NEGT 2018

The Southern California Symposium on
Network Economics and Game Theory

California Institute of Technology

January 12, 2018



This symposium brings together students, professors, and researchers from Southern California who use game theory to analyze, design, and assess networked interactions across Economics, Computer Science, Engineering, Management and other disciplines. We hope to highlight connections between research areas and stimulate conversations about the benefits and limitations of game theory as a tool for understanding networked systems. Our community is interested in both the application of game theory to network related problems and in the development of novel game-theoretic methods; we also have a broader interest in learning and mechanism design.

Local organizer: Omer Tamuz. Administrative assistant: Letty Diaz. Email: Funding provided by Caltech IST and the Linde Institute.

1 Program

9:00 Haifeng Xu (USC)The Mysteries of Security Games:
Equilibrium Computation Becomes Combinatorial Algorithm Design
9:30 Pathikrit Basu (Caltech)Repeated coordination with private learningWe study a repeated game with payoff externalities and observable actions where two players receive information over time about an underlying payoff-relevant state, and strategically coordinate their actions. Players learn about the true state from private signals, as well as the actions of others. They commonly learn the true state (Cripps et al., 2008), but do not coordinate in every equilibrium. We show that it is possible to construct equilibria in which players eventually coordinate on the correct action, for any discount factor. For high discount factors we show that in addition players can also achieve efficient payoffs.
10:00 Jiasi Chen (UCR)Economic Analysis of Wireless Network Infrastructure Sharing
10:30 break
11:00 Vijay Vazirani (UCI)Google's AdWords Market: How Theory Influenced Practice I will talk about an optimal online algorithm for Internet Ad markets, such as Google's AdWords market. Although this result was obtained over a decade ago when the algorithmic and economic/game-theoretic issues of this marketplace were just being understood, its impact is becoming clear only in recent years. Our result addresses a central algorithmic issue underlying this marketplace: how to match query keywords to advertisers so as to maximize Google's revenue.

I will give an overview of the novel LP-based techniques that led to this result and the simple heuristic, of bid scaling, that is suggested by our algorithm. I will also present a formal framework for thinking about budgeted auctions more generally. These ideas have been widely adopted by Google and other search engine companies.

Purely theoretical work, from the 1990s, on the online bipartite matching problem greatly benefitted our work. The latter problem, while mathematically clean and elegant, appears to have no applications. On the other hand, the multi-billion dollar online ad industry has become the key source of revenues for several Internet companies. For algorithms designers, this a very happy story of practical impact from rich theory.

This talk is intended for a wide audience
12:00 lunch
1:00 Hamid Nazerzadeh (USC)Balancing Ride-hailing Marketplaces via Surge Pricing
1:30 Philipp Strack (UC Berkeley)Turning Up the Heat: The Demoralizing Effect of Competition in ContestsWe demonstrate, in the standard, symmetric complete-information all-pay contest framework, that, whenever marginal effort costs are increasing, increasing contest competitiveness demoralizes contestants, thereby reducing the mean and increasing the riskiness of output. This demoralization effect imposes testable restrictions on the mean and variability of output which we verify using extant experimental and field data. These results have significant implications: Although often criticized as evidence of laxity or cronyism, muting competition (e.g., adopting softer grading curves or less high-powered promotion systems) both reduces inequality and increases expected output. Reducing class size, simply by reducing classroom competition, improves student performance.
2:00 break
2:30 Elchanan Mossel (MIT)Information flow on (deep) networks
3:30 break

Vahab Mirrokni (Google) Robust Ad Allocation: Online Optimization & Auction Design

5:00 poster session

2 Venue and directions

The symposium will take place in the auditorium of the Beckman Institute at Caltech. The nearest parking structure is North Wilson.

3 Registration

Registration is free, via eventbrite closed.

4 Contact

Omer Tamuz: