Research by PhD student Stefanie Stantcheva touches on taxation, student loans and education incentives.
A new interdepartmental Decision Sciences Program to be headed by Dr. Thomas L. Magnanti, George Eastman Professor of Management, has been announced by Dr. David Litster, vice president and dean of research.
Decision sciences is an interdisciplinary topic emerging in importance in many industries in both service and manufacturing sectors. It is a field that helps to guide individuals and firms in making informed decisions, often concerning the effective use of scarce financial, human, and technical resources.
The explosive development of new sources of information and the continuing emergence of ready access to faster and more powerful computers has created both enormous potential and considerable need for more informed decision making. The new Decision Sciences Program aims to exploit these opportunities by developing innovative educational and research programs and by working closely with industry.
Decision Sciences is a field with numerous applications in engineering, management and the life and physical sciences, and in industries as wide ranging as biotechnology, manufacturing, public systems, telecommunications, and transportation. An example is complex computer-based "passenger yield management" systems and models that the airlines use to adjust pricing of each flight's seats in order to maximize revenue and profitability to the airline. Another example is optimal scheduling of servers at a retail establishment to provide good service to customers while at the same time maintaining operating costs at a reasonable competitive level.
A decision scientist often addresses "how to" questions: how to design a computer chip or a long distance telephone network, how to evaluate risks of transporting hazardous waste materials, how to assess the effects of teams and group decision making, how to organize waiting lines in a bank or to schedule and route material through a manufacturing facility, or how to choose an investment portfolio.
An Interdisciplinary Field
The disciplinary base of the decision sciences ranges from behavioral decision theory and psychology to applied mathematics, economics, operations research, optimal control, and statistics, all topics that help inform individuals and organizations in making decisions. Recognizing the interdisciplinary nature of the field, the DSP will draw upon faculty and student interests from all five schools at MIT and coordinate activities across several existing centers and laboratories.
As Dean Litster notes, "decision sciences is the type of activity that should thrive at MIT. It is inherently multidisciplinary and can draw upon the broad expertise of the MIT Faculty. Decision sciences also requires a strong foundation in operations research, which has existed at MIT since the Operations Research Center was founded by Philip Morse in 1952. The new Decision Sciences Program will address issues that are important to society and I expect it will contribute much to the Institute's fluid intellectual environment."
The new program will initially operate as a "virtual" center or program, in the sense that it will not have its own large administrative and professional staff, or large program office; rather, it will draw upon existing resources and expertise in other centers and programs, including the Operations Research Center, the Laboratory for Information and Decision Systems, the Center for Transportation Studies, the Leaders for Manufacturing Program, and the Center for Computational Research in Economics and Management Science.
As Joel Moses, dean of engineering, states, "the Decision Sciences Program is an experiment in a new type of organizational form at MIT, one that is small and nimble and that can cut across the existing departmental and laboratory matrix structure that has evolved over the past 40 years.
"I am optimistic that programs like this will permit the Institute to develop new ways for our faculty and students to work together, ultimately resulting in new and innovative uses of information to improve corporate and societal decisions. Two similar structured programs announced earlier this year are the Program in Environmental Engineering Education and Research, directed by Professor David Marks, and the Technology, Management, and Policy Program, directed by Professor Daniel Roos."
One of the primary goals of the new program is to support graduate education in the decision sciences at MIT, enriching ongoing programs like those at the ORC, and creating new opportunities in fields like statistics and decision processes. The ORC program focuses on models, algorithms, and applications. The statistics program will address not only traditional material on statistical theory, but also place considerable emphasis on issues of statistics in the new information and technology era, including issues of computation, large-scale information data sets, process control, total quality management, and various allied applications. The decision processes track will consider such issues as human processing of information, man-systems interactions, and group decision making.
A centerpiece of the Decision Sciences Program will be a Transactional Data Laboratory. Examples of transactional data include bar-code laser scanned data from supermarkets, customer interaction data from ATM's (Automatic Teller Machines), materials movement and position data from assembly lines, stocks and options trading data, point of sale data from retailers, and satellite data giving estimates of vehicle locations. Transactional data sets derived from newly created data acquisition technologies are usually huge, and the recipient of the data is often unaware of the tremendous information content contained within the data set.
Figuratively speaking, in 1849 the gold was in the streams of California; in the 1990's the "gold" is often buried within gigabytes of data, and new sophisticated statistical information extraction procedures are needed to mine the gold. Examples of recent MIT-based successes in this area include the work of Institute Professor John D. C. Little on information extraction from supermarket bar-code scanned data and the work on inferring queue behavior from the start and stop times of customer service by Professors Richard C. Larson and Dimitris J. Bertsimas.
The Transactional Data Laboratory will be equipped with a sample of the new transactional data acquisition technologies, so it would resemble in many ways more traditional laboratories at MIT. In addition to computers, the Laboratory will include examples of data acquisition technologies that are giving rise to the gigabyte data explosion. These would include, for instance, pressure sensitive mats for detecting the presence of customers, infrared and ultrasonic customer detectors, various portable vehicle presence detectors and professional videotaping equipment to monitor the accuracy of recorded data from actual operations.
Students and faculty will often take a data acquisition device outside the laboratory into the real world and observe and model its performance in real world settings. Such trial implementations, coupled with videotaping capabilities, would also allow the controlled scientific testing of proposed statistical inference procedures developed by MIT DSP researchers. Additional performance inference experiments could be conducted within the laboratory, perhaps for example having student volunteers serve as simulated customers in some retail establishment.
The DSP will actively promote and foster interactions with industry. As examples, together with the Center for Transportation Studies and the Operations Research Center, the Program recently hosted a "Symposium on Use of Transactional Data in Transportation and Other Service Industries." From August 11-14, together with the ORC, the Program will host a major international symposium on information systems and operations research in telecommunications and transportation.
The Decision Sciences Program's Steering Committee represents a number of current Institute undertakings that the Program will attempt to bridge. Its members are: Professor John Carroll from the behavioral and policy sciences area of the Sloan School of Management; Professor Richard Larson of electrical engineering and computer science who is co-director of the Operations Research Center; Professor Steven Lerman, director of the Intelligent Engineering Systems Laboratory and of the CECI; Professor Sanjoy Mitter, co-director of LIDS and director of the Center for Intelligent Control Systems; Professor Amedeo Odoni of aeronautics and astronautics; Professor Yosef Sheffi, director of the Center for Transportation Sciences; and Professor Roy Welsch, director of the Center for Computational Research in Economics and Management Science. The chair of this committee is Professor Magnanti, who will direct the Decision Science Program. Professor Magnanti currently also serves as co-director of the Leaders for Manufacturing Program and of the Operations Research Center. Professor Larson will serve as a co-director of the DSP and as director of its Transactional Data Laboratory.
In launching the DSP, Professor Magnanti notes "the possibility of building upon the rich intellectual traditions of MIT in the decision sciences, of creating new opportunities for our faculty and students, and of bridging theory and practice in the best tradition of the Institute is very exciting. I would invite anyone from the MIT community, as well as from government or industry, who would like to learn more about this new program to contact me or any member of the steering committee. We very much welcome your participation."
A version of this article appeared in the May 20, 1992 issue of MIT Tech Talk (Volume 36, Number 31).