Wednesday August 22:
(Dept of Mathematics)
Title: Modelling Epidemiology and the Microeconomics of Infectious Disease Policy
Abstract: Over the last 50 years, theoretical epidemiologists have developed broad and powerful biology-based theories of infectious disease dynamics. But as we transition from growth-reliant to sustainable economies, health problems are becoming as much economic, demographic, and social as epidemiological. In recent years, innovative modelling approaches incorporating social networks, game theory, information propagation, and agent-based simulation have been proposed to represent the extra-biological factors interacting with the biology of disease transmission. In this talk, I will review some of the recent research exploring behavioral aspects of infectious disease management, with particular emphasis on the use population games to merge epidemiology with economics. The resulting models provide predictions of when intervention scenarios will precipitate policy-resisting and policy-reinforcing responses from the public.
Wednesday August 1:
(ARL / Dept of Mathematics)
Title: A Personally Biased Introduction to Early Results in Game Theory
Abstract: We will provide an overview of the larger areas of modern Game Theory broken down by the dichotomous areas of differential equation based and non-differential equation based. Our talk will be biased toward classical game theory with an emphasis on the connections between optimization and Game Theory. Beginning with Game Trees, we will move to matrix games and discuss the relationship between games and optimization problems, in particular highlighting the connection between Kurush-Kuhn-Tucker conditions and equilibria.
Wednesday August 8:
(ARL / Dept of Mathematics)
Title: Integration of Game Theory and Psychology: Overview and Examples
Abstract: Since my 1995 doctoral thesis I argued for the importance of incorporating non-trivial psychology in player models for games, which I call Multidimensional Dynamic Character. By multidimensional, I meant multiple and possibly conflicting decision algorithms were at work within each player, and by dynamic I meant that the balance of control among these competing algorithms, in a dynamic or repeated play context, could be affected by the history of play. While perhaps intuitively appealing in some regards, the approach raises legitimate concerns: is there a set of algorithms that adequately model all observed human behavior in all games? Are we risking the credibility of the entire theory by allowing arbitrary irrational decision making? Will models become so ad hoc that it becomes impossible to compare results from one model to another? John Nash has argued that "game theory" should be about the games rather than how people will play them, and suggested that this approach is better termed an "Integration of Game Theory and Psychology". Marvin Minsky's "Society of Mind" and related work in evolutionary psychology by Steve Pinker, Jerry Fodor, and others offers hope of a sound theoretical basis for the theory, and collaboration with psychologists can help ensure a level of consistency and credibility to avoid the temptation arising in applied settings to tailor models with particular results in mind. Old and new models will be presented to illustrate the ideas, with time allotted biased toward my current research, since a few members of this seminar have already seen the older models.
Wednesday August 15:
(Dept of Electrical Engineering / Dept of Computer Science & Engineering)
Title: Stochastic Loss Aversion for Random Medium Access
Abstract: We consider a slotted-ALOHA local area network with loss-averse, noncooperative greedy users. To avoid non-Pareto equilibria, particularly deadlock, we assume probabilistic loss-averse behavior. This behavior is modeled as a modulated white noise term, in addition to the greedy term, creating a diffusion process modeling the game. For a special case, we observe that when players modulate with their throughput, a more efficient exploration of play-space (by Gibbs sampling) results, and so finding a Pareto equilibrium is more likely over a given interval of time.