Boundaries enable positive material-informational feedback loops

(Cross-posted from my blog)

[epistemic sta­tus: ob­vi­ous once con­sid­ered, I think]

If you want to get big things done, you al­most cer­tainly need pos­i­tive feed­back loops. Un­less you can already do all the nec­es­sary things, you need to do/​make things that al­low you to do/​make more things in the fu­ture. This dy­namic can be found in RPG and econ­omy-man­age­ment games, and in some ac­tual eco­nomic sys­tems, such as in­dus­tri­al­iz­ing economies.

Ma­te­rial, in­for­ma­tion, and economy

Some goods that can be used in a pos­i­tive feed­back loop, such as soft­ware and in­ven­tions, are in­for­ma­tional. Once pro­duced, they can be used in­definitely in the fu­ture. In eco­nomic terms, they are non­ri­val­ous.

Other goods are ma­te­rial, such as man­u­fac­tured goods and en­ergy. They can’t be copied cheaply. In eco­nomic terms, they are ri­valrous.

In prac­tice, any long-last­ing pos­i­tive feed­back loop con­tains both in­for­ma­tional and ma­te­rial goods, as pro­duc­tion of in­for­ma­tion re­quires a phys­i­cal sub­strate. While en­sur­ing that in­for­ma­tional goods can be used in the fu­ture is an or­ga­ni­za­tion and com­mu­ni­ca­tion prob­lem (a sub­ject be­yond the scope of this post), the prob­lem of en­sur­ing that ma­te­rial goods can be used in the fu­ture is ad­di­tion­ally a se­cu­rity prob­lem.

An im­por­tant ques­tion to ask is: why haven’t ma­te­rial-in­for­ma­tional pos­i­tive feed­back loops already taken over the world? Why don’t we have so much stuff by now that pro­vid­ing for peo­ple’s ma­te­rial needs (such as food and hous­ing) is triv­ial?

To some ex­tent, ma­te­rial-in­for­ma­tional pos­i­tive feed­back loops have taken over the world, but they seem much slower than one would naively ex­pect. See cost dis­ease. As an ex­am­ple of cost dis­ease, the av­er­age cost of a new house in the USA has quadru­pled over a 60-year pe­riod (ad­justed for in­fla­tion!), whereas mod­els of cap­i­tal­ism based on econ­omy-man­age­ment games such as Fac­to­rio (or, more aca­dem­i­cally, ac­cord­ing to the la­bor/​cap­i­tal based eco­nomic mod­els of clas­si­cal economists such as David Ri­cardo) would sug­gest that houses would be plen­tiful by now. (And no, this isn’t just be­cause of land prices; it costs about $300K to build a house in the US in 2018)

Se­cu­rity and boundaries

I’ve already kind of an­swered this ques­tion by say­ing that en­sur­ing that ma­te­rial goods can be used in the fu­ture is a se­cu­rity prob­lem. If you use one of your ma­te­rial goods to pro­duce an­other ma­te­rial good, and some­one takes this new good, then you can’t put this good back into your pro­duc­tion pro­cess. Thus, what would have been a pos­i­tive feed­back loop is in­stead a nega­tive feed­back loop, as it leaks goods faster than it pro­duces them.

Solv­ing se­cu­rity is­sues gen­er­ally re­quires bound­aries. You need to draw a bound­ary in ma­te­rial space some­where, differ­en­ti­at­ing the in­side from the out­side, such that ma­te­rial goods (such as en­ergy) on the in­side don’t leak out, and can po­ten­tially have pos­i­tive feed­back loops. There are many ways to pre­vent leaks across a bound­ary while still al­low­ing in­for­ma­tional and ma­te­rial to pass through some­times, such as semi­porous phys­i­cal bar­ri­ers and ac­tive polic­ing. Re­gard­less of the method to en­force the bound­ary, the bound­ary has to ex­ist in some ge­o­met­ri­cal sense for it to make sense to say that e.g. en­ergy in­creases within this sys­tem.

Not all se­cu­rity is­sues are from other agents; some are from non-agen­tic pro­cesses. Con­sider a home­o­static an­i­mal. If the an­i­mal ex­pends en­ergy to warm its body, and this warmth es­capes, the an­i­mal will fail to re­al­ize gains from the en­ergy ex­pen­di­ture. Thus, the an­i­mal has a bound­ary (namely, skin) to solve this “se­cu­rity prob­lem”. The cold air par­ti­cles that take away heat from the an­i­mal are analo­gous to agents that di­rectly take re­sources, though ob­vi­ously less agen­tic. While per­haps my us­age of the word “se­cu­rity” to in­clude re­sponses to nona­gen­tic threats is non­stan­dard, I hope it is clear that these are on the same spec­trum as agen­tic threats, and can be dealt with in some of the same ways.

It is also worth think­ing about semi-agen­tic en­tities, such as microor­ganisms. One of the biggest threats to a food store is microor­ganisms (i.e. rot­ting), and slow­ing the nega­tive feed­back loops de­plet­ing food stores re­quires solv­ing this se­cu­rity prob­lem us­ing a bound­ary (such as a sealed con­tainer or a sub­set of the air that is colder than the out­side air, such as in a re­friger­a­tor).

Prop­erty rights are a sim­ple ex­am­ple of bound­aries. Cer­tain goods are con­sid­ered to be “owned” by differ­ent par­ties, such that there is com­mon agree­ment about who owns what, and peo­ple are for one rea­son or an­other not mo­ti­vated to take other peo­ple’s stuff. Such di­vi­sion of goods into sets owned by differ­ent par­ties is a set of bound­aries en­abling pos­i­tive feed­back loops, which are es­pe­cially salient in cap­i­tal­ism.

