Now, there are other themes arising from our blindness to the Blaek. Swan:

Transcription

1 ! 50 UMBERTO ECO'S ANflLlBllARY Swan: Now thee ae othe themes aising fom ou blindness to the Blaek a. We foeus on peseleeted segments of the seen and genealize fom it to the unseen: thc eo of confimation. b. We fool ouselves with stoies that cate to ou Platonic thist fo distinct pattens: the naative fallacy. e. We behave as if the Blaek Swan does not exist: human natue is not pogammed fo Blaek Swans. d. What we see is not neeessaily al! that is thee. Histoy hides Black Swans fom us and gives us a mistaken idea about the odds of thcse events: this is the distotion of silent evidenee. e. We "tunnel": that is we foeus on a few well-defined souces of uncetainty on too specifie a list of Blaek Swans (at the expense of the othes that do not easily come to mind). Cha pe Flve CONFRMATON SHMONFRMATON! 1 will diseuss eaeh of the points in the next five ehaptes. Then in the eonclusion of Pat One 1will show how in effeet they ae the sane topic. have so much evidence-can Zoogles be (sometimes) Boogles?- Cooboation shmooboation-poppe's idea As much as it is ingained in ou habits and conventional wisdom confimation can be a dangeous eo. Assume 1 told you that 1 had evidence that the football playe O. J. Simpson (who was accused of killing his wife in the 1990s) was not a ciminal. Look the othe day 1 had beakfast with him and he didn't kill _ ybody. 1am seious 1did not see him kill a single pesono Wouldn't that ;' eonfim his innocenee? f 1 said such a thing you. would cetainly call a 'hink an ambulance o pehaps even the police sinee you might think.. ~.hat 1 spent too much time in tading ooms o in cafés thinking about this lack Swan topic and that my logic may epesent such an immediate ~ e dllnge to society that 1 myself need to be loeked up immediately. ;('i You would have the same eaction if 1 told you that 1 took a nap the Othc day on [he aiload tae k in New Rochelle New Yok and was not klllcd. Hey look at me 1 am alive 1 would say and that is evidence that lylng on tain tacks is isk-fee. Yet eonside the following. Look again at t'.l~lle 1 in Chapte 4; someone who obseved the tukey's fist thousand ~.Ys(but not the shock of the thousand and fist) would tell you and [jlllhtly so that thee is no evidence of the possibility of lage events i.e.

2 UMBERTO ECO'S ANTlLlBRAY Black Swans. You ae likely to confuso hat statencn lady if you do not pay close atention evidence of no possible Black Swans. Even though logical distanee mind beween the two assetions now if you manage to emembe the seeond Black Suians. call his eonfusion mens ae not intecbangeable. Such confusion (but cucial) os no ae pofessos plicated equations! all Moslems is tue that 99 peeent thousand. oveestimate to tivial immune Assume mistake lege. Yet anothe elevato This inability simplify the fom the same confusion: it. ae Moslems" mean that since thee ae moe " neve mean teoists one in a Moslem Mili once complained. ound-tip peson fallacy the unfainess of aeas in the Unied States have suffeed people ae geneally is chonic: of the object with ou bu they still and sophisti- 1 depend on what evolutionay psycholo- he evento The c1assoom John Stuat if you tell people that the Logieal poblems diffeently ap- in daily even when it is exact does not often lead to appopiate if we do not pay attention plied pofessos of statistics t tistical questions even when we ae expets. tend to leave thei bains engage in the mos tivial infeential i it and how suounds system. is a do- not on its acions beca use we tend to foget what we know o foget how o pocess :1 "'.. is a quite By domain-specific ou intuitions poached one way in the c1assoom might be teated l fe. ndeed hey ae teated diffeently in daily life. " teets. n 1971 the psychologiss stupid. O social-emotional.