Absurd rules and social order
echoweaver 's parents have been visiting this week. During a dinner conversation, one of them mentioned a point that has always bothered me:
"95% of the rules exist for 5% of the people"
I'm sure we've all run into this and loathe it equally -- why should many people be punished for the bad behavior of a few bad actors? At the largest scale, this produces the kind of labyrinthine legal systems that all first world nations live under (to my knowledge, anyway). On a more day-to-day basis, we get the kinds of stupid bureaucratic piles of regulations that wear us all down -- account your time to 15 minute increments, don't walk on the grass, sit through seventeen ethics training modules, byzantine hiring and firing rules, etc. etc. etc. (Not to mention the annual dread of tax season, with its byzantine, inscrutable rules and absurd 161 page instruction manual.)
I hate this shit. I hate having my life curtailed because some other asshole didn't feel like playing nice with the rest of us. I hate being subject to lawyers and accountants because our legal and financial systems are too complicated for anybody but a lifelong, dedicated expert to understand. I hate getting slapped with a new "you can't do that" every time some moron comes up with a creative new way to fuck the rest of us over.
So what I've been wondering is whether there could be some way to structure society so that we don't have to have to have complicated rule sets and yet we still enforce mostly good behavior?
The problem, as I see it, is that we currently govern entirely by rule sets. Essentially, every time someone does something evil or stupid, we introduce another exception case that says, "No, you can't do that thing either, dumbass." The result is that our legal system comes out looking like a program written by a freshman -- it's a huge, nested set of "if" and "case" statements. And, as hackers all know, such structures are insanely hard to debug, test, verify, validate, or modify. Thus, we spend a lot of expense and effort developing and maintaining a set of hyper-specialized "programmers" to maintain this one code base (we call them lawyers and judges).
So what are the alternatives? Clearly, we can't do without some kind of social regulation mechanism. Anarchy just doesn't work. (Robert Heinlein's wet dreams notwithstanding.)
Well, one possibility is, of course, refactoring -- trying to reduce a complicated rule set to a minimal, effective rule set. Remove nested and redundant ifs, and so on. I don't think that's the answer here. It would, of course, involve an insanely large effort to do (not to mention probably being EXP-complete or NP-insane or something). But this is all a thought exercise anyway, so I'm not worried about that for the moment. No, I'd really just rather do away with the majority of the rule set to begin with. (I think that we may end up needing some, as I'll point out in a moment, but I'd like to minimize the ones that we need.)
So what are the alternatives to rules?
In my observation, there are only two ways to regulate social behavior outside of explicit, authority-enforced rules:
So could we do something with incentive systems to ensure good social behavior? What we want is something like, "We'll pay you to be nice and bankrupt you for being naughty". Note that this is not at all the same proposal as the laissez-faire assertion that "free markets will solve everything" -- here we're explicitly planning to muck with the market by offering incentives and penalties that aren't there "naturally". An example would be something like, "yes, you're free to hire anybody you want, but the cost to hire someone younger than 18 goes up (at least) exponentially quickly the younger they are -- if you really think that it's worth, say, 1024* the hourly rate of an 18 year old to hire an 8 year old, go right ahead."
But could we really get what we want out of such a system? At least partially, I think so.
If you believe modern economic theory, a la Von Neumann and Morgenstern, then humans are utility-maximizing agents. (While I don't completely buy that story, I think it's not a bad first approximation.) Anyway, in this view, humans are constantly trying to solve a huge, complicated constrained optimization problem: the objective function is their general life satisfaction (of which money is one component, and the easiest for us to twiddle socially) and the constraint set is provided by the hard rules imposed by governments, companies, religions, etc. So one thing we can do is convert hard rules into incentives by compiling constraints into the objective function (for those geeky enough to care about how, the KKT conditions guarantee this, at least for the classes of optimization problems that I think we need here). So it should be possible to compile rules of the form "thou shalt not do X" into "we'll penalize you a lot of $$$ for doing X" (or, perhaps, "we'll pay you for doing not X").
So this would get us one thing, at least -- it would get rid of a bunch of annoying "thou shalt not" kinds of absolute rules, replacing them with a more flexible system in which people could make their own decisions about what's in their best interest. Really want more than 2 weeks of vacation a year? Fine, go ahead and take more, but at some economic penalty (possibly a superlinear penalty). Want to come in late to work? Sure, fine, but it could be expensive. Want to walk on the grass? That's ok too -- it'll just be $0.25/step or something. Want to make your workers do more than 40 hrs/week? Great, but their time gets more expensive as you go. (I still don't know why we "exempted" white collar workers from fair treatment -- if construction crews get time-and-a-half, why don't hackers?) Note that this doesn't have to be centralized -- many of these things can be monitored and accounted locally, rather than through a government.
