New mathematical research shows that consideration for others does not always lead to the best outcome - that is, when it’s applied in game theory. This result can have applications in data and traffic networks, peer-to-peer networks and GSP auctions, such as those used by Google Adwords. Bart de Keijzer from Centrum Wiskunde & Informatica (CWI) in Amsterdam studied algorithms for game theory, with a focus on cooperative aspects. The CWI mathematician succesfully defended his PhD thesis ‘Externalities and Cooperation in Algorithmic Game Theory’ on 16 June at VU University.
Game theory is a branch of mathematics that deals with economic strategic decision making. In this new study it is not about purely selfish players anymore, but also about influence by others, where, for example, one player does not want to hurt the other by his decisions. Sometimes this is a more realistic assumption than players being interested in themselves only - as is common in conventional models.
During his PhD research, De Keijzer analyzed different types of auctions. He investigated the impact of cooperation, friendship and animosity on different games. One of his conclusions is that when players behave altruistic, the flow in a road or data network can become worse.
“So, when altruism is introduced into the models, then the outcome could be worse. But I won’t say that math now proves that everyone should be selfish," the researcher laughs. "But it’s certainly a remarkable result that for the mathematical concept of social welfare, one can sometimes better choose at the expense of others, than to change the strategy to please them."
Both researchers and policy makers can take advantage of these more realistic models and make better qualitative predictions with them. Other research results from De Keijzer have applications in procurement auctions, treasury auctions, spectrum auctions and the allocation of housing.
More information:
- personal homepage of Bart de Keijzer
- homepage of promotor Prof. dr. G. Schäfer (CWI en VU)
- CWI's Networks & Optimization Group
