2/15 Matthias Scheutz, Cognitive Science, Computer Science, and Informatics, IUB

The limited utility of communication for simple organisms

Abstract: Many forms of signaling have evolved in the animal kingdom for different purposes (including indicating food, signaling danger, and attracting mates). Yet, there are very few species that are capable of "representational signaling" or what we will call "communication". In this talk, we will report results from extensive artificial life simulations that point to the limited benefits communication has for simple agents in biologically inspired foraging tasks and show that communication can even have detrimental effects on performance in certain environments. Using a combination of statistical analyses of simulation results and complexity-theoretic considerations regarding the (computational) cost of components in agent control architectures, we will argue for the limited utility of communication for any kind of simple organism in an evolutionary setting. The upshot of the argument is that simple agents with simple architectures need very special environmental conditions for communication to benefit them and thus to evolve. We will conclude with a few speculative thoughts about what evolutionary pressures might have been at work in cases where communication did evolve.

Bio: Matthias Scheutz received degrees in philosophy (M.A. 1989, Ph.D. 1995) and formal logic (M.S. 1993) from the University of Vienna and in computer engineering (M.S. 1993) from the Vienna University of Technology (1993) in Austria. He also received the joint Ph.D. in cognitive science and computer science from Indiana University in 1999. Matthias is currently an associate professor of informatics and computer science in the School of Informatics and Computing, associate professor of cognitive science in the Cognitive Science Program and adjunct associate professor in the Department of Psychological and Brain Sciences in the College of Arts and Sciences at Indiana University where he also directs the Human-Robot Interaction Laboratory. He has over 100 peer-reviewed publications in artificial intelligence, artificial life, agent-based computing, natural language processing, cognitive modeling, robotics, swarms, human-robot interaction and foundations of computing and cognitive science. His current research and teaching interests include multi-scale agent-based models of social behavior and complex cognitive and affective robots with natural language capabilities for natural human-robot interaction.

2/22 David Hachen, Department of Sociology, University of Notre Dame

Weighted Reciprocity in a Human Communication Network

Abstract: Weighted reciprocity, the relative balance of interaction flows in a dyad, is an understudied topic in social network research because very few network datasets contain information on the frequency of contact between two people.  Using data on the calling patterns of 9 million subscribers of a cellular telephone network over a two month period in 2008, we find that weighted reciprocity has a long-tailed distribution, with most dyads characterized by reciprocity, but a significant minority of dyads being non-reciprocal.  We show that variability in the level of reciprocity is related to vertex similarity.  Dyads with vertices that are dissimilar in degree (as well as age and physical location) tend to be reciprocal, while dyads with degree similar vertices vary in their reciprocity levels.  Among degree-similar dyads, non-reciprocity is greater when the vertices are more dissimilar in their vertex strength, i.e. the extent  to which each party makes calls to other people.  These findings call into question sociological theories that view reciprocity as an outgrowth of strong ties among similar people.  Instead we propose a dynamic model in which tie decay among degree-similar dyads is less sensitive to  non-reciprocity.  While in general non-reciprocal dyads are less likely to persist than reciprocal ones, some non-reciprocal dyads persist and these dyads tend to be ones in which the vertices are similar. We conclude by speculating about the possible causes of this greater tolerance for non-reciprocity among degree similar dyads.

Bio: David Hachen is Associate Professor of Sociology and Co-Director Inter-Disciplinary Center for Network Science and Applications (iCeNSA) at the University of Notre Dame.  His major interest is in social networks, especially dynamic models of their evolution.  Through iCeNSA David works with other social scientists, physicists, mathematicians, computer scientists and electrical and civil engineers on a wide range of projects involving networks including the “Longitudinal Analysis and Modeling of Large-Scale Social Networks Based on Cell Phone Records,” the “Open Sourcing the Design of Civil Infrastructure,” and the recently created  Social/Cognitive Network Academic Research Center (SCNARC) funded by the Army Research Lab through the Network Science Collaborative Technology Alliance Program.  

