What can poker and AI teach us about politics?
Poker is commonly thought of as a mind game full of intrigue and perception in which the best players are those who can keep a poker face while bluffing and reading their opponents’ strategies. But this is a common misconception that originates from the way humans have traditionally played the game, not a feature of the game itself. In fact, artificial intelligence is exploiting this human weakness to pick apart the world’s best poker players.
Poker is an imperfect information game, and thus any player’s choices must be made with an uncertainty of outcome. Unlike chess or go, which are perfect information games, the result of a poker game cannot be determined by solely analyzing the game’s current state. A game of chess, for example, can be started from any initial configuration, and one can find an optimal strategy. The same is not valid in poker. Here the optimal strategy depends on the entire history of the game, not just the current state of a particular sub-game. Thus one must not only consider the outcome of a single sub-game but of a large number of them even if the current state of the game excludes these alternate sub-games.
The way artificial intelligence has dominated poker is by calculating the expectation values of many possible alternative sub-games and using these results to influence its optimal strategy for the sub-game that is being played out in reality. Thus, one should not only think about how the current sub-game might play out but how all possible sub-games might have played out. This is a task almost impossible for humans due to the large number of possible sub-games.
Poker, being a zero-sum game that ensures that money is conserved, provides that achieving Nash equilibrium is the optimal strategy. This simply means that losses are minimized. By collecting all expectation values of all possible sub-games of a given hand, an AI can evaluate the Nash equilibrium of the hand at the game’s current state and devise an optimal strategy.
Due to many decision points in a poker game, this calculation can never be exact. However, it is currently good enough to beat humans at a rate three times higher than humans can beat other humans.
All this shows that poker is not fundamentally a mind game based on deception but rather a game based on a complicated decision-making process with an optimal strategy known as the Nash equilibrium.
Politics and negotiations resemble poker large regard. They are primarily based on imperfect information and have a traditional focus on intrigues and deception. We can now see that this is not fundamental to making difficult decisions but is simply a result of human limitations in making complex decisions.
Poker is a game in which all the information is contained within a minimal region of information space. By that, I mean that poker is defined by a deck of cards, a few simple rules, and some chips. On the other hand, political decisions and negotiations rely on information that occupies a much larger information space. This increased complexity makes it much more difficult to calculate all possible sub-games’ expectation values independently of the number of sub-games that can exist.
With recent advances in AI, I can imagine a future where artificial intelligence will guide us in making these decisions. Moreover, AI will help us collect all the relevant information concerning a decision, even if that information is, to no small extent, imperfect. I thus imagine a future where politics will transition more into science and decision making will become less polarized.
