Reflections on a paper a day

2020-06-07 3 minute read

Four years ago I read a paper a day for almost a year. Two things stood out to me from the experience: (1) the sense that there is an insurmountable number of papers that need to be read to keep on top of the literature completely disappears when you do this, and (2) holding yourself to reading a paper every day (seven days a week) is a terrible and demoralizing idea.

The motivation for reading a paper every day was that I felt I wasn’t reading enough. If I didn’t have time dedicated each day I found it was easy for weeks to slip by without actually sitting down and reading through anything properly. I would scan the literature every day (mainly using RSS feeds) and note all the papers I found interesting, but rarely would end up going back and actually reading much more than the abstract.

Another fact has become obvious to me as it must to anyone who has ever done a journal club. Reading a paper in enough detail to present to others, either through a journal club presentation or in a blog post, helps me to remember a paper much better than passively reading, which I’ve found to be almost pointless. I’ve opened papers before that have my highlights all throughout but I have no memory of the paper at all. In contrast, I still remember almost every paper I’ve presented at a journal club or written about. I wrote a blog post about FISSEQ six years ago, and I could still remember plenty of details about the method when reading the new expansion sequencing paper.

As I now find myself slipping into the same habits that motivated me four years ago to read a paper every day, I have started thinking how I can use the experience from last time to now come up with a lasting strategy that will enable me to stay on top of the literature in more than a superficial way. I have decided on the following plan:

First, I won’t hold myself to a reading a paper every day, but instead aim for 2-3 papers per week. This should be a more manageable number. Second, rather than just listing the names of the papers I read, I will summarise them and discuss them as a blog post. Third, I will allow combining multiple papers on the same topic into a single post. For example, I’m interested in the graph-based community detection methods, and there are ~6 core papers about this topic that I’d like to read in detail. I’d then allow myself to spend 2 weeks on these 6 papers and write one blog post covering the whole topic after those 2 weeks.

Hopefully this experience will be more enjoyable, sustainable, and productive that the first.