Three Kinds of Research Documents: Clarification, Explanatory, Academic

Epistemic Sta­tus: Low. This was a quick idea, but the group­ing hon­esty doesn’t work as well as I’d like. I still think it could be use­ful to some peo­ple though. Ideas ap­pre­ci­ated.

Re­cently I have started writ­ing more and have been try­ing to be more in­ten­tional with what I ac­com­plish. Differ­ent doc­u­ments have differ­ent pur­poses and it seemed use­ful to help clar­ify this. Here is a list of three spe­cific differ­ent types I think are rele­vant on LessWrong and similar.


I see clar­ifi­ca­tion posts as gen­er­ally the first in­stance of in­for­ma­tion be­ing writ­ten down. Here it is im­por­tant to get the es­sen­tial ideas out there and to cre­ate con­sen­sus around ter­minol­ogy among the most in­ter­ested read­ers. In some cases, the only in­ter­ested reader may be the au­thor, who would use the post just to help ce­ment their ideas for them­selves.

Clar­ifi­ca­tion posts may not be im­me­di­ately use­ful and re­quire later posts or con­text for them to make sense. This is typ­i­cally fine. There’s of­ten not a rush for them to be un­der­stood. In many cases, there is a lot of pos­si­ble in­for­ma­tion to write down, so the first step is to en­sure it’s out there, even if it’s slow, hard to read, or doesn’t much make sense un­til later.

I think of many of Paul Chris­ti­ano’s posts as clar­ifi­ca­tion posts. They’re very nu­mer­ous and novel, but quite con­fus­ing to many read­ers (at least, to my­self and sev­eral peo­ple I’ve talked to). Some­times the ter­minol­ogy changes from one post to the next. I used to see this is some­what of a weak­ness, but now it comes across to me as a prag­matic op­tion. If he were to have tried to make all of this read­able to the av­er­age LessWrong reader, there’s likely no way he could have writ­ten a por­tion as much.

One im­por­tant point here is that if some­thing is a clar­ifi­ca­tion post, then the main rele­vant feed­back is on the core con­tent, not the pre­sen­ta­tion. Giv­ing feed­back on the read­abil­ity can still be use­ful, but it should be un­der­stood and ex­pected that this isn’t the main goal.


Ex­plana­tory posts seek to ex­plain con­tent to peo­ple. The fo­cus here is on ac­cessibil­ity. Often the main ideas are already doc­u­mented some­where, but the au­thor thinks that they could do a bet­ter job ex­plain­ing them to their in­tended au­di­ence.

I would cat­e­go­rize some of the re­cent posts on Embed­ded Agency as be­ing ex­plana­tory. Some of them have very nice di­a­grams and are el­e­gantly laid out. I be­lieve much of the con­tent comes from ear­lier work that was a lot more frag­mented and ex­per­i­men­tal. Zhu­keepa’s re­cent overview of Paul Chris­ti­ano’s work also is a good ex­am­ple.


Aca­demic doc­u­ments, as I in­ter­pret them, aim to be ac­cept­able to the aca­demic com­mu­nity or con­sid­ered aca­demic. Some at­tributes that typ­i­cally go along with this in­clude:

  • The aca­demic ar­ti­cle structure

  • Ci­ta­tions, gen­er­ally of other aca­demic works

  • Dis­cus­sion of how work fits in with ex­ist­ing aca­demic literature

  • A high level of rigor and completeness

  • An ex­pec­ta­tion that the main terms and ideas won’t change much

  • PDF formatting

There can definitely be a lot of sig­nal­ing go­ing on here. Many peo­ple see aca­demic seem­ing ar­ti­cles as sub­stan­tially more trust­wor­thy and im­pres­sive than other works.

That said, I feel like there are some use­ful at­tributes to these works be­sides sig­nal­ing. For one, it’s a for­mat well suited to in­ter­fac­ing with the aca­demic world. In­ter­fac­ing with the aca­demic world can be quite valuable, es­pe­cially in do­mains with sub­stan­tial aca­demic work. Also, the for­mat has be­come pop­u­lar for some valid rea­sons around ro­bust­ness and con­text.

As an ex­am­ple, MIRI’s offi­cial pa­pers fit into this cat­e­gory.

Aca­demic-ori­ented posts don’t need to be PDFs. I would con­sider my post on Pre­dic­tion-Aug­mented Eval­u­a­tion Sys­tems to par­tially be in this cat­e­gory, and sev­eral EA Fo­rum posts to par­tially be in this cat­e­gory (ex­am­ples here, here, and here.)

There are some doc­u­ments that do a good job be­ing both “aca­demic” and “ex­plana­tory.” I think these should be con­sid­ered a mix of both.

Fur­ther Thought

I think the main take away of this post is that some doc­u­ments ex­ist for the main pur­pose of clar­ifi­ca­tion, and should be un­der­stood as such. I my­self cur­rently have a lot of ideas I want to write down and in­tend to fo­cus on clar­ifi­ca­tion posts for a while.

The dis­tinc­tion be­tween ex­plana­tory and aca­demic doc­u­ments doesn’t seem as novel nor as el­e­gant to me. I’d be re­ally cu­ri­ous if read­ers can post in the com­ments with im­prove­ments on this on­tol­ogy or bet­ter ex­am­ples.