Cooperation is for Winners

Cross-posted from Pu­tanu­monit.


My last post was about learn­ing to com­pete well from sports. But why com­pete at all? After all, I’m the one who wrote that win­ning is for losers, and we should avoid get­ting sucked into zero-sum con­tests.

Com­pet­ing well doesn’t nec­es­sar­ily mean hav­ing to com­pete more, for one thing. In­stead, it does al­low you to choose where to com­pete. More im­por­tantly, sports-like com­pe­ti­tions cre­ate pres­tige hi­er­ar­chies (as op­posed to dom­i­nance hi­er­ar­chies). Climb­ing those is not in­tended to make you a fear­some boss but a valuable ally, one that oth­ers want to co­op­er­ate with. Get­ting to co­op­er­a­tion of­ten re­quires win­ning com­pe­ti­tions.

Wei Dai com­mented on the pre­vi­ous post:

More se­ri­ously, these days I think of com­pe­ti­tion as more of a prob­lem than a solu­tion. Some of the most im­por­tant x-risks (e.g., ad­vanced AI) are x-risks mainly be­cause of com­pet­i­tive dy­nam­ics. If peo­ple weren’t com­pet­ing for the pres­tige/​power/​money of be­ing first to cre­ate AGI or to make ad­vances in AI in gen­eral, we’d be able to solve AI safety prob­lems at leisure.

How does one get to solve AI safety prob­lems at their leisure, in a co­op­er­a­tive en­vi­ron­ment? Ob­vi­ously, by work­ing at the Ma­chine In­tel­li­gence Re­search In­sti­tute.

But to work at MIRI one must first get a job at MIRI, and since MIRI has few spots and many ap­pli­cants this means out­com­pet­ing other can­di­dates for the job. MIRI it­self is funded by dona­tions, so it has to out­com­pete other non-prof­its for grants.

Hope­fully, both com­pe­ti­tions are con­ducted in a sports­man­like way by demon­strat­ing re­search achieve­ments in­stead of by sab­o­tag­ing com­peti­tors. But it’s still a com­pe­ti­tion – ap­ply­ing for jobs or grans is stress­ful and en­tails a high risk of failure, of­ten due to cir­cum­stances out­side one’s con­trol.

Or­ga­ni­za­tions have to man­age very care­fully the pro­cess of di­rect­ing the win­ners of a com­pet­i­tive se­lec­tion pro­cess to an ul­ti­mately co­op­er­a­tive en­deavor. Here’s a won­der­ful post look­ing at this is­sue in two al­most iden­ti­cal situ­a­tions: chicken coops and academia.

So in the late 70s and early 80s there were ac­tive breed­ing pro­grams to pro­duce hens who laid more eggs. Take the high­est egg lay­ers in each gen­er­a­tion and let them be the only ones that breed to pro­duce the next gen­er­a­tion. […]
Only one prob­lem. High pro­duc­tivity egg-lay­ing is as­so­ci­ated with ag­gres­sion – in­deed the high­est egg lay­ers are ba­si­cally the ones that beat up the other hens in the coop with them and cap­ture the most re­sources. This led to a choice be­tween two routes. If noth­ing was done, to­tal egg-lay­ing went down as there were too many in­juries and too much mor­tal­ity. The al­ter­na­tive was to con­tinue the se­lec­tion pro­cess but to sup­ple­ment it by en­gag­ing in what many would re­gard as bar­baric prac­tices – beak trim­ming, de­claw­ing etc to pre­vent in­juries from the ag­gres­sion.
Then in the 1980s peo­ple got the idea to use group se­lec­tion. In­stead of pick­ing in­di­vi­d­u­als that were most pro­duc­tive, they se­lected en­tire hen houses that were most pro­duc­tive to pro­duce the next gen­er­a­tion. […]
Now, when I tell fel­low fac­ulty this story, many im­me­di­ately re­spond all on their own “that’s like the se­lec­tion be­ing im­posed for high-perform­ing in­di­vi­d­ual fac­ulty mem­bers at the cost of their de­part­ments and the larger mis­sion”. Are Dean’s and higher ups com­mit­ting the same er­ror as hen breed­ers and rely­ing too much on a one-di­men­sional in­di­vi­d­ual level se­lec­tion that re­sults in de­stroy­ing over­all pro­duc­tivity? Are we pro­duc­ing pro­fes­sors who need to be de­clawed and have their beaks trimmed to main­tain pro­duc­tivity (or less figu­ra­tively is de­part­men­tal col­le­gial­ity suffer­ing as a re­sult)?

