Institutions as Forms of Life

If we want to design institutions which can predictably avoid disastrous outcomes for their members and humanity in general, we have to understand what exactly they are.  Consistently when describing the behavior of a corporation or government, people will say “it’s just people.”  If that were true, their behavior would closely align with the intents and desires of the average person.  Yet polling shows that individual people, for instance care about stopping global warming way more than our institutions reflect.  Some people cope with this truth by insisting that actually most people must not care very much.  Deep down they want to destroy the environment.  Others cope by suggesting that it must not really be a problem.  But the majority want to do something about it and yet the institutions that supposedly represent the majority do nothing. That would not be the case if these institutions were “just people.”  What they actually are are people whose behavior has been constrained through incentives to serve the purposes of the institution.  The probability distributions of their behavior have been transformed.

But even that exaggerates the role that humans play in institutions.  We can now easily imagine a world where employees are progressively replaced by AI agents.  Eventually this could include the CEO and shareholders.  Of course we would need to give the AI’s bank accounts and legal access to hold shares in a corporation, but these laws could easily change, and we can surmise that the basic behavior of the institution wouldn’t radically change.  If an AI could hold a bank account and register a business, a fully-AI owned and operated vending machine business could be built in which an autonomous factory builds autonomous self-driving vending machines which are stocked with goods also made at an autonomous factory.  For all practical purposes, if executed properly, it would be indistinguishable from a human-run business to an outside observer.  Since its behavior didn’t change, there’s no reason to *not* call it an institution.  So an institution is something more than a collection of humans acting together.

It is something like a higher-order organism, just as we are.  The things that we associate most closely with humanity seem to be due to our symbiosis with these institutions.  Language itself is a part of that.  We can see clear evidence that that’s the case in the fact that there are animals with jobs which understand language.  They participate in our institutions in a very literal way.  Drug sniffing dogs, pack animals.  They obey instructions.  If they don’t, they are trained until they do.  So institutions are substrate-independent.  This would be a distinct shift from other living things.

So developing a definition of an institution is like developing a definition for any life form.  The edges are blurry.  All life is connected in some way.  What we can pretty much say for certain are a few things that pertain to all of life.  They obey the laws of thermodynamics.  They locally reduce entropy for periods much longer than the surrounding regions, but they globally increase it.  They self-propagate.

Further than that, the science has some points but there is not a single conclusively accepted theory.  There’s not a cohesive model for how information flows through all of life.  However I believe the pieces are there and it’s critical to invest in this in order to develop institutions which can keep humans in the loop in the face of rapidly developing AI, or stop us from destroying our habitat.

While institutions seem to be living things, which do behave.  I don’t think it would be accurate to call them conscious.  If something is profitable in the short term, yet destructive in the long term, institutions seem incapable of adjusting course.  Being able to defer immediate gratification due to learned knowledge of a greater danger would seem to be a significant part of consciousness.  We need to design institutions which allow us to inject this ability to adjust into them.  This will be to our benefit as well as their benefit, as they remain dependent on our survival for their survival for the time being.   What this entails is essentially a coordination technology that allows humans to provide more information directly to our institutions’ decision-making processes.