By designing a chatbot that harnesses the power of machine learning to connect scientists/educators with existing tools, datasets, and other experts, this sociotechnical project aims to remove significant user support barriers in the US STEM community. A chatbot called Vidura will be developed to provide personalized user support to novice and expert users in the form of a conversational agent to facilitate interdisciplinary research and education collaborations. The chatbot investigations in the project benefit the investments in science gateways (SG) and cyberinfrastructure (CI) made by the NSF and other federal agencies for over two decades. Vidura chatbot works around the clock and makes the greatest impacts when user support by human agents (i.e., domain specialists or CI support persons) is not available and/or is too costly as a science gateway surges in user uptake. The Vidura chatbot will initially be prototyped in CyNeuro, a neuroscience SG, but will be made accessible to be adapted in SGs across multiple domains to benefit the broader research/education communities.
This project integrates human communication science, conversational agent design, recommender algorithms, machine learning techniques, domain topic modeling in a synergistic way that advances social science, computer science, and neuroscience. In addition, this sociotechnical project provides research opportunities to benefit undergraduate and graduate students in both social and computer sciences and creates new interdisciplinary courses. The project activities will benefit students with diverse backgrounds as it will be carried out at two large public universities, one of which is a Hispanic serving institution. The project objectives and activities will focus on answering three main research questions: (i) How to design a chatbot for gathering user requirements and creating user profiles with proficiency in order to provide personalized expert service support and maintain adoption? (ii) How to equip the chatbot communication and custom dialogue flows during support actions with underlying recommender algorithms using un-supervised machine learning within/across science domains? (iii) How to implement a chatbot framework in research and education workflows of data-intensive/computation-intensive application communities (such as the neuroscience community, as well as in CI provider communities of SGCI, XSEDE, NSG, CyVerse, and JetStream) to evaluate utility across domains and identify best practices for ongoing relevance? The project findings will ultimately advance knowledge on how to enable conversational agents implemented as chatbot interfaces with pertinent underlying recommender algorithms in order to provide expert services to scientific domain users with reduced cost and increased convenience.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
