Coercive Formats

In How to Make a Com­plete Map of Every Thought You Think, Lion Kim­bro in­tro­duces the idea of co­er­cive vs un­co­er­cive for­mats for in­for­ma­tion.


  • Sto­ries.

  • Les­son plans; course pre­req­ui­sites.

  • Videos.

  • Lin­ear text which must be read in or­der. In­tro­duc­ing ter­minol­ogy and ideas once (rather than wher­ever they’re needed), with­out an in­dex or other bet­ter refer­ence mechanisms.

  • Ex­plain­ing what you’re get­ting at only via your chain of ar­gu­men­ta­tion, rather than up-front.

  • In­ter­twin­ing de­vel­op­ment of many differ­ent themes (mak­ing it im­pos­si­ble to look for only the in­for­ma­tion you want).


  • Blueprints.

  • Wikis.

  • Refer­ence man­u­als.

  • Hyper­links.

  • TLDRs.

  • Tools which make it easy to quickly get an overview, such as pa­per ab­stracts or in­tro­duc­tions which ac­tu­ally sum­ma­rize a work (rather than teas­ing its con­clu­sions or giv­ing au­to­bi­o­graph­i­cal con­text, etc).

  • Di­a­grams/​ta­bles/​charts and other graphic el­e­ments. (The eye is free to roam to whichever part is of in­ter­est; also, these el­e­ments can of­ten be made fairly self-ex­plana­tory, so that one doesn’t have to read ev­ery­thing else to see what is go­ing on.)

  • Tables of con­tents; out­lines.

  • Chap­ters and sec­tions which can be read in­de­pen­dently.

I ini­tially wrote the above lists as two para­graphs. Turn­ing them into bul­leted lists made them feel less co­er­cive. The rest of this post is co­er­cive, in that I don’t sep­a­rate out differ­ent top­ics in an eas­ily dis­t­in­guish­able way.

I’m not go­ing to claim that “less co­er­cive” is always bet­ter. Co­er­cive for­mats are of­ten more fun to read, for ex­am­ple. On this spec­trum, HPMOR would be more co­er­cive than the se­quences. Fur­ther­more, you might want to pre­sent in­for­ma­tion in a co­er­cive for­mat if get­ting in­for­ma­tion in the wrong or­der is likely to make a per­son ac­tively mi­s­un­der­stand a con­cept rather than only fail to un­der­stand. (Mere pre­req­ui­site con­cepts are not re­ally a good ex­cuse—a wiki-style for­mat, where pre­req­ui­site con­cepts are ob­ses­sively hy­per­linked, works fine.)

A ‘se­quence’ is a some­what co­er­cive for­mat—a set of posts in­tended to be read in a par­tic­u­lar or­der. How­ever, se­quences on LW1.0 were less co­er­cive than se­quences as they ex­ist on LW2.0. LW1.0 se­quences (mostly?) listed the posts in a se­quence with sum­maries of each post,, while LW2.0 se­quences only have a sum­mary of the en­tire se­quence at the top, w/​o a short de­scrip­tion of each post.

A co­er­cive for­mat can also be bet­ter for at­ten­tion man­age­ment. Un­co­er­cive for­mats can cre­ate a ten­dency for at­ten­tion to jump all over the place and not spend enough time on any one topic.

Co­er­cive for­mats seem work­able for small and medium-sized chunks of in­for­ma­tion, but be­come un­wieldy for large amounts of in­for­ma­tion. Imag­ine if an en­cy­clo­pe­dia were writ­ten to be read in a se­quen­tial or­der.

Un­co­er­cive for­mats of­ten take more effort to cre­ate. Draw­ing a good di­a­gram tends to take more effort than writ­ing the in­for­ma­tion in text (though, ob­vi­ously, this de­pends on a lot of fac­tors). Maybe the ideal should be to move in­for­ma­tion from more-co­er­cive to less-co­er­cive for­mats: it is good to ini­tially cre­ate a lot of con­tent in very lin­ear for­mats which have to be read from start to finish, but it is also good to even­tu­ally move things to a wiki or some­thing like that.

Another idea is that in­for­ma­tion should just be available in as broad a va­ri­ety of for­mats as pos­si­ble.

Co­er­cive for­mats seem to be bet­ter for con­vinc­ing peo­ple of a tar­get po­si­tion, whereas un­co­er­cive for­mats seem bet­ter for pre­sent­ing all of the in­for­ma­tion. This would make co­er­cive for­mats a sym­met­ric weapon. This isn’t en­tirely true—ad­ver­tise­ments of­ten seem to use in­for­ma­tion-de­liv­ery strate­gies which I would or­di­nar­ily char­ac­ter­ize as less co­er­cive (such as bul­leted lists, text el­e­ments that don’t need to be read in a par­tic­u­lar or­der, images). Maybe that’s be­cause they need to work with a low at­ten­tion span.

It is pos­si­ble that rigor­ous ar­gu­ment and care­ful anal­y­sis fits bet­ter in lin­ear, co­er­cive for­mats. So, it isn’t clear which for­mat­ting style is bet­ter for healthy epistemics.

Maybe com­mu­ni­ties which share in­for­ma­tion in a man­ner heav­ily bias to­ward co­er­cive for­mats have a greater risk of be­ing con­vinced by long and overly clever ar­gu­ments, whereas com­mu­ni­ties which share in­for­ma­tion heav­ily bi­ased to­ward non­co­er­cive for­mats suffer from overly quick im­pres­sions formed via merely glanc­ing at a mil­lion things.

But, let’s not jump to the con­clu­sion that a healthy bal­ance is suffi­cient. The ideal would be to get the ad­van­tages of both.