Collaborative Research: Elements: TRAnsparency CErtified (TRACE): Trusting Computational Research Without Repeating It

点击翻译按钮获取中文摘要

Research communities across the natural and social sciences are increasingly concerned about the transparency and reproducibility of results obtained by computational means. Calls for increased transparency can be found in the policies of peer-reviewed journals and processing pipelines employed in the creation of research data products made available through science gateways, data portals, and statistical agencies. These communities recognize that the integrity of published results and data products is uncertain when it is not possible to trace their lineage or validate their production. Verifying the transparency or reproducibility of computational artifacts—by repeating computations and comparing results—is expensive, time-consuming, and difficult, and may be infeasible if the research products rely on resources that are subject to legitimate restrictions such as the use of sensitive or proprietary data; streaming, transient, or ephemeral data; and large-scale or specialized computational resources available only to approved or authorized users. The TRACE project is addressing this problem through an approach called certified transparency - a trustworthy record of computations signed by the systems within which they were performed. Using TRACE, system owners and operators certify the original execution of a computational workflow that produces findings or data products. By using a TRACE-enabled system, researchers produce transparent computational artifacts that no longer require verification, reducing burden on journal editors and reviewers seeking to ensure reproducibility and transparency of computational results. TRACE presents an innovative and efficient approach to ensuring the transparency of research that uses computational methods, is consistent with the vision outlined by the National Academies, and enables evidence-based policymaking based on transparent and trustworthy science.The central goal of the TRACE project is the development, validation, and implementation of a technical model of certified transparency. This includes a set of infrastructure elements that can be employed by system owners to (1) declare the dimensions of computational transparency supported by their platforms; (2) certify that a specific computational workflow was executed on the platform; and (3) bundle artifacts, records of their execution, technical metadata about their contents, and certify them for dissemination. The first phase of the project focuses on the development of a conceptual model and technical specification that can be used to certify the description of a system, termed a Transparency-Certified System (TRACE system), and the aggregation of artifacts along with records of their execution, termed Transparency-Certified Research Objects (TROs). The second phase focuses on the development of reusable software components implementing the TRACE model and approach. To demonstrate certified transparency, the toolkit is used to TRACE-enable existing platforms including Whole Tale, SKOPE, and the SLURM workload manager. These TRACE-enabled systems produce certified TROs that can be trusted and do not need to be repeated or re-executed to verify that results were obtained as claimed.This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Division of Social and Economic Sciences within the Directorate for Social, Behavioral and Economic Sciences; and by the Division of Information and Intelligent Systems within the Directorate for Computer and Information Science and Engineering.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.