan­tity could very well be pre­cisely mea­surable, con­cep­tu­ally crisp, and tractable. A sci­en­tific fact that, if proven, would cer­tainly have im­por­tant im­pli­ca­tions in ethics and meta-ethics.


* It’s a shame that Coy­ote Peter­son didn’t rate the pain pro­duced by the var­i­ous wings he ate on the Hot Ones show rel­a­tive to in­sect stings, but that sort of data would be very helpful in es­tab­lish­ing a uni­ver­sal valence scale. More gen­er­ally, stunt-man per­son­al­ities like the L.A. Beast who sub­ject them­selves to ex­tremes of nega­tive valence for In­ter­net points might be an un­tapped gold mine for ex­pe­rience defer­ence data (e.g. How does eat­ing the most bit­ter sub­stance known com­pare with the bul­let ant glove? Ask­ing this guy might be the only way to find out, with­out cre­at­ing more ca­su­alties).

**Base rate of men­tions of worst ex­pe­riences:

[(‘Father death’, 19), (‘Child­birth’, 16), (‘Grand­mother death’, 13), (‘Mother death’, 11), (‘Car ac­ci­dent’, 9), (‘Kid­ney stone’, 9), (‘Mi­graine’, 9), (‘Ro­man­tic breakup’, 9), (‘Bro­ken arm’, 6), (‘Bro­ken leg’, 6), (‘Work failure’, 6), (‘Divorce’, 5), (‘Pet death’, 5), (‘Bro­ken foot’, 4), (‘Bro­ken an­kle’, 4), (‘Bro­ken hand’, 4), (‘Un­speci­fied’, 4), (‘Friend death’, 4), (‘Sister death’, 4), (‘Skin burns’, 3), (‘Skin cut need­ing stitches’, 3), (‘Fi­nan­cial ruin’, 3), (‘Prop­erty loss’, 3), (‘Sprained an­kle’, 3), (‘Gal­l­s­tones’, 3), (‘Fam­ily breakup’, 3), (‘Divorce of par­ents’, 3), (‘C-sec­tion re­cov­ery’, 3), (‘Love failure’, 2), (‘Bro­ken finger’, 2), (‘Un­speci­fied frac­ture’, 2), (‘Bro­ken ribs’, 2), (‘Un­speci­fied fam­ily death’, 2), (‘Bro­ken col­lar­bone’, 2), (‘Grand­father death’, 2), (‘Un­speci­fied ill­ness’, 2), (‘Pe­riod pain’, 2), (‘Be­ing cheated’, 2), (‘Fi­nan­cial loss’, 2), (‘Bro­ken tooth’, 2), (‘Cousin death’, 2), (‘Rel­a­tive with can­cer’, 2), (‘Cluster headache’, 2), (‘Un­speci­fied leg prob­lem’, 2), (‘Root canal’, 2), (‘Back pain’, 2), (‘Bro­ken nose’, 2), (‘Aunt death’, 2), (‘Wis­dom teeth’, 2), (‘Cancer (eye)’, 1), (‘Ap­pendix op­er­a­tion’, 1), (‘Dis­lo­cated elbow’, 1), (‘Con­cus­sion’, 1), (‘Mono’, 1), (‘Sex­ual as­sault’, 1), (‘Kid­ney in­fec­tion’, 1), (‘He­m­or­rhoids’, 1), (‘Tat­too’, 1), (‘Un­speci­fied kid­ney prob­lem’, 1), (‘Un­speci­fied lung prob­lem’, 1), (‘Un­speci­fied can­cer’, 1), (‘Un­speci­fied child­hood sick­ness’, 1), (‘Bro­ken jaw’, 1), (‘Bro­ken elbow’, 1), (‘Thrown out back’, 1), (‘Lost sen­ti­men­tal item’, 1), (‘Abor­tion’, 1), (‘Rup­tured kid­ney’, 1), (‘Big fall’, 1), (‘Torn knee’, 1), (‘Finger hit by ham­mer’, 1), (‘In­jured thumb’, 1), (‘Brother in law death’, 1), (‘Knocked teeth’, 1), (‘Un­speci­fied death’, 1), (‘Rip­ping off finger­nail’, 1), (‘Per­sonal anger’, 1), (‘Wrist pain’, 1), (‘Get­ting the wind knocked out’, 1), (‘Blown knee’, 1), (‘Burst ap­pendix’, 1), (‘Tooth ab­scess’, 1), (‘Ten­dini­tis’, 1), (‘Altru­is­tic frus­tra­tion’, 1), (‘Leg op­er­a­tion’, 1), (‘Gal­lblad­der in­fec­tion’, 1), (‘Bro­ken wrist’, 1), (‘Stomach flu’, 1), (‘Run­ning away from fam­ily’, 1), (‘Child beat­ing’, 1), (‘Sinus in­fec­tion’, 1), (‘Bro­ken thumb’, 1), (‘Fam­ily abuse’, 1), (‘Mis­car­riage’, 1), (‘Tooth ex­trac­tion’, 1), (‘Feel­ing like your soul is lost’, 1), (‘Home­less­ness’, 1), (‘Los­ing your re­li­gion’, 1), (‘Los­ing bike’, 1), (‘Fam­ily mem­ber in prison’, 1), (‘Crohn s dis­ease’, 1), (‘Ir­ri­ta­ble bowel syn­drome’, 1), (‘Fam­ily in­jured’, 1), (‘Un­speci­fied chronic dis­ease’, 1), (‘Fibromyal­gia’, 1), (‘Blood clot in toe’, 1), (‘In­fected c-sec­tion’, 1), (‘Suicide of lover’, 1), (‘Den­tal ex­trac­tion’, 1), (‘Un­speci­fied part­ner abuse’, 1), (‘In­fer­til­ity’, 1), (‘Father in law death’, 1), (‘Bro­ken neck’, 1), (‘Scratched cornea’, 1), (‘Swol­len lymph nodes’, 1), (‘Sun burns’, 1), (‘Tooth ache’, 1), (‘Lost cus­tody of chil­dren’, 1), (‘Un­speci­fied ac­ci­dent’, 1), (‘Bike ac­ci­dent’, 1), (‘Bro­ken hip’, 1), (‘Not be­ing loved by part­ner’, 1), (‘Dog bite’, 1), (‘Bro­ken skull’, 1)]

