In modeling scientific discovery. Over the last several years, this group has developed several computer programs that automate important aspects of scientific discovery in biology, chemistry, physics, and also general experimental science. They have identified promising discovery tasks by studying primary sources such as scientific articles, talking with a broad array of scientists, and even doing "field work" in some science. Automation methods have included the concepts and techniques of heuristic search, of algorithms and optimization, and a elementary mathematics and statistics. Concrete results in various fields back up the hypothesis that a significant fraction of scientific inference can be automated today. There is, however, the hypothesis that mush of scientific inference consists of generic tasks that are quite specific, but that - in their computational essence - are common to multiple sciences. The concept of generic scientific task is a new way to view computing in biology, computational chemistry, and so on. This research explore the concept of "generic scientific task" by investigating whether it leads to practical results (powerful discovery tools for scientists), and how this can provide a more systematic understanding os scientific inference.
