|
||||||
|
||||||
 |
||||||
A team of University of Maryland researchers from the fields of business, computer science and economics has joined forces to solve one of the thornier problems facing the electronic commerce world: Transactions that are tied to the clock. Existing e-commerce technologies--such as online auctions like those hosted by e-Bay--are fine for buying and selling a broad range of goods and services, and the model has worked well for both consumers and businesses. But when the value of an item rises and falls depending upon when it's sold--as in the case of an airplane ticket or a hotel room--the understanding of key economic forces comes up short. Closing that knowledge gap in the e-commerce world is the focus of the Maryland team, which is operating under a $2 million, three-year National Science Foundation grant. Demand for a technological solution to the problem of time-sensitive goods in electronic markets spans a wide range of industries. Just a few examples: Transportation companies make the most money when their trucks are full of cargo. They need to quickly let clients know when vehicles are available so that shippers can bid for the space. Telecommunications firms who own large amounts of fixed bandwidth might wish to sell some of their unused bandwidth at non-peak times. Such firms want to sell this limited-time bandwidth at just the right price, while buyers seek to make the lowest bid that still beats competing offers. Hotel chains want to book as many guests as possible and airlines want to sell all seats for all flights. In businesses like these, where price is often the primary factor in the consumer's decision, setting the right price at the right time can make or break a booking. In each of these cases, once a deadline passes, the value of any possible transaction plummets, or sometimes evaporates altogether, such as when a commercial airliner pulls away from the gate with empty seats. The challenge facing the Maryland researchers is how to get the buyer and seller together, taking into account real-world obstacles such as distance and competition, arrive at a price that is best for both and execute the deal while it still matters. Led by Gnanalingam Anandalingam, the Ralph J. Tyser Professor of Management Science with a joint appointment in the Institute for Systems Research, the eight-member team will study the structures of such markets, develop a software prototype to lend artificial intelligence to the parameters of such transactions, and consider ways to integrate wireless devices into e-commerce. The bulk of the work will be done through the university's Center for Electronic Markets and Enterprises, which Anandalingam co-directs. If successful, the team--which includes members from the Smith School of Business, the Department of Economics and the Department of Computer Science--will open the door to a wealth of commercial applications, says computer science professor Venkatramanan Subrahmanian, who also holds an ISR appointment. Subrahmanian will work on the technological heart of the project, a so-called computerized "agent" that acts on a user's behalf. But only by pooling their talents and expertise, the team members say, can they hope to solve the complex problem the NSF has set before them. The problem itself isn't new to any of the Maryland researchers. All have ongoing investigations in related areas, which Anandalingam says was why the university won the grant in the first place. "When we wrote the proposal, it was clear that we knew what we were talking about," he says. "To get a proposal on a subject from somebody who just started thinking about it that day is different from one from people who have already been working on it." "All of us have research agendas consistent with the goals of the project," adds economics professor Peter Cramton. He and fellow professor Lawrence Ausubel, also on the team, have spent years researching the theory and dynamics of auctions. "We're all going to be pursuing the questions we're currently working on," Cramton says. The team approach to solving multidisciplinary problems, which the university has encouraged in recent years, is critical in solving real-world problems, officials say. The world has grown so complex that isolated research in a single field of study is almost always only one piece of a puzzle. Solutions come from assembling the pieces. "There are certainly studies that an individual can do. But if you're talking about research that's really broad like this, I would be surprised if anyone would take it alone," says Hank Lucas, Robert H. Smith Professor of Information Systems and co-chair of the CEME with Anandalingam. "If somebody in information systems were to work on this project alone, and start making assumptions about auction mechanisms, it would be hard to compete with somebody who's spent their professional career studying auctions."
