Publishing in Economics, Philosophy, and Computer Science

21 June 2020

In a recent paper, economist Gerege Akerlof discusses Sins of Omission and the Practice of Economics. He argues that an excessive focus on hardness, understood as the difficulty of producing precise work, has led economist to neglect in fact important questions. For some important issues, papers of sufficient hardness are not available. I want to focus on one consequence that Akerlof proposes: The curse of the top five.

In economics, the top five journals dominate decisions about tenure and promotion. They have abysmal acceptance rates, apparently around 6%. Authors will often have to substantially revise their papers multiple times. As a result, the length of papers has increased as they attempt to pre-empt any possible objections. This curse of the top five will sound familiar to academic philosophers in the anglophone world.

These trends have been observed for philosophy as well (example 1, 2). But I don’t think the same notion of hardness is in play for philosophy. In the examples of Akerlof, hardness is strongly associated with mathematical formalisation and the need for statistic evidence. Neither enjoy nearly the same prominence in philosophy, from what I can tell. I would think they are much more important in e.g. computer science. After all, mathematical description of new methods, say a new loss-function for neural networks, and statistical evidence figure prominently in many computer science papers. As it turns out, however, computer science does things in a radically different way.

Instead of journals which often take years to publish a paper after first submission, computer science is organised around conferences. The top conferences are considered more prestigious than journals, which also exist. As a consequence, turn-around for the most prestigious venues is faster and it is not possible to request multiple rounds of revision. These conference papers are much shorter, often limited to 8 pages (although they often come with extensive appendices which can make up the difference). Acceptance rates are also considerably higher.

So, yeah, maybe that model is worth a look. The problem might be resetting the discipline to accept it.