Effect of Information in Bayesian Congestion Games
Manxi Wu

Citation
Manxi Wu. "Effect of Information in Bayesian Congestion Games". Talk or presentation, 23, August, 2017.

Abstract
This presentation studies the effects of heterogeneous information in routing games where multiple traveler populations are asymmetrically informed about the network state. We study a Bayesian congestion game, where travelers who have access to an information system belongs to one population. Our equilibrium notion is Bayesian Wardrop Equilibrium, in which each population uses routes with the lowest expected cost based on its belief about the state and information received by other populations. We consider the information environment in which beliefs are based on a common prior. We evaluate the effect of information penetration by studying the behavior and costs of Bayesian Wardrop equilibrium as the relative population sizes change. By introducing a metric to evaluate the "impact of information'', we show that the equilibrium edge load is independent with the relative population size if and only if the impact of information is fully achieved in equilibrium. Specifically, directional perturbation of relative population sizes results in three distinct regimes, where the equilibrium edge load does not change with the sizes in the middle regime, but changes in the two side regimes. Although information systems send correlated information with incomparable accuracies, we obtain a complete order of the information systems in terms of the equilibrium cost of the subscribed population. This order is dependent on the relative population sizes. Furthermore, given any common prior, there always exists an intermediate set of relative population sizes, where the equilibrium outcome is independent of the sizes, and the costs of all populations are identical. Finally, we provide bounds on the equilibrium social cost, and show that the results of worst case inefficiency in complete information games can be extended to heterogeneous information environment.

Electronic downloads

Internal. This publication has been marked by the author for FORCES-only distribution, so electronic downloads are not available without logging in.

Citation formats

• HTML

  Manxi Wu. <a
  href="http://www.cps-forces.org/pubs/266.html"
  ><i>Effect of Information in Bayesian Congestion
  Games</i></a>, Talk or presentation,  23,
  August, 2017.

• Plain text

  Manxi Wu. "Effect of Information in Bayesian Congestion
  Games". Talk or presentation,  23, August, 2017.

• BibTeX

  @presentation{Wu17_EffectOfInformationInBayesianCongestionGames,
      author = {Manxi Wu},
      title = {Effect of Information in Bayesian Congestion Games},
      day = {23},
      month = {August},
      year = {2017},
      abstract = {This presentation studies the effects of
                heterogeneous information in routing games where
                multiple traveler populations are asymmetrically
                informed about the network state. We study a
                Bayesian congestion game, where travelers who have
                access to an information system belongs to one
                population. Our equilibrium notion is Bayesian
                Wardrop Equilibrium, in which each population uses
                routes with the lowest expected cost based on its
                belief about the state and information received by
                other populations. We consider the information
                environment in which beliefs are based on a common
                prior. We evaluate the effect of information
                penetration by studying the behavior and costs of
                Bayesian Wardrop equilibrium as the relative
                population sizes change. By introducing a metric
                to evaluate the "impact of information'', we show
                that the equilibrium edge load is independent with
                the relative population size if and only if the
                impact of information is fully achieved in
                equilibrium. Specifically, directional
                perturbation of relative population sizes results
                in three distinct regimes, where the equilibrium
                edge load does not change with the sizes in the
                middle regime, but changes in the two side
                regimes. Although information systems send
                correlated information with incomparable
                accuracies, we obtain a complete order of the
                information systems in terms of the equilibrium
                cost of the subscribed population. This order is
                dependent on the relative population sizes.
                Furthermore, given any common prior, there always
                exists an intermediate set of relative population
                sizes, where the equilibrium outcome is
                independent of the sizes, and the costs of all
                populations are identical. Finally, we provide
                bounds on the equilibrium social cost, and show
                that the results of worst case inefficiency in
                complete information games can be extended to
                heterogeneous information environment.},
      URL = {http://cps-forces.org/pubs/266.html}
  }
  

Posted by Carolyn Winter on 24 Aug 2017.
Groups: forces
For additional information, see the Publications FAQ or contact webmaster at cps-forces org.