Internet and Network Economics

Volume 5929 of the series Lecture Notes in Computer Science pp 316-327

The Impact of Social Ignorance on Weighted Congestion Games

  • Dimitris FotakisAffiliated withSchool of Electrical and Computer Engineering, National Technical University of Athens
  • , Vasilis GkatzelisAffiliated withComputer Science Department, Courant Institute, New York University
  • , Alexis C. KaporisAffiliated withDepartment of Information and Communication Systems Engineering, University of the AegeanResearch Academic Computer Technology Institute
  • , Paul G. SpirakisAffiliated withResearch Academic Computer Technology Institute

* Final gross prices may vary according to local VAT.

Get Access


We consider weighted linear congestion games, and investigate how social ignorance, namely lack of information about the presence of some players, affects the inefficiency of pure Nash equilibria (PNE) and the convergence rate of the ε-Nash dynamics. To this end, we adopt the model of graphical linear congestion games with weighted players, where the individual cost and the strategy selection of each player only depends on his neighboring players in the social graph. We show that such games admit a potential function, and thus a PNE. Our main result is that the impact of social ignorance on the Price of Anarchy (PoA) and the Price of Stability (PoS) is naturally quantified by the independence number α(G) of the social graph G. In particular, we show that the PoA grows roughly as α(G)(α(G) + 2), which is essentially tight as long as α(G) does not exceed half the number of players, and that the PoS lies between α(G) and 2α(G). Moreover, we show that the ε-Nash dynamics reaches an α(G)(α(G) + 2)-approximate configuration in polynomial time that does not directly depend on the social graph. For unweighted graphical linear games with symmetric strategies, we show that the ε-Nash dynamics reaches an ε-approximate PNE in polynomial time that exceeds the corresponding time for symmetric linear games by a factor at most as large as the number of players.