For the 1 February issue of Nature magazine, I wrote a Toolbox article on interactive figures. Unlike static PDFs or JPEGs, these figures allow users to explore the underlying data and code used to create them, for instance to zoom in on a crowded region of interest, or to probe the robustness of a computational model.
It’s an exceptionally broad and growing field of tech development, and my article name-checks more than a dozen tools. Inevitably, omissions were made, one of which was pointed out within hours of the article going live.
Writing on Twitter, Josh Nicholson, CEO of the online publishing platform Authorea, called me out:
“@j_perkel Great article! is there a reason we were not mentioned? We have been doing interactive publications much longer than F1000Research for example and have integrated with Jupyter, Plotly, Carto, d3js, etc. (see templates here: https://www.authorea.com/featured-templates …)” — 11:03AM 30 Jan 2018
I was particularly embarrassed by the oversight, as I actually wrote about Authorea and related tools in a 2014 Toolbox, which specifically noted the company’s goal of creating “interactive, data-driven” articles. So, I set out to rectify the situation.
As Nicholson told me in a follow-up conversation, Authorea is a collaborative authorship and publishing platform that allows multiple researchers to work on a document simultaneously thanks to the baked-in version-control system, Git. Think of it as a more research-focused version of Google Docs. Articles can be published on the Authorea platform itself, or directly submitted to journal publishers for consideration. As articles are basically Git repositories, other readers can ‘fork’ them to serve as foundations for their own work.
The platform allows users to embed a rich set of interactive elements into their articles, including interactive graphics created using Plotly, D3.js, and Carto (an interactive mapping tool), as well as interactive code exploration using Jupyter notebooks. One article by Nicholson on “the future of research writing and publishing“, for instance, includes a plot of ‘hard science’ and ‘life science’ submissions on the arXiv preprint server from 1991 to 2017. The image itself is a static PNG file. But two nearby icons point to the underlying code and data; click the code link and the system ‘spins up’ the Jupyter notebook that produced the figure, demonstrating how it was made (and allowing other users to download the code, or copy it for their own exploration).
“We want research to be part of the web, and the web is HTML, and HTML supports all types of interactive figures,” Nicholson says. That means moving away from Word and PDF documents, which are “static, closed, [and] hard to data-mine”, and towards a more open, organic infrastructure.
The aim, he says, “is to really rethink the research document…. It’s not just a static snapshot at one point that can never be altered; you can update it just like preprints, and you can make timestamps, etc etc. And this all gets towards the goal of improving how researchers write and communicate their findings.”
Interactive documents represent just a small fraction of articles published on Authorea to date — about 100 of the 7,700+ public documents on the site feature interactivity, Nicholson says, ranging from student essays to preprints to full-blown publications. One student at the NYU Center for Urban Science and Progress authored an essay that uses location data for New York City’s ‘Citi Bikes’ to calculate how far men and women ride. The article includes a static figure, but the underlying data and Jupyter notebook are just a click away.
Such papers represent “the future of publishing,” Nicholson says: “they’re open, they use open data … they include it in their paper, they tie it to Github, they use Jupyter notebooks, they have interactive figures.”
But for the moment, as I wrote in my Toolbox, few journals are capable of actually handling that level of interactivity. As a result, the Authorea system can flatten interactive documents into static PDFs, which can then be passed from person to person for editing and peer review. Those wishing to explore the figures interactively can view them on Authorea.
The platform has instituted a program to entice authors to embrace interactivity, offering free premium memberships to anyone who includes interactive visualizations in their articles.
In the meantime, I’m aware of at least one other research publisher supporting interactivity: the British Medical Journal (e.g., see here). If I’ve missed others, or key interactive tools, feel free to comment below. Mea culpa!