Seeking a “Seeking Whence ‘Seek Whence’” Sequence

One of the sharpest and most im­por­tant tools in the LessWrong cog­ni­tive toolkit is the idea of go­ing meta, also called seek­ing whence or jump­ing out of the sys­tem, all terms crafted by Dou­glas Hofs­tadter. Though pop­u­larized by Hofs­tadter and re­peat­edly em­pha­sized by Eliezer in posts like “Lost Pur­poses” and “Ta­boo Your Words”, Wikipe­dia in­di­cates that similar ideas have been around in philos­o­phy since at least Anax­i­man­der in the form of the Prin­ci­ple of Suffi­cient Rea­son (PSR). I think it’d be only ap­pro­pri­ate to seek whence this idea of seek­ing whence, tak­ing a his­tory of ideas per­spec­tive. I’d also like analy­ses of where the theme shows up and why it’s ap­peal­ing and so on, since again it seems pretty im­por­tant to LessWrong episte­mol­ogy. Topics that I’d like to see dis­cussed are:

  1. How con­ser­va­tion of prob­a­bil­ity in Bayesian prob­a­bil­ity the­ory and con­ser­va­tion of phase space vol­ume in statis­ti­cal me­chan­ics are re­lated—a sum­mary of Eliezer’s posts on the topic would be great.

  2. How con­ser­va­tion of prob­a­bil­ity &c. are re­lated to other phys­i­cal/​math­e­mat­i­cal laws, e.g. Noether’s the­o­rem and quan­tum me­chan­ics’ con­ti­nu­ity equa­tion.

  3. The his­tory of the idea of con­ser­va­tion laws; whether the dis­cov­ery of con­ser­va­tion laws was fueled by PSR-like philo­soph­i­cal-like con­cerns (e.g. Leib­niz?), by lower level in­tu­itive con­cerns, or other means.

  4. How con­ser­va­tion of prob­a­bil­ity &c. are re­lated to the idea of seek­ing whence [pdf] (e.g., “fol­low the im­prob­a­bil­ity”).

  5. How the PSR re­lates to con­ser­va­tion of prob­a­bil­ity &c. and to seek­ing whence.

  6. How go­ing meta and seek­ing whence are re­lated/​equiv­a­lent.

  7. Which philoso­phers have used some­thing like the PSR (e.g. Spinoza, Leib­niz) and which haven’t; those who haven’t, what their rea­sons were for not us­ing it.

  8. What kinds of con­clu­sions are typ­i­cally reached via the PSR or have his­tor­i­cally been jus­tified by the PSR, and whether those con­clu­sions fit with LW’s stan­dard con­clu­sions. If it dis­agrees with LW’s stan­dard con­clu­sions, where does the PSR not ap­ply or not ap­ply as strongly; al­ter­na­tively, why stan­dard LW con­clu­sions might be mis­taken.

  9. Whether Schopen­hauer’s four-fold di­vi­sion of the PSR makes sense. (Schopen­hauer’s a rel­a­tively LW-friendly con­ti­nen­talesque philoso­pher.) A sum­mary of any crit­i­cisms of his four-fold di­vi­sion.

  10. What makes the PSR, go­ing meta, “JOOTS”-ing and seek­ing whence ap­peal­ing, from a meta­phys­i­cal, episte­molog­i­cal, prag­matic, and psy­cholog­i­cal per­spec­tive. What sorts of en­vi­ron­ments or prob­lem sets se­lect for it. (The Bald­win effect and similar phe­nom­ena might be rele­vant.)

  11. What go­ing meta /​ seek­ing whence looks like at differ­ent lev­els of or­ga­ni­za­tion; how one jumps out of sys­tems at vary­ing lev­els.

  12. Eliezer’s rule of deriva­tive val­idity from CFAI and how it re­lates to the PSR; an anal­y­sis of how the (moral, or per­haps UDT-like de­ci­sion-policy-cen­tric) PSR might be rele­vant to Friendli­ness philos­o­phy, e.g. as com­pared with CEV-like pro­pos­als [pdf].

  13. How la­tent Pla­tonic nodes in TDT [pdf] (p. 78) re­late to the PSR.

  14. A gen­er­al­iza­tion of CFAI’s causal val­idity se­man­tics to time­less val­idity se­man­tics in the spirit of the gen­er­al­iza­tion of CDT to TDT, or per­haps even fur­ther gen­er­al­iza­tions of causal val­idity se­man­tics in the spirit of Up­date­less De­ci­sion The­ory or eX­cep­tion­less De­ci­sion The­ory. (ETA: Whoops, Eliezer already dis­cussed the acausal level, but seems to have only men­tioned Pla­tonic forms as an af­terthought. Maybe ig­nore this bul­let point.)

  15. How the PSR and the rule of deriva­tive val­idity re­late to Robin Han­son’s idea of pre-ra­tio­nal­ity and Wei Dai’s ques­tions about ex­tend­ing pre-ra­tio­nal­ity to in­clude past selves’ util­ity func­tions—whether this elu­ci­dates the re­la­tion be­tween XDT and UDT.

  16. Where Hofs­tadter picked up the idea of “go­ing meta” and what led him to think it was im­por­tant. What led Eliezer to rely on it so much and em­pha­size the im­por­tance of avoid­ing lost pur­poses.

Does any­one have the math­e­mat­i­cal, his­tor­i­cal, philo­soph­i­cal, and re­search skills nec­es­sary to write a very long post or two or a se­quence on this? How about just the skills re­quired to tackle one of the above ques­tions, or a ques­tion like those?
Might peo­ple with rele­vant knowl­edge share it in com­ments on this post? I’ll try to write a few com­ments to seed dis­cus­sion of some of the above ques­tions.
(Maybe best dis­cussed el­se­where, but: What other cog­ni­tive tools or themes of­ten used on LessWrong have a long his­tory that we don’t all know even ex­ists?)
I’d also just gen­er­ally ap­pre­ci­ate posts on go­ing meta, es­pe­cially with re­gards to epistemic ra­tio­nal­ity—dis­cus­sion of “go­ing meta” in the do­main of in­stru­men­tal ra­tio­nal­ity tends to de­volve into bor­ing ‘ex­per­i­ment more!’ ‘no, go meta more!’ back-and-forths with­out any ac­com­pa­ny­ing ex­plicit VOI calcu­la­tions. With epistemic ra­tio­nal­ity the benefits are clearer cut. When peo­ple re­flect and try to go meta dis­cus­sions tend to go way bet­ter—e.g., when peo­ple are ex­plic­itly aware that “poli­tics is the mind-kil­ler” they’re less likely to start a dis­tract­ing flame war, and more likely to dis­cuss in­ter­est­ing fac­tual is­sues. (Please, let’s avoid dis­cus­sion of “poli­tics is the mind kil­ler” poli­tics here. ;P ) I think this sort of effect could be in­creased across the board, es­pe­cially for mildly poli­ti­cal sub­jects like whether to donate to x-risk char­i­ties, and that this would in­crease the qual­ity of most ar­gu­ments on LessWrong. Although much of the rele­vant knowl­edge needed to go meta there is knowl­edge of sig­nal­ing game the­ory and so­cial psy­chol­ogy, and can’t be re­placed sim­ply by ap­ply­ing a skill like ‘go meta’, go­ing meta and stay­ing re­flec­tive can still miti­gate many ob­vi­ous failure modes. Posts giv­ing ex­am­ples of (hy­po­thet­i­cal) de­bates and how the de­baters could go meta could be re­ally valuable.
Thanks all!