~ cians i has been shown know bette. Consevative" ou mode of thinking in which the mate is pesented egistes -:~ t popely come fon hei ethnic ae not ciminals by people who should This poblem knowledge logical meit bu on he basis of which famewok i by close to fifty to say that the Conseva ti ves ae geneally meant to say that stupid si de of town steps into the tansfe main; eal life is anothe We eac to a piece of infomation you (unconsciously) individual even if nost ciminals suffe fom díscimination fom the wong to autonaically mean tha ou eactions giss cal! the "domain" makes dawn most of hei ethnic subgoup on thei SAT test might not make i to eol- cation fom one situation to anothe o fom theoy to pactiee distubing attibute of human natue. Knowledge in uban have peson can get vey high scoes on the SATs and still feel a chill of fea when someone This would times! subgoup sweing this kind of question the context thousand steeotypes-minoities mtuinons can make a big diffe- Ae Nof A/ 800g195 that the fis statenent the age of say fifteen and fifty) being a teoist see in this Ou staistical Let us call it the domain speci{icity of ou eactions. (between might mconsequenial in which these subtleties Al! zoogles ae boogles. You saw a boogle. 1s it a zoogle? Not necessaily since not al! boogles ae zoogles adolescents who make a mistake in an- to them (com- all eoists ae teoists the odds of a andomly The eade Zoogle5 logical e- ou knowing and only say ten thousand So the logical almost 110 wih claity of mind). ae Moslems. of Moslems than one billion Moslems hunded in you oí a tivial vey tivial happily "almost ae teoists." 15 not evolved fo a habitat ence. hee. of teoists.001 pecent only about patakes do so without Many people confuse the statement with "alnost a all you will be vey had we ae likely to unwitingly t is woth a deepe examination eo hee but it fallacy since these sta te- paticulaly beca use ou minds outinely 53 killes ae toild animals and most wtld animals ae killes. Thee is an faet vast the thee is poof of we ae not immune do not tend to cohabit Unless we concentate poblem the ound-tip and thinkes SHMONFRMATON fo the othe Ten days fom vesion-that of the two statements logical eo-but it isin the fis statement inaccuate paticu- that thee is will seem vey naow so that one can be easily substituted likely to etain howeve with the statenent CONFRMAON eos Danny with statistical Statisti- in the classoom and once they ae et out on the Kahneman quesions One was simila to the following and Amos Tvesky not phased (changing as sta- the example fu c1aity): Assume that you live in a town with two hospítals-one lage key to success is not always skills they think that you ae telling them that thc ohe smal!. On a given day 60 pecent of those bon in one of the two it is neve ski lis always luck. hospitals Ou infeential machiney that which we use in daily life is not made fo a complicated envionment when its woding is slightly onment in which modified. thee is no consequential a statement Conside diffeence changes makedly that in a pimitive between the staenents envi- most ae boys. Which mude the equivalent of the mistake choosing the lage hospital i- lage hospital is it likely to be? Many (duing a casual 50 pecent convesaion) when in fact the vey basis of statistics samples ae moe stable and should f!uctuate Vt'age-hee statisticians of is that less fom the long-tem fo each of the sexes-han smalle sampies.

3 l' 1 54 UMBERTO ECO'S ANTlLlBRARY These statisticians would have flunkcd thei own exams. Duing my days as a quant counted hundeds of such sevee infeemial mistakes nade by staisticians who fogot that hey wee satisticians Fo anothe illustation oí the way we can be ludicously doma inspecific in daily life go to the luxuy Reebok Spots Club in Ncw Yok City and look at the numbe of people who afte iding the escalato fo a couple of floos head diecly to the StaiMastes. This domain specificity of ou infeences and eactions woks both ways: sone poblems we can undestand in thei applications but not in textbooks; othes we ae bette at captuing in the textbook than in the pactical application. People can manage o effotlessly solve a poblem in a social situation but suggle when it is pesened as an abstact logical poblcm. We tend to use diffeent menal machiney=-so-called