Base rate of men­tions of best ex­pe­riences:

[(‘Fal­ling in love’, 42), (‘Chil­dren born’, 41), (‘Mar­riage’, 21), (‘Sex’, 19), (‘Col­lege grad­u­a­tion’, 13), (‘Or­gasm’, 11), (‘Al­co­hol’, 8), (‘Va­ca­tion’, 6), (‘Get­ting job’, 6), (‘Per­sonal fa­vorite sports win’, 6), (‘Na­ture scene’, 5), (‘Own­ing home’, 5), (‘Sports win’, 4), (‘Grad­u­at­ing high­school’, 4), (‘MDMA’, 4), (‘Get­ting paid for the first time’, 4), (‘Amuse­ment park’, 4), (‘Game of chance earn­ing’, 4), (‘Job achieve­ment’, 4), (‘Get­ting en­gaged’, 4), (‘Cannabis’, 3), (‘Eat­ing fa­vorite food’, 3), (‘Un­ex­pected gift’, 3), (‘Mov­ing to a bet­ter lo­ca­tion’, 3), (‘Travel’, 3), (‘Divorce’, 2), (‘Gift­ing car’, 2), (‘Giv­ing to char­ity’, 2), (‘LSD’, 2), (‘Won con­test’, 2), (‘Friend re­union’, 2), (‘Win­ning bike’, 2), (‘Kiss’, 2), (‘Pet own­er­ship’, 2), (‘Chil­dren’, 1), (‘First air trip’, 1), (‘First kiss’, 1), (‘Public perfor­mance’, 1), (‘Hugs’, 1), (‘Un­speci­fied’, 1), (‘Re­cov­er­ing from un­speci­fied kid­ney prob­lem’, 1), (‘Col­lege party’, 1), (‘Grad­u­ate school start’, 1), (‘Fi­nan­cial suc­cess’, 1), (‘Din­ner with loved one’, 1), (‘Feel­ing sup­ported’, 1), (‘Chil­dren grad­u­ates from col­lege’, 1), (‘Fam­ily event’, 1), (‘Par­ti­ci­pat­ing in TV show’, 1), (‘Psychedelic mush­rooms’, 1), (‘Opi­ates’, 1), (‘Hav­ing own place’, 1), (‘Mak­ing mu­sic’, 1), (‘Be­com­ing en­gaged’, 1), (‘Theater’, 1), (‘Ex­treme sport’, 1), (‘Armed forces grad­u­a­tion’, 1), (‘Birth­day’, 1), (‘Pos­i­tive preg­nancy test’, 1), (‘Feel­ing that God ex­ists’, 1), (‘Belief that Hell does not ex­ist’, 1), (‘Get­ting car’, 1), (‘Aca­demic achieve­ment’, 1), (‘Helping oth­ers’, 1), (‘Meet­ing soul­mate’, 1), (‘Daugh­ter back home’, 1), (‘Win­ning cus­tody of chil­dren’, 1), (‘Friend stops drink­ing’, 1), (‘Mas­tur­ba­tion’, 1), (‘Friend not dead af­ter all’, 1), (‘Child learns to walk’, 1), (‘At­tend­ing wed­ding of loved one’, 1), (‘Chil­dren safe af­ter dan­ger­ous situ­a­tion’, 1), (‘Un­speci­fied good news’, 1), (‘Met per­sonal idol’, 1), (‘Child learns to talk’, 1), (‘Chil­dren good at school’, 1)]

For clar­ity – “Per­sonal fa­vorite sports win” means that the re­spon­dent was a par­ti­ci­pant in the sport as op­posed to a spec­ta­tor (which was la­beled as “Sports win”). The differ­ence be­tween “Sex” and “Or­gasm” is that Sex refers to the en­tire act in­clud­ing fore­play and cud­dles whereas Or­gasm refers to the spe­cific mo­ment of cli­max. For some rea­son peo­ple would ei­ther men­tion one or the other, and em­pha­size very differ­ent as­pects of the ex­pe­rience (e.g. in­ti­macy vs. phys­i­cal sen­sa­tion) so I de­cided to la­bel them differ­ently.

*** It is pos­si­ble that some fine-tun­ing of pa­ram­e­ters could give rise to long-tail ra­tios even with a nor­mal dis­tri­bu­tion (es­pe­cially if the mean is, say, a nega­tive value and the stan­dard de­vi­a­tion is very wide). But in the gen­eral case a nor­mal dis­tri­bu­tion will have a fairly nar­row range for the ra­tios of the “top value di­vided by the sec­ond top value”. So at least as a gen­eral qual­i­ta­tive ar­gu­ment, I think, the simu­la­tions do sug­gest a long-tailed na­ture for the re­ported he­do­nic val­ues.