A guide to backward paper writing for the data sciences

Abstract

In this perspective, we outline a set of best practices for the planning, writing, and revision of scientific papers and other forms of professional communication in the data sciences. We propose a backward approach that begins with clearly identifying the scientific and professional goals motivating the work, followed by a purposeful mapping from those goals to each section of a paper. This approach is motivated by the conviction that manuscript writing can be more effective, efficient, creative, and even enjoyable-particularly for early-career researchers-when the overarching goals of the paper and its individual components are clearly mapped out.

Figure 1

Schematic representation of the process of backward paper writing The high-level steps involved in the process of backward data science manuscript preparation. The square boxes at the top represent the important pre-writing steps in which you clarify the scientific and professional goals motivating your work. The rounded box represents the process of initial writing and revision. Once a draft is complete, the diamond box represents circulating the manuscript to colleagues and mentors for feedback, or submitting for publication, with the expectation that this will result in further revision and updating of your work. The circle represents the typical endpoint of the process: publication in a peer-reviewed outlet, sharing publicly via a preprint server, publishing online via an interactive notebook or app, or the many other ways in which data science research can be disseminated to relevant scientific communities and the public at large. Finally, the dashed arrow represents the potential for post-publication revision in response to feedback and critique or new data. While not required, this type of post-publication revision is increasingly common in data science fields, allows for greater transparency, and may increase the long-term relevance of the published work.