So this gets us some benefits. But it's far from ideal. For one thing, it still requires people to remember a bunch of conditions -- they're just represented as economic algebraic conditions rather than hard rules. We still have the labyrinthine code problem and the "every time a moron breaks it, we need a new condition" problem. And we still need experts to write them all and gobs of monitoring and accounting to enforce them. Not to mention the question of the economic flows -- whether we'd be taking in enough money in penalties to pay for incentives and so on. A bigger problem is that there are plenty of activities that are incommensurable with money (that is, there are objects or actions whose value cannot be expressed in dollars) -- the classical, "what is the value of a human life?" connundrum. (Though it turns out that the latter can actually be quantified.) A bigger problem is that utilities are both nonlinear and individual specific, so it's hard to write an economic rule set that applies to both the beggar and Bill Gates. (We could do some calibration, by making it 10 billion times more expensive for Bill to walk on the grass than the beggar, but it doesn't really scale linearly.)
But it does feel like progress to me, and it may be that there would be ways to solve some of the other issues. For one, having represented things algebraically, it might become clearer which conditions are redundant and could be eliminated.
My bigger worry is what to do about the creative new moron problem. The core difficulty with economic incentives is that, by design, everybody is always trying to, well, game the system. Someone is always gonna get ahead by finding some crack in the system. Typically, such situations lead to tragedies of the commons, ultimately producing drastically anti-social behaviors (like, say, fucking over the entire world economy). So can we design the incentive system to discourage stupid shit that we haven't even thought of, without discouraging creative and good stuff? I think maybe. I conjecture that at least some of these cases arise from corner-cases, in which there's a narrow crevice in between rules. Another thing we could get by moving things to the incentive system is that we could (potentially) smooth off the corners of the constraint polytope by imposing a regularity condition on the objective function.
Hmmm... Yeah, this is going to bear more thinking about. This is mostly in the way of brainstorming -- there are still many open kinks to work out. But it's an interesting thought experiment at the very least...
Comments welcome.
"95% of the rules exist for 5% of the people"
I'm sure we've all run into this and loathe it equally -- why should many people be punished for the bad behavior of a few bad actors? At the largest scale, this produces the kind of labyrinthine legal systems that all first world nations live under (to my knowledge, anyway). On a more day-to-day basis, we get the kinds of stupid bureaucratic piles of regulations that wear us all down -- account your time to 15 minute increments, don't walk on the grass, sit through seventeen ethics training modules, byzantine hiring and firing rules, etc. etc. etc. (Not to mention the annual dread of tax season, with its byzantine, inscrutable rules and absurd 161 page instruction manual.)
I hate this shit. I hate having my life curtailed because some other asshole didn't feel like playing nice with the rest of us. I hate being subject to lawyers and accountants because our legal and financial systems are too complicated for anybody but a lifelong, dedicated expert to understand. I hate getting slapped with a new "you can't do that" every time some moron comes up with a creative new way to fuck the rest of us over.
So what I've been wondering is whether there could be some way to structure society so that we don't have to have to have complicated rule sets and yet we still enforce mostly good behavior?
The problem, as I see it, is that we currently govern entirely by rule sets. Essentially, every time someone does something evil or stupid, we introduce another exception case that says, "No, you can't do that thing either, dumbass." The result is that our legal system comes out looking like a program written by a freshman -- it's a huge, nested set of "if" and "case" statements. And, as hackers all know, such structures are insanely hard to debug, test, verify, validate, or modify. Thus, we spend a lot of expense and effort developing and maintaining a set of hyper-specialized "programmers" to maintain this one code base (we call them lawyers and judges).
So what are the alternatives? Clearly, we can't do without some kind of social regulation mechanism. Anarchy just doesn't work. (Robert Heinlein's wet dreams notwithstanding.)
Well, one possibility is, of course, refactoring -- trying to reduce a complicated rule set to a minimal, effective rule set. Remove nested and redundant ifs, and so on. I don't think that's the answer here. It would, of course, involve an insanely large effort to do (not to mention probably being EXP-complete or NP-insane or something). But this is all a thought exercise anyway, so I'm not worried about that for the moment. No, I'd really just rather do away with the majority of the rule set to begin with. (I think that we may end up needing some, as I'll point out in a moment, but I'd like to minimize the ones that we need.)
So what are the alternatives to rules?