modulesin diffeent situations: ou bain lacks a cental all-pupose conputc that stats with logical ules and applies them equally to all possible situaions. And as 've said we can commit a logical mistake in ealtty but not in the classoom This asymnety is best visible in cance detection. Take doctos examining a patient fo signs of cance; tests ae typically done on patients who want o know if they ae cued o if thee is "ecuencc." (n fact ecuence is a misnone; it simply means tha the teatment did not kill al! the canceous cells and that these undetected malignant cells have stated to multiply out of conto!.) t is not feasible in the pesent state of technology to examine evey single one oí the patient's cells to see if al! of then ae nonnalignant so the docto takes a sample by scanning the body with as much pecision as possible. Then she makes an assunption about what she did no see. 1 was once taken aback when a docto told me afte a outine cance eheckup "Stop woying we have evidence of cue." "Why?" 1 asked. "Thee is evidcnce of no cance" was the eply "How do you know?" 1 asked. He eplied "Thc sean is negative." Ye he went aound calling himself docto! An aconym used in the medical liteatue is NED which stands fo No Evidence of Disease. Thee is no such thing as END Evidenee of No Disease. Yet my expeience discussing his matte with plenty of doctos even those who publish papes on thei esults is that many slip into the ound-tip fallacy duing convesation. Doctos in the midst of the scientific aogance of the 1960s looked down at mothes' milk as something pimitive as if it could be eplicated hy thei laboatoies-not eajizing that mothes' milk migh indude usef : l- CONFRMATON SHMONFRMATON! 55 ~:' fui components that could have eluded thei scientific undesanding=-a. simple confusion of absence of eutdence of the benefis of mothes' milk with evidence of absence of the benefis (anothe case of Platonicity as "it did not make sense" to beas-feed when we could simply use bottles). Many people paid the pice fo his naive infeence: those who wee not beast-fed as infants tuned out to he a an inceased isk of a coljection of health poblens including a highe likelihood of developing cetain types of cance-thee had to be in mothes' milk some necessay nutiens that still elude us. Futhemoe benefits to nohes who beast-feed wee also neglected such as a eduction in he isk of beast cánce Likewise with tonsils: the emoval of onsils may lead to a highe incidence of thoat cance but fo decades doctos neve suspcced tha this "useless" tissue might actually have a use that escaped thei detection. The sane with the dieay fibe found in fuits and vegeables: doctos in the 1960s found it useless beca use they saw no immediate evidenee of its necessity and so they ceated a malnouished geneation. Fibe i tuns out acts to slow down the absoption of sugas in the blood and scapes the intestinal taet of pecanceous cells. lndecd medicine has caused pleny of damage houghout histoy owing to this simple kind of infeential confusiono 1 am not saying hee that doctos should not have beliefs only that some kinds of definitive cjoscd beliefs need to be avoided-his is what Menodotus and his school seemed to be advocating with thei band of skeptical-empiical medicine that avoided theoizing. Medicine has gotten bette-bu many kinds of knowledge have 110t. 4; ). Evidence tby a mental mechanisn 1 call naive empiicism we have a natual ten- '. deney to look fo instances that confin ou stoy and ou vision of the wold+-these instances ae always easy to find. Alas with tools and fools '' anything can be easy to find. You take past instances tha cooboate you theoies and you teat then as euidence. Fo instance a diplomat will show you his "accomplishnens" not what he failed to do. Mathematieians will ty to convince you that thei science is useful o society by pointing out instances whee i po ved helpful no those whee i was a waste of time o wose those numeous nahenaical applications that inflicted asevee cos on society owing to the highly unenpiical natue of clegant mahenatical heoies.

