skip to main content

CSIS Research

The Center for Social Information Sciences (CSIS) studies how markets and other social systems function in the overlapping spheres of economics and information and communication systems. Originally established as the Social and Information Sciences Laboratory (SISL) with funding from the Gordon and Betty Moore Foundation, the center now continues its work as CSIS thanks to Caltech's Ronald and Maxine Linde Institute of Economic and Management Sciences. CSIS combines researchers from economics, computer science, engineering, and mathematics in a truly interdisciplinary environment.

With The Linde Institute's ongoing support, faculty, postdocs, and graduate students come together regularly for seminars featuring internal and external speakers, as well as CSIS-organized workshops and conferences. Through these events and their publications, CSIS researchers strive to improve the basic sciences of complex markets and social/communication networks while helping to develop our understanding of the emerging interaction between the two.

Some of the specific topics being investigated by CSIS researchers include:

Learning in the Presence of Strategic Agents

The ability to learn from data and make decisions in real-time has led to the rapid deployment of machine learning algorithms across many aspects of everyday life. Despite their potential to enable new services, the widespread use of these algorithms has also revealed how susceptible current algorithms are to strategic and adversarial manipulations. This is the inevitable result of deploying algorithms — that were designed to operate in isolation — in uncertain dynamic environments in which they interact with other autonomous agents, algorithms, and human decision makers — all of whom have their own objectives. Addressing these issues requires developing and understanding of the fundamental limits of learning algorithms in the presence of strategic agents. Key questions in this area include: how do we design machine learning algorithms that are robust to strategic manipulations in their data? How does the presence of other learning algorithms affect your own learning? What should algorithms optimize for in multi-agent environments? Work in this area is led by Eric Mazumdar and Adam Wierman.

Risk, Uncertainty, and Information

Critical decisions are often made in environments in which the outcomes of different courses of action are uncertain. Whether these decisions are made by humans or by machines, it is important to understand how to evaluate and compare different risky prospects, and how to assess the cost of information and its value in reducing uncertainty. Constructing theoretical foundations and understanding the mathematics underlying these questions will potentially enhance our understanding of human behavior and our ability to construct better machines. Work in this area is led by Omer Tamuz and Luciano Pomatto.

Social and Economic Networks

The precise structure of social interactions can impact a variety of behaviors and outcomes—learning a new computer or spoken language may depend on the number of acquaintances who already know it, information about job openings may flow through word-of-mouth interactions, financial investments and outcomes may depend on the underlying connections between firms, etc. These observations have opened the door to an array of theoretical and empirical questions: How do individuals and organizations strategically interact with neighbors on complex social and economic networks? What network architectures are more conducive to diffusion of behavior and financial outcomes? How do we quantify the impacts of these networks on outcomes using field and experimental data? Work in this area is led by Omer Tamuz and Marina Agranov.

Rethinking Electricity Markets

Over the coming years, the electricity network will undergo an architectural transformation, similar to what has happened to the communication network over the last decades. The proliferation of renewable sources that are uncertain, uncontrollable, and have virtually zero marginal costs, the participation of a large network of distributed energy resources and financial players, the interaction of strategic behavior of prosumers, network structure, and physical laws may upend traditional market design. The engineering and economic challenges that must be overcome are made more formidable by the fact that the economic market structure and engineering architecture are inherently intertwined in the electricity grid. Work in this area is led by Mani Chandy, Adam Wierman, John Ledyard, and Steven Low. More details can be found at the Smart Grid project page and the Resnick Institute website.

Network Economics

It is almost impossible to study computer networks today without considering economic issues. Economics plays a defining role in routing (e.g., hot-potato routing and net neutrality), and economics has come to play a major role in how protocols are designed and analyzed (e.g., the analysis of TCP and the design of BitTorrent). In fact, even the study of cloud computing cannot be isolated from the strategic economic interactions with respect to pricing and provisioning between infrastructure providers and the services that run on top of them. Work in this area is led by Adam Wierman and Steven Low.

Computational Advertising

Extracting revenue from search algorithms increasingly depends on sophisticated computational algorithms for advertising. Many of these algorithms are based on auction theory. Research in this area thus requires very close attention to the interaction between computation and economics. Caltech was at the forefront of computational advertising from the inception of its use on the Internet. Our involvement began with work on the generalized second-price auction in concert with GoTo.com (which morphed into Overture, which morphed into Yahoo), the company that originated the use of auctions for location on search results pages. This work was both theoretical and experimental and was led by John Ledyard, with former Caltech faculty Matthew Jackson and Simon Wilkie. Later, more advanced and computationally intensive advertising work was led by John Ledyard for television and radio advertising and former Caltech faculty member Preston McAfee for webpages.