Against naming things, and so on

Re­cent dis­cus­sion on nam­ing con­cepts mostly fo­cuses on ar­gu­ments in fa­vor, not­ing only a few caveats, as LW user Conor More­ton in Why and How to Name Things:

What you lose by the pro­lifer­a­tion of jar­gon is ease-of-en­try and cross-cul­tural in­tel­ligi­bil­ity and hard-drive space in the brains of peo­ple try­ing to track all of it.

I think you lose more than that, so I’ll try to name [sic] a few more rea­sons your cau­tion might run deeper.

1. The nomo­th­etic fallacy

Di­ag­no­sis isn’t a cure, but it can bring a sense of re­lief un­teth­ered from prog­no­sis. You might still have the same symp­toms that still need the same treat­ment, but it feels like a large part of the prob­lem has been solved. This is, I’d ar­gue, a de­cent chunk of why know­ing about bi­ases can hurt you. Don’t get com­pla­cent just be­cause you have a name for some­thing.

2. Weaponized rationality

Another big chunk of “know­ing about bi­ases can hurt you” comes from wield­ing con­cepts not in in­tro­spec­tion but against oth­ers. (Like many of the con­sid­er­a­tions be­low, tak­ing this as com­pel­ling is a gen­eral ar­gu­ment against much of the ra­tio­nal­ity pro­ject. But pro­lifer­at­ing jar­gon, to my eye, makes most of the prob­lems here rel­a­tively worse for ex­ter­nal use as com­pared to in­ter­nal use.)

3. Be­ing wrong

Some­times your anal­y­sis of the thing you’re try­ing to crys­tal­lize just isn’t very good. Your anal­ogy doesn’t play out the way you think, the pat­tern doesn’t ac­tu­ally ex­ist, you’re not carv­ing re­al­ity at the joints, your fram­ing could be bet­ter, it’s an in­ap­pro­pri­ate level of ab­strac­tion any­where you’d ac­tu­ally want to use it. But, hey, the name stuck, so you can com­mu­ni­cate all that in just a word!

4. Lossy compression

Maybe you got it right this time, but the name doesn’t cap­ture ev­ery­thing. In prac­tice, no­body’s go­ing to re­mem­ber your en­tire blog post ev­ery time some­one ut­ters the ti­tle. I hope you didn’t need that nu­ance. But at least this one went viral!

5. Think­ing on the page

From Conor’s post:

When you define a con­cept rigor­ously and clearly, you al­most always learn new things from play­ing around with their edges and try­ing to get them to in­ter­face with other rigor­ous, clear con­cepts.

And if you don’t define it rigor­ously and clearly—as is the case for pretty much ev­ery­thing hap­pen­ing here—if you play di­rectly with the com­pressed con­cepts and string them to­gether that way, oops, you just proved that 2+2=5. You lost track of the ac­tual stuff the name referred to, the rules for ma­nipu­lat­ing it while pre­serv­ing truth. You thought you were post-rigor when there isn’t even a rigor­ous stage. You were think­ing on the page. [Oh, look, links to named chun­ked con­cepts. What are they do­ing here, of all places?]

Com­bin­ing frame­works and mak­ing analo­gies be­tween your con­cepts is a good way of gen­er­at­ing new ideas, but it’s far from rigor­ous and if any­thing en­courages slop­piness if you’re not care­ful to rec­og­nize that it’s less for­mal rea­son­ing and more an­other way of point­ing to where the new things to learn might be.

6. Illu­sion of transparency

Peo­ple will as­sume they know what you mean. You’ll as­sume you know what they mean in re­ply—that is, what a co­in­ci­dence, that they know what you mean. They don’t.

7. Lift­ing the pot by one handle

The trou­ble with ra­tio­nal­ist skills is that the op­po­site of ev­ery ra­tio­nal­ist skill is also a ra­tio­nal­ist skill.


In the com­ments to Epistemic Learned Hel­pless­ness (now there’s a name, eh?), kom­pon­isto writes:

We have the In­side View, and the Out­side View. Over­con­fi­dence is a prob­lem, but so is un­der­con­fi­dence. You’re sup­posed to listen to the tiniest note of men­tal dis­cord, yet some­times it’s nec­es­sary to shut loud men­tal voices out. And while know­ing the stan­dard cat­a­log of bi­ases is ob­vi­ously cru­cial for the as­piring ra­tio­nal­ist, it can also hurt you. Et cetera, et cetera.