4/5 Yves Gingras, Canada Research Chair in History and Sociology of Science, Department of History, Université du Québec à Montréal, Canada

Evolution of Interdisciplinarity (1900-2008)

Abstract: Since the beginning of the 1990s, interdisciplinarity has often been promoted as a value in itself assuring either a superior kind of knowledge or a better way to tackle important social problems, viewed as always tresspassing disciplinary boundaries. But beyond the unending debates surrounding the definition of terms, like multi- inter- and trans-disciplinarity, the hard question remains: how can we measure interdisciplinary practices? The obvious way from a bibliometric point of view is of course to look at the pattern of journal citations. Though limited, this indicator at least provides a first empirical measure of the links between disciplines and specialties. Most papers on this topic however have analyzed only the recent period, usually after 1980. Thanks to the existence of the “Century of Science” database of Thomson-Reuters, we can now look at the long-term evolution of interdisciplinarity over a century and see if we can observe paterns. For the domains of the natural sciences and engineering (NSE) the data suggest a movement of opening up of the major disciplines to contributions from other disciplines over the period 1900-1945, followed by reinforcement of disciplinarity between 1945 and 1975, followed again, in the recent period (1975-2008) by a movement towards more interdisicplinarity. In biomedical sciences, the period 1945-1975, saw a followed since then by a continuous rise of interdisciplinarity, that is, a tendency for most disciplines to open up to other disciplines. The talk will discus the data, as well as the link between interdisciplinarity and scientific impact, in more details and suggest possible interpretations of the trends observed.

Bio: After having been trained in physics (B.Sc., M.Sc.) à Université Laval, Yves Gingras obtained a Ph.D. in History and sociopolitics of Science at Université de Montréal. His research interests cover the formation of scientific disciplines, the mathematization of the sciences and the bibliometric analysis of the transformation of the scientific field in the 20th century. He is Scientific director of the Observatory of Science and Technology (OST) at UQAM. Over the last few years, he published papers in Isis,History of Science, Journal of the History of Biology, Social Epistemology, Science and Engineering Ethics, Social Science information and Journal of the American Society for Information Science and Technology.

4/12 Paul Thompson, Dartmouth Medical School

Collaborative Utility-Theoretic Information Retrieval

Abstract: The Probability Ranking Principle, which states that documents should be ranked in decreasing order of probability of relevance based on all evidence available to the retrieval system in response to a query, has been served as a foundation for probabilistic information retrieval.  Utility-theoretic information retrieval has been proposed, but not developed to the extent which probabilistic retrieval has been.  This talk describes the experience with probabilistic information retrieval for single searchers and suggests implications for the development of collaborative utility-theoretic information retrieval.  This discussion will use the NIH-funded VIVO and eagle-i projects as examples of how utility-theoretic retrieval might be implemented.

Bio: Paul Thompson received his Ph.D. from the University of California, Berkeley, in 1986. His graduate research was on probabilistic information retrieval. He has worked in the field of information studies for over 25 years. He has published numerous papers, journal articles, and book chapters, and has served as a reviewer for various conferences, journals, and the National Science Foundation. From 1986-1988, he was an assistant professor at Drexel University's College of Information Studies. From 1988-1993, he was a member of PRC, Inc.'s (now part of Northrop Grumman) artificial intelligence development group, where he conducted research in natural language understanding and information retrieval. From 1993 until 2001, he worked for West Publishing Company (now West Group), conducting research on natural language understanding, information retrieval, machine learning / text categorization, and text mining. After joining Dartmouth College’s Thayer School of Engineering and Institute for Security Technology Studies in 2001, he continued his earlier research and began new research in the areas of semantic hacking, the application of Semantic Web technology to sensor networks, and question answering. He has also conducted research on control system security in the oil and gas industry with the Institute for Information Infrastructure Protection. He is currently an instructor in the Department of Genetics at the Dartmouth Medical School.