Live Mathematics on Authorea
A Case for Transparency in Science


Authorea is a collaborative platform for writing in research and education, with a focus on web-first, high quality scientific documents.

We offer a tour through our integration of technologies that evolve math-rich papers into transparent, active objects. To enumerate, we currently employ Pandoc and LaTeXML (for authoring), MathJax (for math rendering and clipboard), D3.js (data visualization), iPython (computation), Flotchart and Bokeh (interactive plots).

This paper presents the challenges and rewards of integrating active web components for mathematics, while preserving backwards-compatibility with classic publishing formats. We conclude with an outlook to the next-to-come mathematics enhancements on Authorea, and a technology wishlist for the coming year.



The motivation behind creating Authorea has been to help streamline academic collaboration in writing any flavor of scientific documentation, notably research papers aimed at passing peer-review and getting published as scientific proceedings. While the authorship and submission experience comes first, a goal that comes close second is to also increase the openness of the scientific process, using the final publication as a “looking glass” into the practices and data collection which happened “behind the scenes”.

We proceed to motivate why transparent research has superior properties and use “live mathematics” as one example of how Authorea enables it.

The core of the transparency problem is that we are still using the original publishing metaphor for documents, dating back to the innovations of 16th century Galileo Galilei, while simultaneously working on 21st century projects which are potentially large-scale, high-dimensional, multi-author and/or internationally distributed (Goodman 2014). The usual scientific document submitted to academic venues today is still oriented towards the printed page, remains opaque to the underlying data, of which it presents static snapshots, and is constrained by page count and margin sizes, often preventing it from providing sufficient detail of methodology and experimental setup.

This disconnect between experimental results and publications offers room for unintentional bias and experimental defects to remain unnoticed, making it difficult for reviewers to verify, and for follow-up experiments to continue the work in question. Studies have shown that even journals of the highest impact factors are vulnerable to retractions – see Fig. \ref{fig:retractions} for an illustration derived in (Fang 2011). In 2015 we have also observed a stream of high-profile retractions from some of the best scored journals that illustrate this problem, as tracked and discussed on the website11, seen June 2015 of the recent Retraction Watch initiative (Marcus 2011).