The Center for Open Science’s (COS) preprint platform was designed to serve a variety of scholarly communities, especially lowering barriers to entry for those disciplines new to preprint publication whose needs were not being served by the larger, more highly powered preprint servers like arXiv. This grant funds a project by COS founder Brian Nosek to use the COS preprint platform as the setting for a series of experiments that will test how user trust is affected by different preprint platform features. Nosek proposes to use the launch of already-planned features like annotation and visual icons to run a set of experiments on the assignment of trust by readers of scientific research. While the budget includes some technical development, the bulk of the requested funding will support the COS “metascience” team to take a mixed-methods research approach, combining surveys with analysis of usage data from the preprint servers to understand the impact of annotation and “reproducibility badges” on readers’ perceptions of trustworthiness of individual preprints and of the preprint server overall.

Somewhere between 50% and 60% of internet users use Google Chrome to browse the web. Chrome, unfortunately, doesn’t natively display mathematics using the standard XML markup language MathML. This forces sites like Wikipedia to generate static images of mathematical notation from the underlying MathML when Chrome can’t render the markup on its own. Not only does this have implications for accessibility, it also inhibits the development of innovative new interfaces and applications that would rely on dynamic interaction with mathematical notation via browser-based programming languages like JavaScript. Funds from this grant will support a project led by the National Information Standards Organization to implement full MathML rendering in Chromium (the open source codebase underlying the Chrome web browser). Technical development will be undertaken by developers at Igalia, an open source software consultancy that has played a key role in MathML integration in other major web engines.

Computers do nothing but process logical statements. Mathematics consists of nothing but such statements. It would be reasonable to assume, then, that computers would be adept, perhaps uniquely, at reading, understanding, and cataloging the academic literature of mathematics. Not yet. People and machines, it turns out, speak different mathematical languages. If computers are to help manage mathematical knowledge, they need to be taught how to read math papers. The grant funds efforts by mathematician Thomas Hales to begin that instruction. Hales has raised an international army of graduate students and postdoctoral researchers, which he plans to unleash on the abstracts of thousands of mathematical papers. They will carefully translate the definitions and results that appear in these abstracts into formal programming language. These formalized abstracts—“fabstracts,” for short—can then be used to train machine learning algorithms to “read” textual mathematics.   

Yale University Library digital archivist Euan Cochrane leads one of the most ambitious software archiving programs in United States research libraries. Currently accessible to Yale faculty and students, the Yale software collection relies on open source software called bwFLA that enables the creation, management, and distribution of “virtual machines” which can simulate the hardware of an older computer on a newer computer and then run older software on the simulated machine. In practice this means that if you have the right credentials, you can go to the Yale Library website, click a link, and suddenly be running Windows 3.1, the original MacOS, or any other operating system and software, right in your browser. This grant supports efforts by Cochrane and his team at Yale to further develop this infrastructure and, working with the Software Preservation Network, to cultivate this capability at other institutions. The grant will support focused work on four use cases: scientific software, CD-ROM archiving, restricted-access reading rooms, and a “Universal Virtual Interactor” that would automatically launch the correct software and version to open any given digital file. Other supported activities include technical refinements to the bwFLA platform and the archiving of the National Software Reference Library currently held by the National Institutes of Standards and Technology.