Judgment Day: Insights from ‘Judgment in Managerial Decision Making’

I found this book through the CFAR read­ing list. Some con­tent was pre­vi­ously posted on my short­form feed.


The more broadly I read and learn, the more I bump into im­plicit self-con­cep­tions and self-bound­aries. I was slightly averse to learn­ing from a man­ager-ori­ented text­book be­cause I’m not a man­ager, but also be­cause I… didn’t see my­self as the kind of per­son who could learn about a “busi­ness”-y con­text? I also didn’t see my­self as the kind of per­son who could read and do math, which now seems ridicu­lous.

Although the read was fast and easy and of­ten fa­mil­iar, I un­earthed a few gems.

Judg­ment in Man­age­rial De­ci­sion Making

A tip of du­bi­ous eth­i­cal­ity for lazy min-max­ers:

Man­agers give more weight to perfor­mance dur­ing the three months prior to [a perfor­mance] eval­u­a­tion than to the pre­vi­ous nine months… be­cause it is more available in mem­ory.

Unified ex­pla­na­tion of biases

The au­thors group bi­ases as stem­ming from the Big Three of availa­bil­ity, rep­re­sen­ta­tive­ness, and con­fir­ma­tion. The model I took away re­lies on a mechanism some­what similar to at­ten­tion in neu­ral net­works: due to how the brain performs time-limited search, more salient/​re­cent mem­o­ries get pri­ori­tized for re­call.

The availa­bil­ity heuris­tic goes wrong when our saliency-weighted per­cep­tions of the fre­quency of events is a bi­ased es­ti­ma­tor of the real fre­quency, when we hap­pen to be ex­trap­o­lat­ing off of a very small sam­ple size, or when our mem­ory struc­ture makes re­call­ing some kinds of things harder (e.g. words start­ing with ‘a’ ver­sus words whose third let­ter is ‘a’). Con­cepts get in­ap­pro­pri­ately ac­ti­vated in our mind, and we there­fore rea­son in­cor­rectly. At­ten­tion also ex­plains an­chor­ing: you can more read­ily bring to mind things re­lated to your an­chor due to salience.

The rep­re­sen­ta­tive­ness heuris­tic can be un­der­stood as highly salient con­cept-ac­ti­va­tions in­ap­pro­pri­ately dom­i­nat­ing our rea­son­ing. We then ig­nore e.g. base rates, sam­ple size, statis­ti­cal phe­nom­ena (re­gres­sion to the mean), and the con­junc­tive bur­den of propo­si­tions. Con­sider how neu­ral net­work ac­ti­va­tions could ex­plain the fol­low­ing:

In­tu­itively, think­ing of Linda as a fem­i­nist bank tel­ler “feels” more cor­rect than think­ing of her as only a bank tel­ler.

The case for con­fir­ma­tion bias seems to be a lit­tle more in­volved. We had evolu­tion­ary pres­sure to win ar­gu­ments, which might mean our cog­ni­tive search aims to find sup­port­ive ar­gu­ments and avoid even sub­con­sciously sig­nal­ling that we are aware of the ex­is­tence of coun­ter­ar­gu­ments. This means that those sup­port­ive ar­gu­ments feel salient, and we (per­haps by “de­sign”) get to feel un­bi­ased—we aren’t con­sciously dis­card­ing ev­i­dence, we’re just fol­low­ing our nor­mal search/​rea­son­ing pro­cess! This is what our search al­gorithm feels like from the in­side.

Mak­ing heads and tails of prob­a­bil­is­tic reasoning

In Sub­jec­tive Prob­a­bil­ity: A Judg­ment of Rep­re­sen­ta­tive­ness, Kah­ne­man and Tver­sky hy­poth­e­size that

since sam­ple size does not rep­re­sent any prop­erty of the pop­u­la­tion, it is ex­pected to have lit­tle or no effect on judg­ment of like­li­hood,

which lines up with the above at­ten­tion/​ac­ti­va­tion model. Any­ways, par­ti­ci­pants judged that the se­quence of birth sexes is more likely than (ob­vi­ously, they have equal prob­a­bil­ity). K&T chalk this up to the first se­quence seem­ing more “rep­re­sen­ta­tive” of a “ran­dom” pro­cess. If you’re con­sid­er­ing whether the set of all se­quences which “look like” the former is more likely than the set of se­quences re­sem­bling the lat­ter, then this an­swer could be cor­rect.

