The effects of altruism and spite on games.

다국어 초록 (Multilingual Abstract)

Standard game theory assumes purely selfish or rational individual behavior, which means that every player will just act to optimize his own payoff function regardless of the effects that their choices may have on the others. However, many phenomena where people do care about others' benefits can be observed in the real world. Experiments also show discrepancy between experimental results and theoretical predictions with the assumption of selfishness. Various explanations with "not entirely selfish" players perceiving "other-regarding" payoffs have been proposed. One of them is altruism and spite among players.
Selfish outcomes have been observed to be drastically downgraded from the optimal one in several natural games. Since players are not totally selfish, these predictions may have been simply too pessimistic. Our goal in this thesis is studying the impact of partially altruistic and spiteful behavior on the outcome of games, and specifically the social welfare, in a social or economic network environment.
We develop and analyze a game-theoretic model with partially altruistic and spiteful players situated in an economic or social network environment. We show the effects of such a model on several problems: traffic routing, congestion games, network vaccination, and auctions. The trend of impact from altruism is different across classes of games. In particular, improvements on the Price of Anarchy are shown in routing games with non-atomic partially altruistic users. However, this trend is not the case for congestion games with atomic partially altruistic players in which the Price of Anarchy is increasing with altruism, yet some special cases of congestion games still exhibit the trend of improvements. Introducing partial altruism into network vaccination games can result in no stable outcome, which even changes the game dynamics completely. We then draw a roadmap of a few interesting future directions.