What about trust be­tween differ­ent en­tities? A com­plex ecosys­tem will con­tain en­tities satis­fy­ing a va­ri­ety of niches, which in­clude par­a­sitism and pre­da­tion (which are on the same spec­trum). A trust net­work can be thought of as a way for differ­ent en­tities to draw var­i­ous bound­aries, of­ten fuzzy ones, that mostly ex­clude par­a­sites/​preda­tors, such that there are few leaks from in­side this bound­ary to out­side this bound­ary (which would in­clude par­a­sitism/​pre­da­tion by en­tities out­side the bound­ary). There are “those who you trust” and “those who you don’t trust” (both fuzzy sets), and you as­sign more util­ity to giv­ing re­sources to those you trust, as this al­lows for pos­i­tive feed­back loops within a sys­tem that con­tains you (namely, the trust net­work).

Ex­ter­nal­ities and sustainability

Since no sub­sys­tem of the world is causally closed, all pos­i­tive feed­back loops have ex­ter­nal­ities. By defi­ni­tion, the out­side world is only di­rectly af­fected by these ex­ter­nal­ities, and is only af­fected by what hap­pens within the bound­ary to the ex­tent that this even­tu­ally leads to ex­ter­nal­ities. A wise de­signer of a pos­i­tive feed­back loop will an­ti­ci­pate its ex­ter­nal­ities, and set it up such that the ex­ter­nal­ities are over­all de­sir­able to the de­signer. After all, there is no point to cre­at­ing a pos­i­tive feed­back loop un­less its ex­ter­nal­ities are mostly pos­i­tive [EDIT: un­less the bound­ary con­tains things that have in­trin­sic value].

A pos­i­tive feed­back loop’s ex­ter­nal­ities mod­ify its en­vi­ron­ment, af­fect­ing its own abil­ity to con­tinue; for ex­am­ple, a pos­i­tive feed­back loop of microor­ganisms eat­ing food will ex­haust it­self by con­sum­ing the food. So, differ­ent pos­i­tive feed­back loops are en­vi­ron­men­tally sus­tain­able to differ­ent ex­tents. Both pro­duc­tion and con­quest gen­er­ate pos­i­tive feed­back loops, as Ben Hoff­man dis­cusses in this post, but pro­duc­tion is much more en­vi­ron­men­tally sus­tain­able than con­quest.

One way to in­crease en­vi­ron­men­tal sus­tain­abil­ity is to move more pro­cesses to the in­side of the bound­ary. For ex­am­ple, a coun­try that is con­sum­ing large amounts of iron (driv­ing up iron prices) may con­sider set­ting up its own iron mines. Thus, the in­side of the bound­ary be­comes more like an econ­omy of its own. This is some­times known as im­port re­place­ment.

Of course, the en­vi­ron­men­tal sus­tain­abil­ity of a pos­i­tive feed­back loop can also be a nega­tive, as it is bet­ter for some pro­cesses (such as rot­ting) to limit or ex­haust them­selves, thus tran­si­tion­ing to nega­tive feed­back or a com­bi­na­tion of pos­i­tive and nega­tive feed­back. Pro­cesses that in­clude in­ten­tion­ally-de­signed pos­i­tive and nega­tive feed­back can be much more en­vi­ron­men­tally sus­tain­able than pro­cesses that only have pos­i­tive feed­back loops de­signed in, since they can limit their growth when such growth would be un­sus­tain­able.

While in the­ory the philos­o­phy of effec­tive al­tru­ism would im­ply a strong (and likely over­whelming) em­pha­sis on cre­at­ing and main­tain­ing en­vi­ron­men­tally sus­tain­able pos­i­tive feed­back loops with pos­i­tive ex­ter­nal­ities, typ­i­cally-recom­mended EA prac­tices (such as giv­ing away 10% of one’s in­come) are nega­tive feed­back loops (the more you make, the more you give away). While in the­ory the place the re­sources are given to could have a faster pos­i­tive feed­back loop than just in­vest­ing in your­self, your friends, and your pro­jects, in prac­tice I rarely be­lieve claims of this form that come from the EA move­ment; for ex­am­ple, if a coun­try has a high rate of poverty, that in­di­cates that the nega­tive feed­back loops (such as cor­rup­tion) are likely stronger than the pos­i­tive ones, and that giv­ing re­sources is in­effec­tive. Thus, I can­not in good con­science al­low any­thing like cur­rent EA ide­ol­ogy to sub­stan­tially con­trol re­source al­lo­ca­tion in most sys­tems I cre­ate, even though EA philos­o­phy taken to its log­i­cal con­clu­sion would get the right an­swer on the im­por­tance of bound­aries and pos­i­tive feed­back loops.

Policy suggestions

How do these ideas trans­late to ac­tion? One sug­ges­tion is that, if you are try­ing to do some­thing big, you use one or more pos­i­tive feed­back loops, and ask your­self the fol­low­ing ques­tions about each one:

  1. What’s the gen­er­a­tor of my pos­i­tive feed­back loop (i.e. what’s the pro­cess that turns stuff into more stuff)?

  2. What is the bound­ary within which the pos­i­tive feed­back in­creases re­sources?

  3. How am I re­duc­ing leak­age across this bound­ary?

  4. What are the ex­ter­nal­ities of this pos­i­tive feed­back loop?

  5. How en­vi­ron­men­tally sus­tain­able is this pos­i­tive feed­back loop?

  6. Are there built-in nega­tive feed­back loops that in­crease en­vi­ron­men­tal sus­tain­abil­ity?


(thanks to Bryce Hidy­smith for a con­ver­sa­tion that led to this post)