Hav­ing to com­pete for the op­por­tu­nity to co­op­er­ate is how dat­ing works too. Your goal is to reach co­op­er­a­tion (for a night or for a life­time) with the per­son you’re pur­su­ing. But you’re com­pet­ing with ev­ery­one else: com­pet­ing to be no­ticed, to go on a date, to earn their com­mit­ment. A lot of peo­ple I know miss out on re­la­tion­ships be­cause they’re afraid of los­ing the com­pe­ti­tion – get­ting re­jected.

They are ways to sneak around this fear, but they’re not op­ti­mal. Guys can use Tin­der, which is set up to make “not be­ing swiped right on” feel as lit­tle as pos­si­ble like re­jec­tion. But Tin­der is su­perfi­cial, has poor norms around it, and only re­ally works for the 20% of men who are most at­trac­tive.

Women can just choose not to ask any­one out at all, whether on­line or in phys­i­cal spaces. But that limits a woman’s choices to her OkCupid in­box or the men who hit on her, which se­lects for men who are ag­gres­sive flirters in­stead of the other at­tributes that are much more im­por­tant in a boyfriend. I think that most women can do much bet­ter by reach­ing out them­selves to bet­ter guys than their in­box pro­vides.

Get­ting over the fear of re­jec­tion is hard, but here again, sports can show the way. Los­ing a ten­nis match is less bad but has a lot in com­mon with get­ting turned down in a ro­man­tic set­ting. In the mo­ment, it feels ter­rible. But af­ter a while, you re­flect that los­ing the match/​date is not a whole­sale con­dem­na­tion of your worth as a per­son or your so­cial stand­ing. It tells you where your ex­pec­ta­tions should be set in the fu­ture and, most cru­cially, how you should im­prove.

There are com­pet­i­tive el­e­ments even in things that don’t look like a com­pe­ti­tion at all, like blog­ging. I en­joy read­ing all across the ra­tio­nal­ist blo­go­sphere. We build on each other’s ideas, link to each other’s posts, and sus­tain a shared com­mu­nity. But no­body (with one ex­cep­tion) has time to read all the blogs, so we are also com­pet­ing for reader at­ten­tion, links, and space on Scott’s blogroll.

This doesn’t mean that I post with the in­ten­tion of steal­ing read­ers away from John or Sarah. But it im­plies that I think that my blog is com­pet­i­tive with theirs, that it’s worth some­one’s time to read Pu­tanu­monit on their re­stroom break. Blog­ging also means I’m will­ing to take an oc­ca­sional L: writ­ing some­thing stupid or poli­ti­cal or just poorly writ­ten and be­ing told by peo­ple that I wasted their re­stroom break and they’re not get­ting it back. Hope­fully, these losses make Pu­tanu­monit bet­ter.

Only 20% of LessWrong par­ti­ci­pants ac­tive enough to fill out a sur­vey have ever writ­ten a post. When I ask my friends who are in the re­main­ing 80%, they don’t say it’s be­cause they have liter­ally noth­ing to con­tribute. They’re afraid that what they write will not be good, and if it’s good it won’t be origi­nal, and if it’s good and origi­nal then Scott will write about the same thing bet­ter in a few weeks any­way. They don’t like the karma sys­tem of up­votes and down­votes. In short: they know it’s a com­pe­ti­tion, and they don’t want to com­pete.

No one gets to hear their ideas, en­joy their writ­ing, or cor­rect their mis­takes. If co­op­er­a­tion is hap­pen­ing, they’re not a part of it.

Win­ning is for Losers was about one sort of mis­take peo­ple can make: they see a hi­er­ar­chy of points and rank­ings (e.g., the en­tire Amer­i­can ed­u­ca­tion sys­tem) and ded­i­cat­ing all their re­sources to win­ning in­stead of ask­ing whether the game is even worth it. But I’m see­ing the op­po­site mis­take more and more, peo­ple see­ing com­pe­ti­tion and flinch­ing away in­stinc­tively. They tell them­selves that they’re the good guys, co­op­er­at­ing in­stead of fight­ing.

But if win­ning is for losers, co­op­er­a­tion is for the win­ners.