How­ever, check­ing the origi­nal pa­per, this was con­trol­led for; K&T em­pha­sized that the ex­act or­der of births was as de­scribed. They go on:

One may won­der whether [sub­jects] do not sim­ply ig­nore or­der in­for­ma­tion, and an­swer the ques­tion by eval­u­at­ing the fre­quency of fam­i­lies of five boys and one girl rel­a­tive to that of fam­i­lies of three boys and three girls. How­ever, when we asked the same [sub­jects] to es­ti­mate the fre­quency of the se­quence , they viewed it as sig­nifi­cantly less likely than (), pre­sum­ably be­cause the former ap­pears less ran­dom. Order in­for­ma­tion, there­fore, is not sim­ply ig­nored.

Share your unique in­for­ma­tion in groups

Groups are good be­cause they pool knowl­edge and ex­per­tise. How­ever, stud­ies show that by de­fault, shared knowl­edge is much more likely to be dis­cussed than un­shared knowl­edge, which can sig­nifi­cantly worsen de­ci­sion-mak­ing. The au­thors give the ex­am­ple of a group ini­tially dis­fa­vor­ing a stu­dent coun­cil can­di­date. One per­son is pri­vately aware of cru­cial pos­i­tive in­for­ma­tion about the can­di­date, and groups in which all mem­bers knew the info were likely to fa­vor the can­di­date. The in­for­ma­tion wasn’t usu­ally shared, and the can­di­date was passed over.

[Ame­lio­ra­tive] strate­gies in­clude fore­warn­ing the group in ad­vance of the unique knowl­edge of differ­ent mem­bers and iden­ti­fing ex­per­tise in the group be­fore the dis­cus­sion be­gins.

Open sub­ac­counts for savings

You can avoid psy­cholog­i­cal an­noy­ances through­out the year (tick­ets, unan­ti­ci­pated fees, etc.) and coun­ter­act the bud­get-plan­ning fal­lacy by, at the be­gin­ning of each year, al­lo­cat­ing money to goal-spe­cific sub­ac­counts. Then, you can for­get about it dur­ing the year, and (per­haps) donate the re­main­der to an effec­tive char­ity.

Be more risk-neutral

Paul Sa­muel­son… offered a col­league a coin-toss gam­ble. If the col­league won the coin toss, he would re­ceive $200, but if he lost, he would lose $100. Sa­muel­son was offer­ing his col­league a pos­i­tive ex­pected value with risk. The col­league, be­ing risk-averse, re­fused the sin­gle bet, but said that he would be happy to toss the coin 100 times! The col­league un­der­stood that the bet had a pos­i­tive ex­pected value and that across lots of bets, the odds vir­tu­ally guaran­teed a profit. Yet with only one trial, he had a 50% chance of re­gret­ting tak­ing the bet.

Notably, Sa­muel­son‘s col­league doubtless faced many gam­bles in life… He would have fared bet­ter in the long run by max­i­miz­ing his ex­pected value on each de­ci­sion… all of us en­counter such “small gam­bles” in life, and we should try to fol­low the same strat­egy. Risk aver­sion is likely to tempt us to turn down each in­di­vi­d­ual op­por­tu­nity for gain. Yet the ag­gre­gated risk of all of the pos­i­tive ex­pected value gam­bles that we come across would even­tu­ally be­come in­finites­i­mal, and po­ten­tial profit quite large.

Biolog­i­cal ex­pla­na­tion for he­do­nic tread­mill?

