Advice for Academic Writing About Data

As the Managing Editor of an economics journal, I’m always intrigued by advice about what goes into writing a good academic paper. 

Jon Zelner, Kelly Broen, and Ella August take their whack at this pinata in “A guide to backward paper writing for the data sciences” (Patterns, January 2022). As is usual with these kinds of papers, some of the advice is worthy but dull and very basic. But there are also some insights that resonated with me, and I’ll emphasize those here.

From the introduction:

Academic and applied research in data-intensive fields requires the development of a diverse skillset, of which clear writing and communication are among the most important. However, in our experience, the art of scientific communication is often ignored in the formal training of data scientists in diverse fields spanning the life, social, physical, mathematical, and medical sciences. Instead, clear communication is assumed to be learned via osmosis, through the practice of reading the work of others, writing up our own work, and receiving feedback from mentors and colleagues.

It won’t come as a shock to any reader of academic literature that this “learning by osmosis” is at best a partial success in producing clear writing and communication.

A well-crafted data science paper is a pedagogical tool that not only conveys information from author to reader but facilitates the understanding of complex concepts. This works in both directions: The paper-writing process is an opportunity for the writers to learn about and clarify their understanding of the topic in addition to communicating it to someone else. If we can accept the idea of this kind of writing as teaching, we can take a lesson from research and practice in the field of educational development, particularly the backward approach to curriculum design, introduced by Williams and McTighe in their book Understanding by Design: “Our lessons, units, and courses should be logically inferred from the [learning outcomes] sought, not derived from the methods, books, and activities with which we are most comfortable. Curriculum should lay out the most effective ways of achieving specific results … the best designs derive backward from the learnings sought.“

This rings true to me in several ways. When you write, you should listen closely to what you are saying, for the sake of understanding yourself. The great author Flannery O’Connor once wrote: “I don’t know so well what I think until I see what I say.” Writing out results should help you to understand yourself. Another part of the task is not just to tell others what you did, but to teach them. (Surely, all academics know the difference between telling and teaching, right?) Finally, the idea that the paper should emerge from the learning outcome that is sought, not from “the methods, books, and activities with which we are most comfortable” seems worth taking to heart. This approach is what the authors refer to as a “backward-design approach” to academic writing.

Under a backward-design approach, the overarching goals of a course are defined first, and then used to motivate and shape everything from the assignments students will complete, the nature and volume of reading material, and the way class meetings will be used to advance toward these goals. In this way of thinking, a course has a set of standard components—assignments, reading, class time—but the way in which they are devised and arranged is organized around supporting the learning goals of the class. The same approach can be applied to the construction of a research paper: even though most papers have the same sections (introduction, methods, results, discussion) early-career researchers may underestimate the amount of flexibility and room for creativity they have in using these components to achieve their scientific and professional development goals. The backward approach we lay out here is about starting at the end by answering the questions of “What do I want accomplish with this paper?” and scaffolding each piece to help serve those goals. This is contrasted with the more ad hoc forward approach most of us have learned to live with, in which we begin with the introduction and struggle through to the conclusion with the primary goal of simply finishing the manuscript.

The article goes into the backward-design approach in more detail. Finally, I appreciated these thoughts about figures and tables:

If it can be conveyed visually, do it! Prefer figures over tables and in-text descriptions where you can. …. Reasonably informed readers should be able to get what is going on from looking at your figure and reading the legend, even if they have not read the rest of the paper. This is not a hard-and-fast rule, but if you work toward it you will ensure that the figures convey as much information as possible. … If you must make a results table, keep it small and simple. Big, complex tables are where reader attention goes to die. If information is best conveyed by a table, be sure to include only the most essential information. When a table gets too big, it becomes easy to forget what its purpose is. By keeping it short and cutting out extraneous information, you are better able to keep the focus on your message.

I fully intend to steal the phrase that “big, complex tables are where reader attention goes to die” for future editorial comments.