4 !.. 56 UMBERTO ECO'S AN1LlBRARY Even in testing a hypothesis we tend to look fo instances whee the hypothesis poved tue. Of couse we can easily find confimation; al! we have to do is look o have a eseache do it fo US 1 can find confimation fo just about anything the way a skilled London cabbie can find taffic to incease the fae even on a holiday Sone people go futhe and give me exanples of events that we have been able to foesee with some success-indeed thee ae a few like landing aman on the moon and the economic gowth of the twenty-fist centuy. One can find plenty of "couneevidence" to the points in this book the best being that newspapes ae excellent at pedicting movie and theate schedules. Look 1 pedicted yesteday that the sun would ise today and it did! NEGATVE EMPRCSM ;i The good news is that thee is a way aound this naive empiicism. 1 am saying that a seies of cooboative facts is not necessaily evidence. Seeing white swans does not confim the nonexistence of black swans. Thee is an exception howeve: 1know what statenent is wong bu not necessaily what staenen is coecto f 1 see a black swan 1 can cetify that all swans ae not white! f 1 see someone kill then 1can be pactically cetain hat he is a ciminal. f don't see him kill 1cannot be cetain that he is innocent. The sane applies to cance detection: the finding ot a single malignant tumo poves that you have cance but the absence of such a finding cannot allow you to say with cetainty that you ae cance-fee. We can get close to the tuth by negative instances not by veification! is misleading to build a geneal ule fom obseved facts. Contay to conventional wisdon ou body of knowledge does not incease fom a seies of confimatoy obsevations like the tukey's But thee ae some things 1 can ema in skeptical about and othes 1 can safely conside cetain. This makes the consequences of obsevations one-sided. t is not much moe difficult than that This asymmety is immensely pactica]. t te lis us that we do not have to be complete skeptics just semiskeptics. The subtlety of eal life ove the books is that in you decision making you need be inteested only in one side of the stoy: if you seek cetainty about whethe the patient has cance not cetainty about whethe he is healthy then you might be satisfied with negative infeence since it wil! supply you the cetainty you seek. So CONF<MATON ':" we can 1" a lo fon d"-b" no' as much as SHMONFRMATON! w expcct. 57 Sometimes a lot of data can be meaningless; a othe times one single piece of infomaion can be vey meaningful. is tue hat a thousand days cannot pove you ight but one day can pove you to be wong The peson who is cedited with the pomotion of tbis idea of one-sided semiskepticism is Si Dokto Pofesso Kal Raimund Poppe who may be the only philosophe of science who is actually ead and discussed by actos in the eal wold (though not as enhusiastically by pofessional philosophes). As 1am wiing these lines a black-and-white pictue of him is hanging on the wal! of my sudy t was a gift 1 got in Munich fom the essayist Jochen Wegne who like me consides Poppe to be about al! "we've got" among moden thinkes=-well almost. He wites to us not to othe philosophes. "We" ae the empiical decision makes who hold that uncetainty is ou discipline and that undestanding how to act unde conditions of incomplete infomation is the highest and most ugent human pusuit. Poppe geneated a lage-scale theoy aound this asynmety based on a technique called "falsification" (o falsify is o po ve wong) meant to distinguish between science and nonscience and people immediaely stated splitting hais about its technicalities even though it is not the most inteesting o the nost oiginal of Poppe's ideas. This idea about the asynnety of knowledge is so iked by pactitiones because it is obvious to then; it is the way they un thei business The pbilosophe maudit Chales Sandes Peice who like an atist got only posthumous espect also came up with a vesion of tbis Black Swan solution wben Poppe was weaing diapes-some people even callecl it the Peice-Poppe appoach. Poppe's fa moe poweful and oiginal idea is the "open " society one that elies on skepticism as a modus opeandi efusing and esisting definitive tuths. He accused Plato of closing ou minds accoding to the agumens 1 descibed in he Pologue. But Poppe's biggest idea was his insight concening the fundamental sevee and incuable unpedictability of the wold and that 1willleave fo the chapte on pediction.. Of couse it is no so easy o "falsify" i.e. to state that something is wong with ful! cetainty mpefections in you testing method may yield a mistaken "no" The docto discoveing cancel' cells might have faulty Neihe Peice no philosophe Vico 1878 as if i wee business-ancients pises. Poppe was the fist to come up wih this asynmey. The Bochad mentioned the impotance of negaive empiicism in a matte held by the enpiicists to be the sound way to do undesood it inplicitly Out-of-pin books delive many su-

