Recommending collaborators

The technological advances of the latter part of the 20th Century have opened the door for scholarly collaboration on an unparalleled level. Such collaboration between researchers has become one of the drivers of scientific research. The scientific community is now faced with many Grand Challenges: in environmental science, how to address local and regional climate variability; in the health sciences, how to create new vaccines in health science; in neuroscience, how the brain produces mental activity. These challenges require the combined efforts of several scientists and engineers with complementary expertise. Federal agencies mandated to promote the growth of science, such as the National Science Foundation, place a strong emphasis on cross-institutional multidisciplinary collaboration when awarding grants. Academic researchers are also faced with the challenge of how to be successful professionals. The are multiple pressures and demands of academic life; publishing and obtaining grant funding are two that in some institutions are closely tied to obtaining tenure. These pressures are particularly intense on incoming junior faculty. One way to meet these obligations is to engage in multidisciplinary collaboration. However, discovering collaborators in unfamiliar disciplines is not an easy task. Collaborators are usually discovered through encounters at conferences or by the recommendation of a colleague. However, these methods work best for locating collaborators in allied disciplines. Researchers do not have the time to wrestle with multiple domain taxonomies to utilize alternate solutions such as search engines, expert locators, and community portals. Thus, there is a perceived need on a micro and macro level to facilitate multidisciplinary collaboration. My research explores the potential of a recommender system to recommend multidisciplinary collaboration partners for researchers in an academic context. My initial approach proxies funded grant proposal as collaboration experiences, from which repeating patterns of collaboration are extracted. A researcher can be matched to these patterns and the best match can be used to recommend collaboration partners. The advantage over current methods is that the researcher only need specify their own information and does not have to give any specifics on the collaborators sought. A researcher would specify information about themselves such as their research interests, seniority, department, and home institution. The system would then seek out collaboration patterns which include a member with characteristics similar to the researcher. The recommendation would then be to collaborate with individuals similar to the remains collaborators in the pattern. This information could then be entered into existing solutions such as an expert locator to discover specific collaborators