In my observation, there are only two ways to regulate social behavior outside of explicit, authority-enforced rules:
- Social mores: Don't do this thing because your neigbors will shun you and the cool kids will laugh at you if you do it. Conversely, you'll win friends, influence people, and get lots of sex if you do the good thing.
- Incentive systems: You'll make more money or get more swag if you act in way A and you'll make less or lose money/swag if you act counter to it.
So could we do something with incentive systems to ensure good social behavior? What we want is something like, "We'll pay you to be nice and bankrupt you for being naughty". Note that this is not at all the same proposal as the laissez-faire assertion that "free markets will solve everything" -- here we're explicitly planning to muck with the market by offering incentives and penalties that aren't there "naturally". An example would be something like, "yes, you're free to hire anybody you want, but the cost to hire someone younger than 18 goes up (at least) exponentially quickly the younger they are -- if you really think that it's worth, say, 1024* the hourly rate of an 18 year old to hire an 8 year old, go right ahead."
But could we really get what we want out of such a system? At least partially, I think so.
If you believe modern economic theory, a la Von Neumann and Morgenstern, then humans are utility-maximizing agents. (While I don't completely buy that story, I think it's not a bad first approximation.) Anyway, in this view, humans are constantly trying to solve a huge, complicated constrained optimization problem: the objective function is their general life satisfaction (of which money is one component, and the easiest for us to twiddle socially) and the constraint set is provided by the hard rules imposed by governments, companies, religions, etc. So one thing we can do is convert hard rules into incentives by compiling constraints into the objective function (for those geeky enough to care about how, the KKT conditions guarantee this, at least for the classes of optimization problems that I think we need here). So it should be possible to compile rules of the form "thou shalt not do X" into "we'll penalize you a lot of $$$ for doing X" (or, perhaps, "we'll pay you for doing not X").
So this would get us one thing, at least -- it would get rid of a bunch of annoying "thou shalt not" kinds of absolute rules, replacing them with a more flexible system in which people could make their own decisions about what's in their best interest. Really want more than 2 weeks of vacation a year? Fine, go ahead and take more, but at some economic penalty (possibly a superlinear penalty). Want to come in late to work? Sure, fine, but it could be expensive. Want to walk on the grass? That's ok too -- it'll just be $0.25/step or something. Want to make your workers do more than 40 hrs/week? Great, but their time gets more expensive as you go. (I still don't know why we "exempted" white collar workers from fair treatment -- if construction crews get time-and-a-half, why don't hackers?) Note that this doesn't have to be centralized -- many of these things can be monitored and accounted locally, rather than through a government.
So this gets us some benefits. But it's far from ideal. For one thing, it still requires people to remember a bunch of conditions -- they're just represented as economic algebraic conditions rather than hard rules. We still have the labyrinthine code problem and the "every time a moron breaks it, we need a new condition" problem. And we still need experts to write them all and gobs of monitoring and accounting to enforce them. Not to mention the question of the economic flows -- whether we'd be taking in enough money in penalties to pay for incentives and so on. A bigger problem is that there are plenty of activities that are incommensurable with money (that is, there are objects or actions whose value cannot be expressed in dollars) -- the classical, "what is the value of a human life?" connundrum. (Though it turns out that the latter can actually be quantified.) A bigger problem is that utilities are both nonlinear and individual specific, so it's hard to write an economic rule set that applies to both the beggar and Bill Gates. (We could do some calibration, by making it 10 billion times more expensive for Bill to walk on the grass than the beggar, but it doesn't really scale linearly.)
But it does feel like progress to me, and it may be that there would be ways to solve some of the other issues. For one, having represented things algebraically, it might become clearer which conditions are redundant and could be eliminated.
My bigger worry is what to do about the creative new moron problem. The core difficulty with economic incentives is that, by design, everybody is always trying to, well, game the system. Someone is always gonna get ahead by finding some crack in the system. Typically, such situations lead to tragedies of the commons, ultimately producing drastically anti-social behaviors (like, say, fucking over the entire world economy). So can we design the incentive system to discourage stupid shit that we haven't even thought of, without discouraging creative and good stuff? I think maybe. I conjecture that at least some of these cases arise from corner-cases, in which there's a narrow crevice in between rules. Another thing we could get by moving things to the incentive system is that we could (potentially) smooth off the corners of the constraint polytope by imposing a regularity condition on the objective function.
Hmmm... Yeah, this is going to bear more thinking about. This is mostly in the way of brainstorming -- there are still many open kinks to work out. But it's an interesting thought experiment at the very least...
Comments welcome.