Collaborative Research: Disciplinary Improvements: Repeto: Building a Network for Practical Reproducibility in Experimental Computer Science

New York University, New York NY

Investigators

Abstract

The Repeto project will foster community practices to make reproducibility a part of mainstream research and education activities in computer science. The project seeks to understand the cost/benefit equation of reproducibility for the computer science systems community, the factors that make reproducibility feasible or infeasible, as well as isolate factors (be they technical or usage oriented) that make practical reproducibility of experiments difficult. This research coordination network will develop a range of activities from teaching methodology for packaging experiments for cost-effective replication; using reproducibility in teaching; collaboration with reproducibility initiatives sponsored through conferences and institutions; community events emphasizing repeating or replicating experiments such as hackathons, competitions, or rankings; fostering repositories of replicable experiments and monitoring their usage/replication; to reporting on state of art and emergent requirements for the support of practical reproducibility. The outcomes of the proposal will be a collection of computer science experiments replicable on open platforms, an understanding of how much and to what extent they are used in mainstream research and education activities via relevant metrics, as well as a series of reports on current enablers and obstacles towards mainstream use of reproducibility in computer science research. Replicable experiments will be created using platform programmability tools including the Chameleon environment and associated software such as CHI, Trovi, and Jupyter notebooks. This platform programmability approach allows experimenters to express complex experimental topologies in repeatable and persistent ways. Combining platform programmability with executable notebooks will allow investigators to capture the full experimental process for subsequent replication by other researchers. This award by the CISE Office of Advanced Cyberinfrastructure is jointly supported by the CISE Computer and Networked Systems Division. 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.