Economics, Computer Science, and Policy: Cross-Fertilization of Ideas and Techniques between Economics and Computer Science Is Yielding Fresh Insights That Can Help Inform Policy Decisions

Article excerpt

Perhaps as little as a decade ago, it might have seemed far-fetched for scientists to apply similar methodologies to problems as diverse as vaccination against infectious disease, the eradication of email spam, screening baggage for explosives, and packet forwarding in computer networks. But there are at least two compelling commonalities between these and many other problems. The first is that they can be expressed in a strongly economic or game-theoretic framework. For instance, individuals deciding whether to seek vaccination against a disease may consider how infectious the overall population is, which in turn depends on the vaccination decisions of others. The second commonality is that the problems considered take place over an underlying network structure that may be quite complex and asymmetric. The vulnerability of a party to infectious disease or spam or explosives depends strongly on the party's interactions with other parties.

The growing importance of network views of scientific and social problems has by now been well documented and even popularized in books such as Malcolm Gladwell's The Tipping Point, but the central relevance of economic principles in such problems is only beginning to be studied and understood. The interaction between the network and economic approaches to diverse and challenging problems, as well as the impact that this interaction can have on matters of policy, are the subjects I will explore here. And nowhere is this interaction more relevant and actively studied than in the field of computer science.

Research at the intersection of computer science and economics has flourished in recent years and is a source of great interest and excitement for both disciplines. One of the drivers of this exchange has been the realization that many aspects of our most important information networks, such as the Internet, might be better understood, managed, and improved when viewed as economic systems rather than as purely technological ones. Indeed, such networks display all of the properties classically associated with economic behavior, including decentralization, mixtures of competition and cooperation, adaptation, free riding, and tragedies of the commons.

I will begin with simple but compelling examples of economic thought in computer science, including its potential applications to policy issues such as the management of spam. Later, I will argue that the power and scale of the models and algorithms that computer scientists have developed may in turn provide new opportunities for traditional economic modeling.

The economics of computer science

The Internet provides perhaps the richest source of examples of economic inspiration within computer science. These examples range from macroscopic insights about the economic incentives of Internet users and their service providers to very specific game-the-oretic models for the behavior of low-level Internet protocols for basic functionality, such as packet routing. Across this entire range, the economic insights often suggest potential solutions to difficult problems.

To elaborate on these insights, let us begin with some background. At practically every level of detail, the Internet exhibits one of the most basic hallmarks of economic systems: decentralization. It is clear that the human users of the Internet are a decentralized population with heterogeneous needs, interests, and incentives. What is less widely known is that the same statement applies to the organizations that build, manage, and maintain what we call monolithically the Internet. In addition to being physically distributed, the Internet is a loose and continually changing amalgamation of administratively and economically distinct and disparate subnetworks (often called autonomous systems). These subnetworks vary dramatically in size and may be operated by institutions that simply need to provide local connectivity (such as the autonomous system administered by the University of Pennsylvania), or they may be in the business of providing services at a profit (such as large backbone providers like AT & T). …