Is there a social/structural notion of entropy?

Asta [AMP]

Is there a social/structural notion of entropy? As Shannon entropy measures the amount of information that can be passed through a medium, is there a similar notion of the corruptibility of a system or structure? That is, if we assume a system has a single (or set of) mode of how it is ‘supposed’ to work, we could define all alternate modes of the system to be “corrupted” and take the measure of both to determine how many alternate ways the system could be used (and therefore quantify, to an extent, the likelihood of a system corrupting over time)

Consider the U.S. second amendment (regarding the “right to bear arms” and a “well regulated militia”); there’s probably a set of meanings, in the original context, that was intended to convey. But it’s subject to alternate linguistic meanings (both ones that existed before and ones that exist now) that means the amendment can be used to justify a whole slew of things. I assume this type of thing is why legal language is as specific and precise as it is (so that it is either free to be abused or more resilient against it depending on who draft it). If a social system or entity is composed largely of written bylaws, articles of incorporation, internal role descriptions, etc, as well as legal and social context surrounding it, it seems that in theory if one were able to find the alternate interpretations of all the language used in each individual component and then find the total set of these alternate meanings which do not conflict with each other, you would have a measure of how many alternate configurations of the system can exist and therefore could calculate the entropy and use it as a measure of how many possible ways the system could “go wrong”.

d@nny "disc@" mc²

@aud distributed blackness poses a challenge to shannon information as an insufficient metric of communication by limiting to the components chosen to be measured

Asta [AMP]

Obviously this would be incredibly difficult to calculate for large or even moderately sized entities (even with fixing certain things in place, such as the meaning of words that exists currently), which is usually why "keep people with good intentions involved in your shit" is the best policy.

But it might be interesting to calculate and see just how risky your language or entity is: assuming that the language is not ignored or changed (big assumption), how susceptible to abuse is the system you are creating?

d@nny "disc@" mc²

@aud i think that a purely textual analysis of laws elide the power that creates them. judicial review was not in the constitution but arose as an interpretation which has since become accepted

Asta [AMP]

@[email protected] ooooooooooooh I like that

To be fair I don't think Shannon entropy was meant to be anything except about communication/information in a very, very narrow and technical sense, so I'd certainly say that outside of contexts such as "how much information can we shove through this wire" it is definitely the wrong metric.

Asta [AMP]

@[email protected] Right; it's definitely completely unable to account for this kind of stuff. So if anything, it'd be a, shall we say, lower bound on the measure of the thing as it exists under current understanding and in the current context.

So almost certainly not worth the cost of calculating, really. But I wonder if the idea in abstract holds some value (which I think is also a way Shannon entropy contributes, rather than just as a purely mathematical equation).

d@nny "disc@" mc²

@aud law captures power as it exists in a single moment and cauterizes it, like taking a 2d slice of a 3d structure (or a 3d slice of a 4d structure). legal analysis is not just of textual interpretations but about the intent of laws, to admit alternate interpretations. legal scholars like @/[email protected] discuss the intent of e.g. the sherman act and FTC act just as often as (if not more so than) their textual content. the intent which led to these laws is the basis of judicial and regulatory decisions, and is not captured in the text

Asta [AMP]

@[email protected] The thought that gave rise to this was less about the specific laws on the books and more about what type of rules/bylaws/etc a group might wish to have on the book. I think, absent good intentions, it's impossible to build any structure which will defy rotting over time (much as a physical structure is subject to rot), which is where the idea of entropy started to come in. You want something that can endure in the sense of "even if terrible people are in the organization, what kind of measures can we build in that limit their ability to subvert?" In the end, if the entire entity is populated with bad actors, there's nothing to defend against that, I don't think. But constructing a system where you can weather some bad actors in important roles can be made easier or harder depending on what linguistic structures you use to build and confine the entity.

So that's the sort of context I'm thinking of, here.

d@nny "disc@" mc²

@aud governance of an open source project similarly is far more than what's captured in the code of conduct, as the code of conduct is selectively enforced typically by an unwritten alternate code. writing down more of these things tends to make the system more accountable, and often equitable, but this dissonance complicates any purely textual analysis imho

Asta [AMP]

@[email protected] I think it's impossible to write a framework that is "self correcting" absent people acting in good faith because people can (and do) simply ignore written rules. Still; I'm just wondering. Even if you have a bdfl of a project and they are truly benevolent, is there any way to write a system of rules that leads to less abuse after they're gone rather than something that is immediately exploitable?*

*(probably not).

