Categorizing Has Consequences

Among the many ge­netic vari­a­tions and mu­ta­tions you carry in your genome, there are a very few alle­les you prob­a­bly know—in­clud­ing those de­ter­min­ing your blood type: the pres­ence or ab­sence of the A, B, and + anti­gens. If you re­ceive a blood trans­fu­sion con­tain­ing an anti­gen you don’t have, it will trig­ger an aller­gic re­ac­tion. It was Karl Land­steiner’s dis­cov­ery of this fact, and how to test for com­pat­i­ble blood types, that made it pos­si­ble to trans­fuse blood with­out kil­ling the pa­tient. (1930 No­bel Prize in Medicine.) Also, if a mother with blood type A (for ex­am­ple) bears a child with blood type A+, the mother may ac­quire an aller­gic re­ac­tion to the + anti­gen; if she has an­other child with blood type A+, the child will be in dan­ger, un­less the mother takes an aller­gic sup­pres­sant dur­ing preg­nancy. Thus peo­ple learn their blood types be­fore they marry.

Oh, and also: peo­ple with blood type A are earnest and cre­ative, while peo­ple with blood type B are wild and cheer­ful. Peo­ple with type O are agree­able and so­cia­ble, while peo­ple with type AB are cool and con­trol­led. (You would think that O would be the ab­sence of A and B, while AB would just be A plus B, but no...) All this, ac­cord­ing to the Ja­panese blood type the­ory of per­son­al­ity. It would seem that blood type plays the role in Ja­pan that as­trolog­i­cal signs play in the West, right down to blood type horo­scopes in the daily news­pa­per.

This fad is es­pe­cially odd be­cause blood types have never been mys­te­ri­ous, not in Ja­pan and not any­where. We only know blood types even ex­ist thanks to Karl Land­steiner. No mys­tic witch doc­tor, no ven­er­a­ble sor­cerer, ever said a word about blood types; there are no an­cient, dusty scrolls to shroud the er­ror in the aura of an­tiquity. If the med­i­cal pro­fes­sion claimed to­mor­row that it had all been a colos­sal hoax, we layfolk would not have one scrap of ev­i­dence from our un­aided senses to con­tra­dict them.

There’s never been a war be­tween blood types. There’s never even been a poli­ti­cal con­flict be­tween blood types. The stereo­types must have arisen strictly from the mere ex­is­tence of the la­bels.

Now, some­one is bound to point out that this is a story of cat­e­go­riz­ing hu­mans. Does the same thing hap­pen if you cat­e­go­rize plants, or rocks, or office fur­ni­ture? I can’t re­call read­ing about such an ex­per­i­ment, but of course, that doesn’t mean one hasn’t been done. (I’d ex­pect the chief difficulty of do­ing such an ex­per­i­ment would be find­ing a pro­to­col that didn’t mis­lead the sub­jects into think­ing that, since the la­bel was given you, it must be sig­nifi­cant some­how.) So while I don’t mean to up­date on imag­i­nary ev­i­dence, I would pre­dict a pos­i­tive re­sult for the ex­per­i­ment: I would ex­pect them to find that mere la­bel­ing had power over all things, at least in the hu­man imag­i­na­tion.

You can see this in terms of similar­ity clusters: once you draw a bound­ary around a group, the mind starts try­ing to har­vest similar­i­ties from the group. And un­for­tu­nately the hu­man pat­tern-de­tec­tors seem to op­er­ate in such over­drive that we see pat­terns whether they’re there or not; a weakly nega­tive cor­re­la­tion can be mis­taken for a strong pos­i­tive one with a bit of se­lec­tive mem­ory.

You can see this in terms of neu­ral al­gorithms: cre­at­ing a name for a set of things is like al­lo­cat­ing a sub­net­work to find pat­terns in them.

You can see this in terms of a com­pres­sion fal­lacy: things given the same name end up dumped into the same men­tal bucket, blur­ring them to­gether into the same point on the map.

Or you can see this in terms of the bound­less hu­man abil­ity to make stuff up out of thin air and be­lieve it be­cause no one can prove it’s wrong. As soon as you name the cat­e­gory, you can start mak­ing up stuff about it. The named thing doesn’t have to be per­cep­ti­ble; it doesn’t have to ex­ist; it doesn’t even have to be co­her­ent.

And no, it’s not just Ja­pan: Here in the West, a blood-type-based diet book called Eat Right 4 Your Type was a best­sel­ler.

Any way you look at it, draw­ing a bound­ary in thingspace is not a neu­tral act. Maybe a more cleanly de­signed, more purely Bayesian AI could pon­der an ar­bi­trary class and not be in­fluenced by it. But you, a hu­man, do not have that op­tion. Cat­e­gories are not static things in the con­text of a hu­man brain; as soon as you ac­tu­ally think of them, they ex­ert force on your mind. One more rea­son not to be­lieve you can define a word any way you like.