Is it lo­tus eat­ing or self-care? Should you re­verse any ad­vice you hear or not? Re­mem­ber­ing a blog­post ti­tle isn’t go­ing to tell you which situ­a­tion you’re in; it’s at best a han­dle for notic­ing you’re in a situ­a­tion where you might want to course cor­rect fol­low­ing deeper anal­y­sis. But im­plicit in the as­sign­ment of han­dles is a claim that you ought to ad­just more than you do in a cer­tain di­rec­tion. And you don’t want achiev­ing a bal­ance of op­po­site mis­takes to de­pend on the rel­a­tive catch­i­ness of named con­cepts. When you lift the pot by one han­dle, the soup spills out.

(This isn’t, on its own, an ar­gu­ment against giv­ing your ideas catchy names, so much as an ar­gu­ment for spend­ing plenty of time dis­cussing trade-offs and how to do the nec­es­sary deeper anal­y­sis. But re­call that it’s still mostly the han­dle that gets re­mem­bered.)

8. Reification

You’ll tend to treat your named ideas as more con­crete than they are, as mean­ing one par­tic­u­lar thing to ev­ery­one, as hav­ing agency or causal pow­ers, as un­chang­ing and non-re­frame­able, as uni­ver­sal and con­text-in­de­pen­dent, as ter­ri­tory, as bi­nary, as ob­jec­tive.

(I won­der if this is re­lated to a lot of our ap­par­ent con­fu­sion about dou­ble crux—peo­ple (in­clud­ing me) ques­tion­ing its sta­tus as The Tech­nique, while Dun­can keeps try­ing to clar­ify how it’s not about ex­e­cut­ing an al­gorithm, it’s about in­ter­nal­iz­ing what gen­er­ates the al­gorithm; it’s not de­signed for use in earnest, it’s a ped­a­gog­i­cal tool for prac­tic­ing men­tal habits; it’s not about Dou­ble Crux, it’s about dou­ble-crux­i­ness.)

9. Round­ing error

Conor again:

[What you gain by the pro­lifer­a­tion of jar­gon in­cludes] be­ing em­bed­ded in a cul­ture where peo­ple are dili­gently seek­ing out and pop­u­lariz­ing such dis­tinc­tions makes a given in­di­vi­d­ual far more likely to pay at­ten­tion to sub­tle dis­tinc­tions them­selves, ac­cel­er­at­ing the pro­cess of cul­tural ac­cu­mu­la­tion of nu­ance and de­tail.

Or, well, the op­po­site of that, where you round ev­ery­thing off to in­group-ap­proved jar­gon.


You can’t stop ti­tling your blog posts, but don’t force it, or you’ll end up with viral­ity track­ing the wrong things (more so than usual, any­way). In par­tic­u­lar, maybe don’t op­ti­mize so much to­wards things like catch­i­ness, metaphor­i­cal weight, unique­ness, and com­mu­ni­ca­bil­ity—even for the sake of rapid, rigor­ous, dis­tinc­tion-rich, high-level dis­cus­sion (or per­haps es­pe­cially for that sake, since this kind of chunk­ing is ei­ther un­nec­es­sary or de­struc­tive for all of these but speed)—over choos­ing words that pre­pare peo­ple to read and un­der­stand what you say.

(And even as far as speed of com­mu­ni­ca­tion, what’s weird is that the ra­tio­nal­ist writ­ing stereo­type is ab­surdly pro­lix—it feels like the com­pres­sion isn’t hap­pen­ing where it mat­ters, just where it feels pow­er­ful for hid­ing com­plex ab­strac­tion in a min­i­mally in­tel­ligible way.)

You shouldn’t stop try­ing to draw more dis­tinc­tions, but maybe you can do that with­out en­courag­ing one par­tic­u­lar way of point­ing to your dis­tinc­tion to reify and metas­ta­size.

You won’t stop slop­pily com­press­ing and com­bin­ing com­plex ideas, but maybe worry more about un­pack­ing things, giv­ing ex­am­ples, check­ing un­der­stand­ing, and ask­ing ques­tions.