Research Talks - Abstracts, Spring 2002 Here are the abstracts for the Brown Bag research talks and other talks organized by INFORMS. If you would like to present your research in one of these meetings, please contact Gary Gaukler, gaukler@stanford.edu. DR. HAO LI Some Economics of On-line Dating Because on-line dating services rely on information provided voluntarily by participants to form matches, the business currently suffers from rampant self exaggeration and efficiency loss. As a potential remedy, consider how an on-line matchmaker may use a schedule of entrance fees to sort different types of agents on the two sides of a matching market into different "chat rooms," where agents randomly form pair wise matches. If matching characteristics of males and females are complementary to each other, there are fee schedules that improve matching efficiency. Further, if the complementarities are strong enough, the revenue-maximizing fee schedule achieves maximum matching efficiency. Applications to exclusion policies, one-sided subsidies and price competition are provided. Dr. Hao Li is a 2002-3 Hoover National Fellow at Stanford Hoover Institution and an associate professor in the economics department at the University of Toronto. ERIC COPE Memory-based Stochastic Approximation New techniques for finding the maximum of an unknown regression function assumed to be unimodal or concave will be discussed. Sample points are sequentially selected according to an adaptive scheme based on current regression estimates or estimates generated by sampling from a posterior distribution on an appropriate function space. These algorithms are shown to be convergent, and more efficient in small sample than standard stochastic approximation. Convergence rates are also discussed. Eric Cope is a PhD candidate in the Management Science and Engineering Department and is working with Professor Nicholas Bambos. QI LI Life, Death, and the Economy: Mortality Change in Overlapping-Generation Increasing life span poses many questions and challenges for fiscal systems. Demographers have shown that there are regularities in mortality change over time, and have used these to forecast changes due to population aging. Such models leave out potential economic feedbacks that should be captured by dynamic models such as the general-equilibrium, overlapping-generations model first studied by Yaari and Blanchard. Previous analytical and simple numerical work by economists has focused on comparative statics and used simplistic representations of mortality, such as the assumption of a constant age-independent death rate, or some parametric approximation to a survival curve. We show that it is straightforward to analyze equilibria in such models if we work with the probability distribution of the age at death. US and other data show that this distribution can be plausibly described by a normal distribution -- for this case we obtain analytical results. For the general case we have numerical results. We show that a proper accounting for the uncertainty of when one dies has significant qualitative and quantitative effects on the equilibria of such economic models. There will be, in turn, significant lessons to be drawn for models of future fiscal policy. Qi Li is a PhD candidate in the Management Science and Engineering Department and this is a joint work with Professor Shripad Tuljapurkar in biological science at Stanford. MIKE BEIGEL RFID: What Penguins have to do with optimization? Designing and Optimizing RFID systems. Radio Frequency Identification (RFID) is an electromagnetic proximity identification and data transaction system. Using “RFID tags” on objects or assets, and “readers” to gather the tag information, RFID represents an improvement over bar codes in terms of non-optical proximity communication, information density, and two-way communication ability. Operational RFID systems involve tags and readers interacting with objects (assets) and database systems to provide an information and/or operational function. This presentation is a summary of current RFID technology, a unique story of some early RFID history, and a vivid depiction of a scientific RFID project in Antarctica in which prominent scientists and engineers learned important optimization lessons from penguins. Mike Beigel is President and Project Director of Beigel Technology Corporation, in Encinitas, California. BTC is an R&D and consulting organization specializing in fields including RFID, computers, technology development and applications. Mike is a graduate of MIT in EE and Humanities. The first part of his career was in Electronic Musical Products. His first RFID project was in 1978, and led to the development of many of the RFID technologies in use today. His present involvement is in RFID, other industrial products, international projects, numerous inventions and patents.