Lucas will take part in designing the technology, but his chief role will be to evaluate both the software ideas and the underlying assumptions about how buyers and sellers behave. That will require close involvement with other team members as Lucas determines how best to test the team's assumptions. Like any other component of interdisciplinary research, the validation depends on input from everyone. To illustrate that, Lucas points to another example that the researchers often cite in explaining what they are studying: Selling tickets to sporting events or concerts. "We have to have some idea of what we're trying to evaluate," he says. But there are key questions to be answered. "Are we looking at the interface, how easy it is to put in a bid [for tickets]? Do we want to compare two or three different ways to find out how they perform against each other? Is our objective to maximize the number of seats that change hands, or to maximize the revenue to people who are selling tickets? Is it to minimize what people pay to buy a ticket?" Meanwhile, Michael Ball, the Orkand Professor of Management Science, also with an appointment to the ISR, is involved in designing a laboratory for experimental work and contemplating possible markets for the technology. Drawing upon research he's conducting for the Federal Aviation Administration, Ball sees a potential use for the technology in that it could allow airlines to buy and sell coveted berths, called "slots," at airports in near real-time. Airlines trade the slots now, but they can't do it quickly and there's no mechanism to add money to the mix. "The solution to most real complex problems doesn't confine itself to one discipline," Ball says. "If you want to attack the complex problems of the world today, you have to get a team. If you look at the most exciting things [in contemporary academic research], they're not coming from a single discipline." In designing the actual software, Subrahmanian plans to draw on existing technology first. And in order to know what he needs to create, he'll have to understand the potential markets. "What we already have is an infrastructure to build agents on top of legacy code and take actions based on it. What we don't have is anything specifically tailored to the marketplace. So we'll have to build that," he says. "We have to do a lot of work dealing with time and uncertainty and decision-making." Most computer users are familiar with simple Internet search agents, like the ones that scan the Web sites of news services and send a daily e-mail with links to articles that meet the search criteria entered by an individual user. The more complex agents Subrahmanian designs are imbued with artificial intelligence and decision-making capability. For example, a hotel chain could use such an agent to learn the room rates offered by competing chains and set its own rates accordingly. Such an agent could even be sophisticated enough to realize that if buyers are attending a conference in a specific hotel, they might pay more to stay in that hotel for the convenience. That kind of technology will be a huge advance for computer science if Subrahmanian can pull it off. At this point, he admits, that's a big "if." "I have some vague idea of how I want to do it," he says. "The skeleton is there. Pulling the pieces together is a different story. It's ambitious. And that's the goal; I prefer to go for the whole canoli and see what happens." Anandalingam chose the team members for their expertise, but building a successful research team takes more than that. It calls for visionaries committed to the team effort, willing to work toward the good of the whole rather than hoarding their best insights for their own advancement. All of the team members know each other, and each has worked with some of the others in the past. The time-sensitive markets project marks the first time all of them have worked together, but Anandalingam is confident that their familiarity with each other will create the necessary collegial atmosphere. As a bonus, most of the team members have worked in private industry. Assistant professor of information technology Subramanian Raghavan, who also is affiliated with ISR, spent four years with the telecommunications company U.S. West before moving into academia. Ball spent two years with Bell Labs, and later launched a consulting and software firm called RouteSmart Technologies in Columbia, Md., with three other professors, which they sold in 1998 after 18 years. Cramton and Ausubel serve as president and vice president respectively at Market Design Inc., a Bethesda, Md., consulting firm. It matters because private sector companies worry about real problems, Anandalingam says. "In industry, the client wants you to solve the problem they have. In academia, the client often wants to 'advance knowledge' rather than solve a problem. It's more theoretical than practical." "If you just think in the abstract, you don't see the complexity of real problems," adds Ball. "The richness and subtleties that make problems interesting are difficult to just think up." He always pumps his funding sources for real problems to work on, he says, preferring to find specific solutions to issues that business entities face rather than writing theoretical papers. As the team moves through the first few months of the project, its members are meeting weekly to discuss ideas, develop specific projects for themselves or their graduate students to tackle, and determine what each of them can contribute. Although they are united by a common goal, coming from different disciplines means they may have different definitions for terms and different modes of approaching problems, Ausubel says. And while the getting-to-know period typically doesn't take long, it's a necessary step to make the three-year project run smoothly. "Different approaches give you quite different answers about what you should do," Cramton says. Because the different disciplines each have their own priorities, it's important for the team to understand each other's point of view early on. It will be Anandalingam's job to keep the team talking, providing the leadership needed to ensure everyone is coordinated and focused on their individual contributions to the whole. "They're experts in one field, but they know something about the other fields," he says of his colleagues. "There is a fair amount of understanding [of the other disciplines]. Having people a little more advanced in their academic careers helps, because they're not so interested in writing that paper that just marginally improves the field." After the Smith School announced the grant, he heard from companies around the country either working on similar technology or in need of it. "It's obvious we hit on something that is very current," he says. "There are a whole lot of companies trying to be in this space. We're hoping we can bring something new to bear."
Maintained by the University of Maryland ElectricPub |
||||||
 |
