Privileging the Hypothesis

Sup­pose that the po­lice of Largeville, a town with a mil­lion in­hab­itants, are in­ves­ti­gat­ing a mur­der in which there are few or no clues—the vic­tim was stabbed to death in an alley, and there are no finger­prints and no wit­nesses.

Then, one of the de­tec­tives says, “Well… we have no idea who did it… no par­tic­u­lar ev­i­dence singling out any of the mil­lion peo­ple in this city… but let’s con­sider the hy­poth­e­sis that this mur­der was com­mit­ted by Mor­timer Q. Sn­od­grass, who lives at 128 Or­di­nary Ln. It could have been him, af­ter all.”

I’ll la­bel this the fal­lacy of priv­ileg­ing the hy­poth­e­sis. (Do let me know if it already has an offi­cial name—I can’t re­call see­ing it de­scribed.)

Now the de­tec­tive may per­haps have some form of ra­tio­nal ev­i­dence that is not le­gal ev­i­dence ad­mis­si­ble in court—hearsay from an in­for­mant, for ex­am­ple. But if the de­tec­tive does not have some jus­tifi­ca­tion already in hand for pro­mot­ing Mor­timer to the po­lice’s spe­cial at­ten­tion—if the name is pul­led en­tirely out of a hat—then Mor­timer’s rights are be­ing vi­o­lated.

And this is true even if the de­tec­tive is not claiming that Mor­timer “did” do it, but only ask­ing the po­lice to spend time pon­der­ing that Mor­timer might have done it—un­jus­tifi­ably pro­mot­ing that par­tic­u­lar hy­poth­e­sis to at­ten­tion. It’s hu­man na­ture to look for con­fir­ma­tion rather than dis­con­fir­ma­tion. Sup­pose that three de­tec­tives each sug­gest their hated en­e­mies, as names to be con­sid­ered; and Mor­timer is brown-haired, Fred­er­ick is black-haired, and He­len is blonde. Then a wit­ness is found who says that the per­son leav­ing the scene was brown-haired. “Aha!” say the po­lice. “We pre­vi­ously had no ev­i­dence to dis­t­in­guish among the pos­si­bil­ities, but now we know that Mor­timer did it!”

This is re­lated to the prin­ci­ple I’ve started call­ing “lo­cat­ing the hy­poth­e­sis,” which is that if you have a billion boxes only one of which con­tains a di­a­mond (the truth), and your de­tec­tors only provide 1 bit of ev­i­dence apiece, then it takes much more ev­i­dence to pro­mote the truth to your par­tic­u­lar at­ten­tion—to nar­row it down to ten good pos­si­bil­ities, each de­serv­ing of our in­di­vi­d­ual at­ten­tion—than it does to figure out which of those ten pos­si­bil­ities is true. It takes 27 bits to nar­row it down to ten, and just an­other 4 bits will give us bet­ter than even odds of hav­ing the right an­swer.

Thus the de­tec­tive, in call­ing Mor­timer to the par­tic­u­lar at­ten­tion of the po­lice, for no rea­son out of a mil­lion other peo­ple, is skip­ping over most of the ev­i­dence that needs to be sup­plied against Mor­timer.

And the de­tec­tive ought to have this ev­i­dence in their pos­ses­sion, at the first mo­ment when they bring Mor­timer to the po­lice’s at­ten­tion at all. It may be mere ra­tio­nal ev­i­dence rather than le­gal ev­i­dence, but if there’s no ev­i­dence then the de­tec­tive is ha­rass­ing and per­se­cut­ing poor Mor­timer.

Dur­ing my re­cent di­avlog with Scott Aaron­son on quan­tum me­chan­ics, I did man­age to cor­ner Scott to the ex­tent of get­ting Scott to ad­mit that there was no con­crete ev­i­dence what­so­ever that fa­vors a col­lapse pos­tu­late or sin­gle-world quan­tum me­chan­ics. But, said Scott, we might en­counter fu­ture ev­i­dence in fa­vor of sin­gle-world quan­tum me­chan­ics, and many-wor­lds still has the open ques­tion of the Born prob­a­bil­ities.

This is in­deed what I would call the fal­lacy of priv­ileg­ing the hy­poth­e­sis. There must be a trillion bet­ter ways to an­swer the Born ques­tion with­out adding a col­lapse pos­tu­late that would be the only non-lin­ear, non-uni­tary, dis­con­ti­nous, non-differ­en­tiable, non-CPT-sym­met­ric, non-lo­cal in the con­figu­ra­tion space, Liou­ville’s-The­o­rem-vi­o­lat­ing, priv­ileged-space-of-si­mul­tane­ity-pos­sess­ing, faster-than-light-in­fluenc­ing, acausal, in­for­mally speci­fied law in all of physics. Some­thing that un­phys­i­cal is not worth say­ing out loud or even think­ing about as a pos­si­bil­i­ty­with­out a rather large weight of ev­i­dence—far more than the cur­rent grand to­tal of zero.

But be­cause of a his­tor­i­cal ac­ci­dent, col­lapse pos­tu­lates and sin­gle-world quan­tum me­chan­ics are in­deed on ev­ery­one’s lips and in ev­ery­one’s mind to be thought of, and so the open ques­tion of the Born prob­a­bil­ities is offered up (by Scott Aaron­son no less!) as ev­i­dence that many-wor­lds can’t yet offer a com­plete pic­ture of the world. Which is taken to mean that sin­gle-world quan­tum me­chan­ics is still in the run­ning some­how.

