Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Actually computing equilibria in even toy mathematical models of social collaboration is horribly intractible even at a single time step. So what you wind up being best off doing is randomly choose your initial location and just follow the local gradient of fun, then iterating this a few times and choosing the best one.


Nobody is suggesting doing game-theoretic analyses before engaging in social interaction. Obviously someone who thinks that they can completely derive their actions from formal rules of behavior is doomed to fail, but it's not like the only two possible choices are "behave completely randomly" and "perform a game-theoretic analysis of the situation based on formal rules". Guides like this provide a good set of heuristics and help the socially unskilled people understand the general ideas and general concepts which socially skilled people seem to just pick up from their environment. This is important so you can pick a reasonable starting location before doing your hill climbing algorithm -- if you just choose a completely random starting point, everything is just extremely discouraging, there's no useful feedback from the environment (other than "YOU FAIL"), and there's absolutely no fun gradient to maximize. (Or at least, the fun gradient is so negative in all directions that it's difficult to figure out which way is up.)

It always amazes me how socially skilled people apparently can't even understand what it means to be socially unskilled. It would be as if everyone who learned to ride a bicycle later completely forgot that they had to learn this skill, then just went around telling everyone who didn't know how to ride a bicycle, "you just get on and pedal, it's so easy, I don't see why you keep falling, you must not be trying".


Bumbling around by semi-randomly trying behaviors learned from pop culture and from other kinds of interactions, discovering empirically what works in what contexts, is actually the normal way to do it. Most people start a lot younger, though. If you make the same mistakes at 18 that other kids did at 13, people aren't going to be very impressed, just like you wouldn't be impressed by a college student taking remedial algebra classes.

The good news is that flirting is a lot like math: most people don't learn any new concepts after the first two years of college. Once you've caught up to that point, it's just a matter of getting from the point where you understand the concepts, but have to think very hard while applying them, to the point where you apply them automatically without thinking about them. It's the difference between understanding the principles required to solve x^2 + x - 6 = 0 but having to think your way through each step of the solution, vs. just looking at the problem and letting your mind produce the solution out of habit.


regarding the first part, you're absolutely right, but understanding the worst case computational complexity of a problem does lend a certain appreciation to the average case complexity that happens when people implicitly approximately solve those problems.

Admission: My current research project involves studying the computational complexity of socially inspired optimization problems. Turns out that even the simplest nontrivial ones are PPAD Hard or #P hard. (so special cases or approximations become key very quickly)

This is why its useful to put social problems into at least semi-mathematical language to make it easier to spell things out to the socially oblivious (but still logically endowed, which sadly isn't always the case) or to make it easier to discuss a complex social issue with a friend


This reminds me of an article I read years ago about cognitive psychologists trying to model how outfielders catch fly balls. Evidently, even given the massively parallel nature of the brain, it was very hard to come up with a model that was effective enough to explain the performance of real outfielders (Little League to Major League) but was simple enough to be neurologically plausible. Just like the social models you describe, none of the mathematically straightforward ways of modeling the fly ball problem were cognitively feasible. Yet, if you were teaching an engineering student to catch a fly ball, they would probably be helpful, even though the final goal would be to induce an entirely unrelated cognitive structure in the engineer's mind.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: