Game Theory Is Counterintuitive • Hacker News Board

Re: Game Theory Is Counterintuitive

#21

Post by marketforlemmas » Tue, May 27, 2014, 2:31 AM UTC

wspaniel wrote: post #17

curiouslearn wrote: post #10

Earlier quoted context omitted.

No, it does not say how people should behave? It analyzes strategic situations using various equilibrium concepts, such as the Nash equilibrium. If you are not sure that your opponent will play the Nash equilibrium strategy, Game theory doesn't tell you what you should do.

That's when you use Bayesian Nash equilibrium or perfect Bayesian equilibrium. No big deal.

I take some issue with the idea that we can simply just rely on another equilibrium refinement and say "no big deal".

It seems a bit silly to observe some behavior in a game and say "See that's a Nash Equilibrium, so game theory works", and then to turn around and observe some non-NE behavior and say "well in this game, BNE is clearly the right model, so game theory still works". And then yet again to observe some more behavior that conflicts with the theory (or to get rid of silly equilibrium) and say "ah, now we simply use perfect Bayes" or trembling hand equilibrium, or actually we were totally using correlated equilibrium this whole time.

In any case, it feels weird that a theory should behave like the "No true Scotsman" fallacy. We can always get the equilibrium by simply redefining what we mean by equilibrium.

Re: Game Theory Is Counterintuitive

#22

Post by gte910h » Tue, May 27, 2014, 2:51 AM UTC

YokoZar wrote: post #14

foxhill wrote: post #6

Earlier quoted context omitted.

well, in his defence - > Every voting system is manipulable

This is actually the least defensible claim there. I posted this in the comments: "Every Voting System is Manipulable" is an exaggeration of the implications of strategic voting. Yes, the Gibbard–Satterthwaite theorem proves that a dishonest vote could profitably exist in any system, however there are already systems where intentionally doing so requires the voter to: 1) Have perfect (or near-perfect) information abo…

Could you go over the good solutions then? I have always heard there are about 4-5 "good but complicated" systems, none of which are easily shown to be best due to strategic voting and haven't seen clear "we should do this one because it beats all those for sure" type recommendations.

Re: Game Theory Is Counterintuitive

#23

Post by YokoZar » Tue, May 27, 2014, 5:06 AM UTC

hcarvalhoalves wrote: post #18

YokoZar wrote: post #15

Earlier quoted context omitted.

Doesn't some of this behavior come from the rather preposterous way simcity models its drivers? A model where there is no persistent data and the heuristic is instead "go to nearest house and declare that home, if someone beat me there go to the next nearest house" naturally lends itself to a tree model, as no one actually wants to go anywhere but the next link in the chain! A more reasonable model (and why gridded s…

SimCity 4 "Rush Hour" introduces a refined traffic simulation where you can inspect the commute of each citizen and it's exactly what you would expect in real life - they go to anywhere in the city, not only nearby. Grid layouts don't optimize for heavy traffic [1] because drivers choose shorter paths, making it worse for all [2]. In real life you don't have grid layouts without large roads, one-way streets, overpass…

Thanks, this is all very interesting. I do think there's a bit more to consider though, such as how it's much harder to design walkably compact neighborhoods with arterial roadways than with gridded ones, meaning you might just get fewer cars in the gridded setup.

Re: Game Theory Is Counterintuitive

#24

Post by hcarvalhoalves » Tue, May 27, 2014, 7:42 PM UTC

YokoZar wrote: post #23

hcarvalhoalves wrote: post #18

Earlier quoted context omitted.

SimCity 4 "Rush Hour" introduces a refined traffic simulation where you can inspect the commute of each citizen and it's exactly what you would expect in real life - they go to anywhere in the city, not only nearby. Grid layouts don't optimize for heavy traffic [1] because drivers choose shorter paths, making it worse for all [2]. In real life you don't have grid layouts without large roads, one-way streets, overpass…

Thanks, this is all very interesting. I do think there's a bit more to consider though, such as how it's much harder to design walkably compact neighborhoods with arterial roadways than with gridded ones, meaning you might just get fewer cars in the gridded setup.

Yes, it's about trade-offs. The point I was making though is how the simulation models the counter-intuitive concept of "closing roads improve traffic" because the simulation embodies a property of game theory that also applies in real life.

Re: Game Theory Is Counterintuitive

#25

Post by vubuntu » Tue, May 27, 2014, 11:03 PM UTC

rogweiof wrote: post #11

The first bullet-point ("People often take aggressive postures that lead to mutually bad outcomes even though mutual cooperation is mutually preferable.") is not backed up by the source, is incorrect, and if it was correct it would be a finding from psychology not game theory. I believe (not an expert) it is more important to minimize losses that to maximize wins in game theory (cold war roots). Viewed in that light…

>I believe (not an expert) it is more important to minimize losses that to maximize wins in game theory (cold war roots).

That sounds meaningless. Aren't they the same thing? Isn't minimizing losses effectively mean maximizing wins? Give me an example where minimizing losses is not the same as maximizing wins. The only cases where it is not is if you haven't vetted out (or could not vet ) the complete outcome of all possible moves that you can make next, for example chess, where the game tree could be of infinite (or deep enough to process in reasonable time) depth.

But chess programs are coded to try all possible tree paths to a reasonable depth given the time constraints and pick a move that has the best weight/score/winning chances. Not a move that has least losses. From that perspective it is all about maximizing the chances of winning. Although I am sure they are also coded to recognize draw conditions and play for draw if that is more effective. So it is not so dry and cut.

So when you are not sure of the certainty of the outcome of the several possible moves you could make next in a game, the best strategy is to pick the one with maximum winning chances not minimal losses. Especially if you think the opposite party is also going to play to win. Next, if your chances of losing are increasing, then play for draw first and lastly mutual destruction (to force the opposite player to draw)

Even in the 'cold war roots' that you mention, the goal is not just to occupy/conquer the opposite country. The goal is also to do so at minimal loss to ourselves. So "winning" in the game of war "is to conquer enemy country at minimal losses or maintain status quo if the losses would be substantial/irrecoverable". So you are always playing to 'win'.