Collaborative Research: CyberTraining: CIU: Toward Distributed and Scalable Personalized Cyber-Training

This project is addressing the challenge of providing distributed, scalable, and personalized training of cyberinfrastructures - systems that offer state-of-the-art cloud services for storing, sharing, and processing scientific data. Today, personalized training of these rapidly evolving, and hence relatively undocumented, systems requires trainer-supervised, hands-on use of these systems. These training sessions require trainees and trainers to be co-located and provide personalized training to a relatively small number of trainees. The project is developing new (a) domain-independent technologies in distributed collaboration and machine learning to reduce all three problems in a concerted manner, and (b) domain-dependent training material targeted at trainees in statistics, physical sciences, computer science, humanities, and medicine. It, thus, serves the national interest, as stated by NSF's mission: to promote the progress of science; to advance the national health, prosperity and welfare.

A key technical insight in this work is that a cyberinfrastructure should not only support data science, but also make use of data science. The project is exploring two related innovations based on this insight: (1) Collaboration technologies that log, visualize and share the work of remote and local trainees to allow trainers to determine the need for remote or face-to-face assistance. (2) Machine-learning technologies that mine trainee and trainer interactions so that trainees can be automatically instructed on how to solve their problems based on similar problems that have been previously solved by trainers and other trainees. The project is leveraging existing technologies and training techniques developed for a widely used NSF-supported cyberinfrastructure, called CyVerse. This system is domain-independent, but so far, its training material has been targeted mainly at plant-science research. The project is extending the command interpreters and GUIs provided by CyVerse. The extended user-interfaces allow (a) trainees to announce difficulties and request recommendations, and (b) trainers to be aware of the progress of remote and local trainees, and remotely intervene when necessary. The functionality behind the user-interfaces is implemented by CyVerse-independent servers based on a general model of cyberinfrastructures, which includes the concepts of sharing and visualization of protected files, creation and execution of parameterized commands composed in workflows, and shareable, persistent work spaces. The project is adapting the CyVerse training material to cover new research domains including Geoscience, Political Science, and Biomedical Engineering. This expanded training material is being used to evaluate the proposed training technologies through training sessions for (a) students in a Statistics, Computer Science, Political Science, and interdisciplinary course, (b) attendees at three conferences targeted at Geoscientists, women, and Hispanics and Native Americans, respectively, (c) subjects in controlled lab studies, and (d) members of research groups at multiple institutes. The proposed qualitative and quantitative evaluation data gathered from these sessions are being used to assess not only the proposed technologies and training material, but also CyVerse and cyberinfrastructures in general.

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.