Marie desJardins

Researcher focused on understanding trust among agents in online communities, desJardins has led the Multi-Agent, Planning and Learning Lab (MAPLE) at UMBC, which focuses on developing A.I. solutions to real world problems. A Professor of Computer Science, she has described what she does as "trying to get computers to do things that you would think were smart if people did them." Within the realm of artificial intelligence, desJardins divides her research interests into the three areas within her lab: Multiagent Systems, Planning, and Machine Learning.

Multiagent Systems deals with the task of getting multiple intelligence systems, like humans or A.I.s, to solve problems together. She is interested in the problem of trust, working to understand how to know which agents—for example, restaurant or movie reviews, or travel services—in an online community are trustworthy and which are not. desJardins has been working with a referring agent that she knew would overestimate an individual's ability and provide her with biased positive referrals. A biased agent, she explained, leads to the phenomenon of optimistic and pessimistic referrals.

She is quoted as saying, "Planning focuses on the problem of trying to pre-plan in complex domains where planning is hard." desJardins has compared her work to the job of a logistics planner for a FedEx fleet who is bombarded with last-minute pick-ups and deliveries that dynamically change his anticipated plan. In both cases, the task is the same: "What can you do in advance to anticipate what the likely kinds of requests are and be prepared to change things quickly?"

Machine Learning deals with building models to classify data or to make predictions. She explained the concept with an example close to home: predicting whether or not students would pass a class. "What are the attributes that actually lead to success or failure in that context," she said, "That's the model building question." In some cases, however, there was not enough data to build a model. If the model-builder did not know information such as the number of hours each student spent doing homework, for example, it became difficult to predict their success in class. desJardins explained that this is where cost-sensitive feature acquisition comes into play—meaning that certain information could be collected, but if the model relied on that acquired information, it became severely limited by its necessity to have that information for all future predictions.

She is especially interested in collaborating with students and helping them develop their own research interests, noting that nearly ninety-five percent of her research has been conducted with students. "I like the students to learn about a problem and find something that they think is interesting," she said. desJardins has also mentioned that she is increasingly thinking about research methodology more explicitly and is interested in writing about how to do research effectively. Her ultimate vision—though she acknowledged it was probably too ambitious to realize—is an all-purpose A.I. that helps with computer maintenance and other tasks. "What I would love to exist by the time that I retire is a true intelligent agent that would live on your laptop and monitor your life," she said.