5 ':-1 - i.;:"!"i~ j i - 58 UMBERTO ECO'S ANTLlBRARY equal in impotance. Poppe intoduced the mechanism of conjectues and efutations which woks as follows: you fomulate a (bold) conjectue and you stat looking fo the obsevation that would peve you wong. This is the altenative to ou seach fo confimatoy instances. f you think the task is easy you wil! be disappointed-few humans have a natual ability to do this. 1confess that 1 am not one of them; it does not eome natually to me. * Counting to ttvee \ 1 1 Cognitive scientists have studied ou natual tendency to look only fo cooboation; they caj\ this vulneability to the cooboation eo the confimatían bias. Thee ae some expeiments showing that people focus only on the books ead in Umbeto Eco's libay. You can test a given ule eithe diectly by looking at instances whee it woks o indiectly by focusing on whee it does not wok. As we saw ealie disconfiming instances ae fa moe poweful in establishing tuth. Yet we tend to not be awae of this popety. The fist expeiment 1 know of concening this phenomenon was done by the psychologist P. C. Wason. He pesented subjects with the theenumbe sequence and asked them to ty to guess the ule geneating it. Thei method of guessing was to poduce othe thee-numbe sequences to which the expeimente would espond "yes" o "no" depending on whethe the new sequences wee consistent with the ule. Once confident with thei answes the subjeets would fomulate the ule. (Note the similaity of this expeiment to the diseussion in Chapte 1 of the way histoy pesents iself to us: assuming histoy is geneated accoding to some logic we see only the events neve the ules but need to guess how it woks.) The coeet ule was "numbes in ascending ode" nothing moe. Vey few subjects discoveed it beca use in ode to do so they had to ohe a seies in descending ode (that the expeimente would say "no" o). Wason noticed that the subjects had a ule in mind but gave As 1 said in the Pologue the likely no happening fiming the likely is equivalent to confiming is also a Black Swan. So discon- the unlikely. SHMONFlliMATON 59 him exanples aimed at confiming it instead of tying to supply seies that wee inconsistcn with thei hypothesis. Subjects tenaciously kept tying to confim the ules that they had made up. This expeiment inspied a collection of simila tests of which anothe example: Subjects wee asked which questions tu ask to find out whethe a peson was extoveted o not pupotedly fo anothe type of expeimento t was established that subjects supplied mostly questions fo which a "yes" answe would suppot the hypothesis. But thee ae exceptions. Among them figue chess gand mastes who i has been shown actually do focus on whee a speculative move might be weak; ookies by conpaison look fo confimatoy instances instead of falsifying ones. But don't play chess to pactiee skepticism. Scientists believe that it is the seach fo thei own wcaknesses that makcs then good chess playes not the pacice of chess that tuns then ino skeptics. Similaly the speculato Geoge Soos when making a financial bet kecps looking fo instances that would pove his initia theoy wong. This pehaps is tue self-confidenee: the ability tu look at the wold without the need to find signs that soke one's ego." Sadly the notion of cooboation is ooted in ou intellecual habits and discouse. Conside this comnent by he wite and citic John Updike: "Wben Julian Jaynes... speculates that until late in the second millenniun B.C. nen had no consciousness but wee automatically obeying the voices of gods we ae astounded but compelled to follow this enakable thesis though al! the cooboative evidence." Jaynes's thesis may be ight but M. Updike the cental poblem of knowledge (and the point of this chapte) is that thce is no such animal as cooboatiue evidence. equipment causing optical illusions; o he could be a bell-cuve-using economist disguised as a docto. An eyewitness to a cime might be dunk. But it emains the case that you know what is ioong with a lot moe confidence than you knaw what is ight. A pieces of infomation ae not CONFRMATON.1 Saw Anofhe :. Red Mini! The following point futhe illustates the absudity of confimation. f you believe that witnessing an additional white swan will bing confimation that thee ae no blaek swans then you should also accept the sta te- {.0:- ~ j.~. ti This confimation poblem pevades ou noden liíe since most conflicts have at thei oot the following mental bias: when Aa bs and lsaelis watch news epots hey see diffeent stoies in the sane succession of events Likcwise Democats and Republicans look a diffeent pats of he sane data and neve convege to [he same opinions. Once you mind is inhabied with a ceain view of the wo1d you will tend to only conside instances poving you to be ight. Paadoxical1y the moe infonation you have he moe justified you will feel in you views ~!~~.