Asta [AMP]

@[email protected] but I think on the flip side, it is probably an easy problem to write a set of rules that is easy to exploit and subvert.

d@nny "disc@" mc²

@aud for the second amendment i think the concept of textual interpretation is actually a dodge and the issue is more around selective enforcement. is the nation a nation of laws? is there a rule of law? we say this but largely don't practice it. the idea that the second amendment admits ambiguity is a work around the fact that there is ambiguity in much of the constitution and it is always selectively enforced, and gun owners tend to hold power in us society and can effectively sway the populace as it aligns with concepts of nationalism and independence that hold sway. i think this should be considered more of the problem than the text itself

Asta [AMP]

@[email protected] (and also, of course, as you pointed out, 'proper enforcement' of a code of conduct often requires basically ignoring the text of the rules and just solving the problem by removing the problematic person, since when they need to people acting in bad faith will stick to the appearance of rules).

This started as a mostly abstract question... and is probably to some degree not even a very good one as reality has no enforcement mechanism for the rules that people write.

Charles U. Farley

@aud I'm not sure if anyone has written about it, but I guess the entropy of an organization would be defined as something like how much additional text you need beyond its constitution to describe its behavior, if you wanted it to correspond in some way to corruption. But that could also have a high value for an organization that had a lot of unwritten tradition.

There are also complexity measures, and I can imagine possible definitions for temperature as well, specifically as a measure of dissatisfaction, leading to "evaporation" and "evaporative cooling".

In tech orgs, technical debt may be the best analog to entropy. I imagine all organizations must have equivalents.

Asta [AMP]

@[email protected]

or the second amendment i think the concept of textual interpretation is actually a dodge and the issue is more around selective enforcement

oh yeah, I totally agree. That is the problem, and indeed much of the problem with writing rules in general, it seems. Definitely gun owners are acting in serious bad faith here and it seems given enough time there's no way they can't torture the language to mean whatever they want it to mean.

which, so, maybe the entropy in reality is "infinite, boundless, and uncaring".

d@nny "disc@" mc²

@aud i think writing down a system of rules and their enforcement which does not depend on assumed benevolence (something the us constitution does do in some respects but not others, as it was written to defend against monarchic and not oligarchic power) is both possible to achieve and maintain

Charles U. Farley

@aud The Second Amendment is a very interesting example, because it's been "obsoleted" by both the 14th amendment and the fact that we now have a large, permanent standing army, large reserves, the national guard, and paramilitary police forces. I'm pretty sure it was originally intended to protect the states from the federal government, and that it didn't really occur to the framers that states would try to disarm their own people because that would be disarming *themselves* since they didn't have their own militaries.

There are those who argue that the 14th Amendment makes the Constitution a completely different document by applying it to the states. So that would have been a step change in the entropy of the country.

Asta [AMP]

@[email protected] I think you're right. I think this thought exercise just made me marginally more of a fan of the central idea of anarchy (no hierarchies!) if anything.

Even if you could measure it at a moment in time, you can't avoid capture of the organization (which means who cares what you wrote down) any more than you can avoid the linguistic, cultural, and legal landscape changing.

Sometimes I think, it would be nice to build something that lasted beyond me, right? But it doesn't matter if I'm corruptible, because even if I am not, I will one day die and not be in charge of it or I will be removed. And I struggle with this because I don't think it's better to not build. You just have to be careful about what you do build. Except you can't possibly see all ends.

d@nny "disc@" mc²

@aud i'm saying that like all metrics it measures only what the measurer is measuring, and the claims of universality associated to shannon information make it an extremely intoxicating viewpoint to misuse—but if we want to be fair, this is not a problem with shannon information but its usage. to some extent there are very meaningful forms of power and intent that aren't written down and run through all forms of language which i think is important

Asta [AMP]

@[email protected] This is definitely one area where I am waaaay outside of my depths regarding the specifics as I am not a legal scholar nor do I pretend to understand the way in which the language involved in the amendment(s) is now applied or commonly believed to apply.

However, now that you bring it up, the fact that there's been such... shifting around of the ideas even when the language hasn't changed (and as you point out, changing the number of possible ... I'm gonna use 'modes' instead of 'states' here to talk about the number of possible configurations that exist depending on how the language is interpreted. But anyway, it definitely changes the number of modes depending on which entities people think the language does or does not apply to).

Maddening.