The strik­ing as­pect about fram­ing and refer­ence-point effects is that they sug­gest the pres­ence of un­der­ly­ing men­tal pro­cesses that are more com­pli­cated than a ra­tio­nal de­ci­sion-maker would em­ploy. Ra­tional de­ci­sion-mak­ers would sim­ply seek to max­i­mize the ex­pected value of their choices. Whether these out­comes rep­re­sent gains or losses would be ir­rele­vant, and con­sid­er­a­tion of the out­come rel­a­tive to the sta­tus quo would be a su­perflu­ous con­sid­er­a­tion. in­stead, we ad­just to the sta­tus quo, and then think of the changes from that point as gains or losses.

Rayo and Becker (2007) pre­sent a per­sua­sive ex­pla­na­tion for why evolu­tion pro­grammed us with ex­tra ma­chin­ery that im­pairs our de­ci­sions. Ac­cord­ing to their ex­pla­na­tion, our re­li­ance on frames and refer­ence points to as­sess out­comes is an el­e­gant solu­tion to a prob­le­matic biolog­i­cal con­straint. The con­straint is that our “sub­jec­tive util­ity scale” – our abil­ity to ex­pe­rience plea­sure and pain – is not in­finitely sen­si­tive. Was Bill Gates’s 50th billion dol­lars as satis­fy­ing as his first? cer­tainly not. the limited sen­si­tivity of our sub­jec­tive util­ity scale is pre­cisely the rea­son why we ex­pe­rience de­clin­ing marginal util­ity for both gains and losses…

Given this biolog­i­cal con­straint on the sen­si­tivity of our sub­jec­tive util­ity scale, we need to read­just our refer­ence point by get­ting used to what we’ve got and then tak­ing it for granted. If we didn’t ad­just our refer­ence point, we could quickly hit the max­i­mum of our util­ity scale, and re­al­ize that noth­ing we could ever do would make us hap­pier. That would effec­tively kill our mo­ti­va­tion to work harder, be­come richer, and achieve more. In re­al­ity, of course, what hap­pens is that we get used to our cur­rent level of wealth, sta­tus, and achieve­ment, and are then mo­ti­vated to seek more, be­liev­ing that it will make us hap­pier.

The irony of this mo­ti­va­tional sys­tem is that for it to keep work­ing, we have to ha­bit­u­ate to our new con­di­tion but not an­ti­ci­pate this ha­bit­u­a­tion. Ev­i­dence does in­deed con­firm that peo­ple ad­just to both pos­i­tive and nega­tive changes in cir­cum­stances with sur­pris­ing speed, and then promptly for­get that they did so. Thus, we find our­selves on a he­do­nic tread­mill in which we strive for an imag­ined hap­piness that for­ever slips out of our grasp, beck­on­ing us on­ward.

Ne­go­ti­a­tion tips

Chap­ters 9 and 10 con­tain a wealth of (seem­ingly) good ne­go­ti­a­tion ad­vice. Be­ing a good ne­go­tia­tor and me­di­a­tor seems like an im­por­tant gen­er­al­ist life skill.

  • Be­fore ne­go­ti­a­tion, con­sider all rele­vant is­sues and their im­por­tance to you and then to your part­ner. Try to spot places where you can make effi­cient pos­i­tive-sum bar­gains. Figure out your next-best al­ter­na­tive if a deal can­not be struck, and an­ti­ci­pate what your part­ner’s will be as well.

  • Find the in­tent gen­er­at­ing their stated po­si­tion. Maybe your boss states they don’t want you in­stal­ling a stand­ing desk, but they’re se­cretly wor­ried it’ll lead to a slip­pery slope of em­ploy­ees in­stal­ling in­creas­ingly dis­tract­ing ac­ces­sories. If you can find this out, you can offer your sup­port in pre­vent­ing a slip­pery slope, in­stead of try­ing to push on the more difficult all-or-noth­ing po­si­tion.

  • Ne­go­ti­ate mul­ti­ple is­sues si­mul­ta­neously. You’re bet­ter able to find pos­i­tive-sum agree­ments when con­sid­er­ing mul­ti­ple axes at once. Also, you’ll avoid mak­ing them com­pro­mise too hard and too early on things which aren’t im­por­tant to you, which can make the later part of ne­go­ti­a­tion less fruit­ful for you.

