| Marie desJardins | |
| Birth Place: | Washington, D.C., US |
| Fields: | Computer Science |
| Workplaces: | University of Maryland, Baltimore County SRI International Simmons University |
| Alma Mater: | Harvard University University of California, Berkeley |
| Doctoral Advisor: | Stuart J. Russell |
| Known For: | artificial intelligence and computer science education |
| Awards: | AAAI Fellow (2018)AAAS Fellow (2022) |
Marie desJardins is an American computer scientist, known for her research on artificial intelligence and computer science education. She is also active in broadening participation in computing.
DesJardins grew up in Columbia, Maryland. She received an A. B. in Engineering and Computer Science from Harvard University in 1985. She received a Ph.D in Computer Science from University of Berkeley in 1992.
In 1991 she joined SRI International, working in the ArtificialIntelligence Center. In 2001 she joined the Department of Computer Scienceand Electrical Engineering at the University of Maryland, Baltimore County as an Assistant Professor. While there she waspromoted to Associate Professor in 2007 and to Professor in 2011. In 2015, she was appointed Associate Dean for Academic Affairs in UMBCCollege of Engineering and Information Technology. She left UMBC[1] in 2018 to become the Founding Dean of the College ofOrganizational, Computational, and Information Sciences[2] atSimmons University in Boston.
DesJardins has explored the effect of the network topology on the efficiency ofteam formation in multi-agent systems, showing that scale-free networks areoften the most effective topologies for facilitating team formation andleading to the development of learning methods for agents to adapt theirbehavioral strategies.[3]
She has shown the first approach to trust modeling that explicitlyseparates the effect of competence (that is, the degree to which an agent isable to carry out its commitments) and integrity (that is, the degree to whichan agent is actually committed to complete its part of a joint action) ondecision making. This framework was later extended to incorporate reputation(indirect observations provided by third-party agents, with applications toonline rating systems and supply chain formation.[4]
In many domains, when a set of items is presented as a collection, interactions between the items may increase (due to complementarity) or decrease (due to redundancy or incompatibility) the quality of the set as a whole. Although this “portfolio effect” had occasionally been mentioned in the literature, this work was the first to address this problem a general way, by modeling the tradeoff between the “depth” of the set (i.e., which characteristics of the individual items are seen as more or less desirable) and its “diversity” (i.e., how broadly or narrowly distributed the objects in the set are over their possible range).[5]
This work presented a heuristic method for taking advantage of taxonomies, orhierarchies of values, in Bayesian network learning by searching for the mosteffective level of abstraction within the taxonomy, discovering whichdistinctions are relevant for the input data, and ignoring the others.This process reduces the number of parameters that must be estimated, andsimplifies the representation, while preserving the meaningful distinctions inthe domain.[6]
This paper, presenting comprehensive advice to help graduate students navigatethe process of earning an M.S. or Ph.D. and develop strong mentoringrelationships, has been circulated widely to graduate students around theworld and has been translated into multiple languages.[7] It has also been published in IAPPP Communications (Winter 1995,no. 58) and excerpted in SHPE (the official magazine of the Society ofHispanic Professional Engineers), Winter 2000, and in IEEE Potentials (August/September 1996).
In 2018, she became an AAAI Fellow[8] and an AAAS Fellow in 2022.[9]
Her other notable awards include: