Altruistic rational agent

Question

is there an agent in Game Theory (or even AGT), which goal is to minimizes other's costs? I've seen some agents that I would call "socialist", which goal is to minimize the sum, but including their cost. If you know any paper referring to this, I would be interested.

Answer 1

What you are asking about can be studied using normal game theory without adaption.

If all the agents share a goal, simply design a game in which they all have the same payoffs in each outcome. Then they will act in a purely cooperative manner.

Normally this can be studied using decision theory, since you can just treat the agents like a single "hive-mind" making all the decisions.

There are some cases in which this is not the case:

• Communication is restricted (such as in the Mermin-Peres magic square game). Then agents need to devise a strategy allowing each of them to make optimal decisions even if they have different information. These can be solved using extensive form games.
• The agents do not know each other's motives. That is, they do not know that they are working towards a common goal. This can be studied with bayesian game theory.
• The agents have different beliefs. I am not sure which form of game theory studies this.

If only some of the agents are altruistic, in that they want to maximize the payoffs of the other agents, you just need to make sure the altruistic agent's payoff is a monotonic function of all the other agent's payoffs. They still are technically maximizing "their" payoff, but when they do this they maximize the other agent's payoffs. For example, if each player has a certain amount of wealth, an altruistic agent's payoff only depends on how wealthy the other agents are, not their own wealth. This can be studied using the same game theory used for other games. Their payoff will likely be some sort of social welfare function, except that it does not take their own welfare into account.