There were a lot more helpful take­aways, and I plan on reread­ing Ch. 9 be­fore con­duct­ing any im­por­tant ne­go­ti­a­tions.


This book was a lit­tle slow at times, both be­cause of ex­ces­sive pream­ble/​sign­post­ing, and my already be­ing fa­mil­iar with much of the liter­a­ture. Still, I’m glad I read it.

Hello again

It’s been a long while since my last re­view. After in­jur­ing my­self last sum­mer, I wasn’t able to type re­li­ably un­til early this sum­mer. This de­railed the pos­i­tive feed­back loop I had around “learn math” → “write about what I learned” → “sa­vor karma”. Pro­tect your feed­back loops.

I run into fewer ba­sic con­fu­sions than when I was just start­ing at math, so I gen­er­ally have less to talk about. This means I’ll be chang­ing the style of any up­com­ing re­views, in­stead fo­cus­ing on deeply ex­plain­ing the things I found coolest.

Since Jan­uary, I’ve read Vi­sual Group The­ory, Un­der­stand­ing Ma­chine Learn­ing, Com­pu­ta­tional Com­plex­ity: A Con­cep­tual Per­spec­tive, In­tro­duc­tion to the The­ory of Com­pu­ta­tion, An Illus­trated The­ory of Num­bers, most of Tadel­lis’ Game The­ory, the be­gin­ning of Mul­ti­a­gent Sys­tems, parts of sev­eral graph the­ory text­books, and I’m go­ing through Munkres’ Topol­ogy right now. I’ve got­ten through the first fifth of the first Feyn­man lec­tures, which has given me an un­be­liev­able amount of mileage for gen­er­ally rea­son­ing about physics.

My “plan” is to keep learn­ing math un­til the low grad­u­ate level (I still need to at least do com­plex anal­y­sis, topol­ogy, field /​ ring the­ory, ODEs/​PDEs, and some­thing to shore up my atro­cious trig skills, and prob­a­bly more)[1], and then branch off into physics + a “softer” sci­ence (any­thing from microe­con to psy­chol­ogy).

New year, new decade

In the new year, I’m go­ing to fo­cus hard on rais­ing the level of my cog­ni­tive game.

Read­ing the Se­quences qual­i­ta­tively lev­el­led me up, and I want to do that again. My thought pro­cesses are still in­suffi­ciently trans­par­ent: I need to flag mo­ti­vated rea­son­ing more of­ten. I still fall prey to the plan­ning fal­lacy (but some­what less than two years ago). I don’t no­tice my con­fu­sion nearly as of­ten as I should.

Not notic­ing con­fu­sion of­ten has a cost mea­sured in hours (or more). Let me give you an ex­am­ple. Last night, I went to speak with Sen. Amy Klobuchar about effec­tive al­tru­ism. It was my un­der­stand­ing that the event would be a meet-and-greet. I planned to query her in­ter­est in e.g. set­ting up a grant­ing agency dis­burs­ing funds based on sci­en­tific ev­i­dence of high im­pact, with the de­tails to be worked out in con­junc­tion with rele­vant pro­fes­sion­als in EA and the gov­ern­ment.

While I was wait­ing for her to ar­rive, I no­ticed that peo­ple were writ­ing on pa­per and hand­ing it to other peo­ple. I rounded this off as com­mit-to-cau­cus cards, which, if I ac­tu­ally thought about it, makes no sense – you keep your com­mit-to-cau­cus card. They were, in fact, pro­vid­ing writ­ten ques­tions, some of which Sen. Klobuchar would later an­swer. If I had just no­ticed this, I could have writ­ten a ques­tion and then left, sav­ing my­self two hours.

The list of things I’ve no­ticed I failed to no­tice in the last month is sur­pris­ingly long. I don’t think I’m bad at this in a rel­a­tive sense – just in an ab­solute sense.

This new year, I’m go­ing to be­come a less oblivi­ous, less stupid, and less wrong per­son.

  1. I also still want to learn Bayes nets, cat­e­gory the­ory, get a much deeper un­der­stand­ing of prob­a­bil­ity the­ory, prov­abil­ity logic, and de­ci­sion the­ory. ↩︎