Post has attachment
"We can think of the trained AI agents as an approximation to economics’ rational agent model “homo economicus”. Hence, such models give us the unique ability to test policies and interventions into simulated systems of interacting agents — both human and artificial."
Post has attachment
Techniques learned by playing Rock, Paper, Scissors can also be applied to all sorts of other realms of trying to predict human behavior, whether it’s Target figuring out when women are pregnant or Siri determining what people are asking for.
Hello. Nice to meet you. Let me introduce myself. I am Seung-Ho Lee from Korea. I study in the game theory in Busan national University. Hope to connect lots of people through this community. ^^ ﻿
Post has shared content
The Betrayer's Banquet
The Betrayer's Banquet (http://www.betrayersbanquet.com/) is a game-theory--themed dinner, which I have to say sounds like a lot of fun. The principle is simple: there's a table with 48 seats at it, and everyone is going to eat a 32-course meal. At the top of the table, the food is excellent; at the bottom of the table, the food is horrible. And between courses, you each play a simple game.

At first, this game sounds like the Prisoner's Dilemma. The diners sitting next to each other compete: they each choose whether to cooperate with each other, or betray one another. If both cooperate, they both move six spaces up; if both betray, six spaces down. If one cooperates and the other betrays, the traitor moves up ten, while his unlucky victim moves down ten.

If you're familiar with game theory, you know that the winning strategy in the iterated Prisoner's Dilemma (where you play it repeatedly with someone) is tit for tat: on the first round cooperate, then do whatever your opponent did on the previous round. But that won't work in the Betrayer's Banquet, for two different reasons.

First, you aren't playing against the same person repeatedly, but against different people every time. I suspect that you could resolve this by playing a variant of tit for tat, always doing unto others what they did to their partners the previous turn, although I haven't done the math.

But second and more importantly, this is a zero-sum game: if everyone at the table cooperates, everyone moves up six spaces, which is to say that nobody moves at all. Traitors are the only people who will advance, at the expense of cooperators and of traitors who ran into each other.

If you're near the top of the table, there's little reason to betray. You stand to gain relatively little, but to lose a lot if you get marked as a traitor by your top-of-the-table peers. The rich should cooperate.

On the other hand, if you're near the bottom of the table, there's little reason to cooperate. You have little to lose, and everything to gain. The peasants ought to revolt -- but of course, if they all revolt, then they are simply stabbing one another in the back, and nobody gets anywhere.

The middle of the table is, in a way, the most interesting: whether or not you wish to betray depends on your tolerance for risk. This turns it into a poker game: whether or not you wish to betray depends, not only on your own tolerance for risk, but on your estimate of your opponent's tolerance for risk. If you're seated opposite a daredevil, you ought to betray them. Of course, by this logic, if you're seated opposite someone risk-averse, you ought to betray them, too. Your best move here is to betray people while looking like the sort of person who would never do so -- which is hard if people can see what you did before. The right move is probably to keep playing nice, unless you're against someone who will obviously betray you, until you're in a good position to unexpectedly betray someone.

It's interesting to note that the closest thing possible to a "fair" outcome -- where everyone cycles through all of the seats equally -- is nearly impossible. Different situations give people very different incentives.

This is a rather good game, I think...

via
Post has attachment
I'm posting this here, in case anyone has any suggestions for John Baez (or if you're just interested in finding out about good undergrad-level game theory texts).  If you do, you can either comment there and see the full discussion or comment here, and I'll be sure to transfer your comments over to the main discussion.