6 60 UMBERTO ECO'S ANTLBRARY ment on pue\y logical gounds that the sighting lent to "all nonwhite objects ae not swans." would infe that the sighting of a nonwhite object that is not a swan should do x was ediscoveed This agument known as Hempe!'s by my fiend the (thinking) Dupie duing one of ou intense meditating intense walk-discussions instincts "Look Nassim lookl No Black Swan!" dominated by ae-vey thousands and thousands biologically Not Eveyfhing plague us eveywhee. But such smat pockets of inductive to involve events that we have encounteed mattes in ou natual n company tend chalaan envionmen membe when childen ae pesented with the pictue of a goup and ae asked to guess the popeties membes they ae capable of selecing which attibutes a child a photogaph of someone oveweight weigh-challenged. But she would espond diffeently t wil! not attack of known in- judgment miles tall t akes a lot moe secuity wild animals Black Swans) emains envionment new enemies enough athe analys is not a past hunans have multiplied they wee and abupt athe weathe quickly and to of souces of uncetainty ingained be- imited fo us to have built an in- to make infeences (i.e. focus on a smal! numbe will not is "ou ally" and quickly. of Black Swans today n the pimitive a make a county uso n the distant fa moe accuately the souces measuability.' "tunnel" is 3 it is impossible- seveal hunded a poject is hopeless nate fea of them. This instinct oí the tibe ae volving skin colo. f you show he people of dak complexion un into ahuman changes These events wee epeatable she will (mos to genealizations being Extemistan can delive a Black Swan afte will no go bust a bokeage-house to newly encounteed Show tell he that he is a membe that all the membes than we (and ou insti- wold days to accept that a wite is ungifed o a neighbo Futhemoe yond 1.- of othe unseen to genealize. of a tibe and ask he to descibe the est of the population: likely) not jump to he conclusion of a single '. in- ule these events out. But the sales of a book o the mag- could make infeenees fom which we have leaned to avoid foolish genealizaíon. Fo instance ae-events. cash a wa will no happen does not skepticism The moden post-alphabet envionment. of white ones so we need to withhold impossibje-to than a thousand we have neve seen him die o that soneone is innocent of mude beca use of na'ive genealization complex in the hail fom but these to the pesent nitude of social events do not follow such sictues. beca use 61 speculate fo suvival fo longe than we ae inclined to. As 1 said in Chape so ou intuitions we have neve seen him kil!. The poblem adequate is a bit moe complex uions) seem to ealize. How? of those We ae not naíve enough to believe that someone will be immotal and statistically fom ou ancestos. the instincts not well adapted ndeed ou envionment Buno intense to the point of ou not noticing the ain. He pointed to a ed Mini and shouted ae cetainly tcnsely infomational ayen paa- walks in London-one things wong inheited Enst Afican Geat Lakes egion whee we pesumably bent mathematician SHMONFRMATON! Medlo""'o' hee that we pobably What confims the latte statea mind with a confimation lo And we may have leaned is equiva- ment should confim the fome. Theefoe bing such confimation. 'oek of a ed Mini Coope should confim that thee ae no black swans. Why? Jus conside that the statement "al\ swans ae white" CONFRMATlON o causes in uso This instinct in a wod is ou pedicamento and ask he to descibe thei co-tibesmen she will assume that they too have dak skin. So it seems that we ae endowed with speciflc and e1aboate inductive instincts showing us the way. Contay to the opinion held by he geat David that helief aises fom custhat we lean genealizations solely fom expeience Hune and that of the Bitish empiicist tadiion tom as they assumed and empiical obsevations that we come equipped genealize leaning with mental machiney fom expeiences in so me domains i was shown fom studies of infant behavio (i.e. to selectively but enain skeptical fom a me e thousand fom the leaning that causes us to selectively acquie in othes). days bu benefiting of ou ancestos-which inductive leaning By doing so we ae not thanks to evolution found its wa)' into ou biology. c~ " \~....~ i i:' f ";t' Clealy weahe-elated and geodesic events (such as tonadoes and eathquakes) have not changed much ove the past millennium but what have changed ae the socioeconomic consequences of such occuences. Today an eathquake o huicane conmands moe and moe sevee economic consequences than it did in the past beca use of the intelocking elationships between economic enities and the intensification of the "netwok effects" that we wil! discuss in Pan Thee. Mattes that used to ha ve mi d effects now command a high impacto Tokyo's 1923 eathquake caused a dop of about a thid in japan's GNP. Extapolating fom the tagedy of Kobe in 1994 we can easily infe tha the consequences of anothe such eahquake in Tokyo would be fa costlie than that of its pedecesso