In the minds of hu­man be­ings, if you can get them to think about this par­tic­u­lar hy­poth­e­sis rather than the trillion other pos­si­bil­ities that are no more com­pli­cated or un­likely, you re­ally have done a huge chunk of the work of per­sua­sion. Any­thing thought about is treated as “in the run­ning,” and if other run­ners seem to fall be­hind in the race a lit­tle, it’s as­sumed that this run­ner is edg­ing for­ward or even en­ter­ing the lead.

And yes, this is just the same fal­lacy com­mit­ted, on a much more blatant scale, by the the­ist who points out that mod­ern sci­ence does not offer an ab­solutely com­plete ex­pla­na­tion of the en­tire uni­verse, and takes this as ev­i­dence for the ex­is­tence of Je­ho­vah. Rather than Allah, the Fly­ing Spaghetti Mon­ster, or a trillion other gods no less com­pli­cated—never mind the space of nat­u­ral­is­tic ex­pla­na­tions!

To talk about “in­tel­li­gent de­sign” when­ever you point to a pur­ported flaw or open prob­lem in evolu­tion­ary the­ory is, again, priv­ileg­ing the hy­poth­e­sis—you must have ev­i­dence already in hand that points to in­tel­li­gent de­sign speci­fi­cally in or­der to jus­tify rais­ing that par­tic­u­lar idea to our at­ten­tion, rather than a thou­sand oth­ers.

So that’s the sane rule. And the cor­re­spond­ing anti-episte­mol­ogy is to talk end­lessly of “pos­si­bil­ity” and how you “can’t dis­prove” an idea, to hope that fu­ture ev­i­dence may con­firm it with­out pre­sent­ing past ev­i­dence already in hand, to dwell and dwell on pos­si­bil­ities with­out eval­u­at­ing pos­si­bly un­fa­vor­able ev­i­dence, to draw glow­ing word-pic­tures of con­firm­ing ob­ser­va­tions that could hap­pen but haven’t hap­pened yet, or to try and show that piece af­ter piece of nega­tive ev­i­dence is “not con­clu­sive.”

Just as Oc­cam’s Ra­zor says that more com­pli­cated propo­si­tions re­quire more ev­i­dence to be­lieve, more com­pli­cated propo­si­tions also ought to re­quire more work to raise to at­ten­tion. Just as the prin­ci­ple of bur­den­some de­tails re­quires that each part of a be­lief be sep­a­rately jus­tified, it re­quires that each part be sep­a­rately raised to at­ten­tion.

As dis­cussed in Per­pet­ual Mo­tion Beliefs, faith and type 2 per­pet­ual mo­tion ma­chines (wa­ter ice cubes elec­tric­ity) have in com­mon that they pur­port to man­u­fac­ture im­prob­a­bil­ity from nowhere, whether the im­prob­a­bil­ity of wa­ter form­ing ice cubes or the im­prob­a­bil­ity of ar­riv­ing at cor­rect be­liefs with­out ob­ser­va­tion. Some­times most of the anti-work in­volved in man­u­fac­tur­ing this im­prob­a­bil­ity is get­ting us to pay at­ten­tion to an un­war­ranted be­lief—think­ing on it, dwelling on it. In large an­swer spaces, at­ten­tion with­out ev­i­dence is more than halfway to be­lief with­out ev­i­dence.

Some­one who spends all day think­ing about whether the Trinity does or does not ex­ist, rather than Allah or Thor or the Fly­ing Spaghetti Mon­ster, is more than halfway to Chris­ti­an­ity. If leav­ing, they’re less than half de­parted; if ar­riv­ing, they’re more than halfway there.

An oft-en­coun­tered mode of priv­ilege is to try to make un­cer­tainty within a space, slop out­side of that space onto the priv­ileged hy­poth­e­sis. For ex­am­ple, a cre­ation­ist seizes on some (allegedly) de­bated as­pect of con­tem­po­rary the­ory, ar­gues that sci­en­tists are un­cer­tain about evolu­tion, and then says, “We don’t re­ally know which the­ory is right, so maybe in­tel­li­gent de­sign is right.” But the un­cer­tainty is un­cer­tainty within the realm of nat­u­ral­is­tic the­o­ries of evolu­tion—we have no rea­son to be­lieve that we’ll need to leave that realm to deal with our un­cer­tainty, still less that we would jump out of the realm of stan­dard sci­ence and land on Je­ho­vah in par­tic­u­lar. That is priv­ileg­ing the hy­poth­e­sis—tak­ing doubt within a nor­mal space, and try­ing to slop doubt out of the nor­mal space, onto a priv­ileged (and usu­ally dis­cred­ited) ex­tremely ab­nor­mal tar­get.

Similarly, our un­cer­tainty about where the Born statis­tics come from should be un­cer­tainty within the space of quan­tum the­o­ries that are con­tin­u­ous, lin­ear, uni­tary, slower-than-light, lo­cal, causal, nat­u­ral­is­tic, et cetera—the usual char­ac­ter of phys­i­cal law. Some of that un­cer­tainty might slop out­side the stan­dard space onto the­o­ries that vi­o­late one of these stan­dard char­ac­ter­is­tics. It’s in­deed pos­si­ble that we might have to think out­side the box. But sin­gle-world the­o­ries vi­o­late all these char­ac­ter­is­tics, and there is no rea­son to priv­ilege that hy­poth­e­sis.