← back to the blog

Journal Response Time (JRT) - Nightmare of PhD students!

Posted on April 30th, 2018 in Miscellanea by Besim Bilalli

In the world of research or otherwise known as the world of "publish or perish", publishing a paper is of "life or death" importance. One of the most relevant measures of success/usefulness/value in this world is the number of publications one has, and therefore the goal of every PhD student is to maximize it.

In general, however, the problem is not as naive as that. That is, in reality, a PhD student faces a more complex multi-objective optimization problem (in case we treat constraints as objectives, otherwise it is a constrained single objective problem).
One constraint of this optimization problem is "publish in a good venue". Now, a good venue can be defined subjectively, yet, we assume the existence of an oracle that can tell whether a venue is good or not. Now, one has to maximize the number of publications in good venues. An additional constraint in this problem is time. Time is usually defined by the project a student participates in. Hence, finally one has to maximize the number of publications p, in good venues v within a given budget of time t. If any one of the objectives is not met, the student may end up as a "failure"!

Yet, the paradox of maximizing the number of publications in good venues is that, you can publish in a good venue only if you have spend a lot of time on it. Therefore, given that you have a fixed budget of time from the start, you can produce a fixed number of publications (in good venues), where the number is upper bound by budgetOfTime/timeForGoodVenue. Therefore, the maximization function is bounded.

Furthermore, here timeForGoodVenue, measures the time one needs to spend on finding a novel solution for a certain problem, discussing with his advisor, designing the solution/architecture, writing the code, running experiments, then running experiments again, discussing with his advisor, running experiments, writing the first draft, improving the first draft, ..., etc. (normally, the number of drafts required for the completion is a 2 digit number). Hence, this is the time it takes one to complete his work, let us denote it with x.

Now, at least in the Computer Science field there are two possible (there are more, but here we are concerned with two) places one can submit his work. Submit for publication on the proceedings of a Conference, or submit for publication on a Journal. Usually in a journal one submits a more complete work --- more time spent. Now, the normal procedure is that, once you have spent x time for completing your work, you need to wait an y amount of time for the reviewers to review and decide whether to accept or reject your work. For the sake of simplicity, let us assume that we will always be accepted (which is quite an optimistic assumption). Now, in case of a submission to a conference, this time y is bounded. Yet, in case of a journal the time y is bound to an undeterministic function.

To ease the understanding of the problem and to show where we come into play, in the following we use an example. Let us assume that one is given a budget of time of 3 years, for publishing at least two articles in indexed journals, and two conferences. Next, assume that for a journal one has to work 7 months, and for a conference 4 months (values are based on real experience). According to these assumptions, to reach the final goal one needs to spend 2*7 + 2*4 = 22 months. Now, given the budget of time, which is 36 months, one has 36-22 = 14 months in disposal, for the whole reviewing processes. Assume for conferences the review process is 1.5 months. Thus, one is left with 11 months for the reviews of the journals. Recall that this model does not assume failure (i.e., rejection of a paper).

Now, the question is, "are 11 months enough for the reviewing of two journal articles?", or let us generalize it to "are m months enough for reviewing a paper p on a journal j"? Although it sounds as a simple question, it is not trivial to answer. Because the function is undeterministic (given the same input can give two different outputs, which by definition is not a function, yet assume it is :)), and depends on several parameters like the journal, the time submitted, the impact factor of the journal, the number of reviewers for the journal, the efficiency of the editor, etc. In the best case one may expect an average, rough number as an answer. This is where we actually come into play.

We developed a tool that based on empirical evaluation/computation provides an answer to this question. More precisely our tool helps on modeling this function by using empirical data for different journals. Note that this empirical data is not available but needs to be extracted --- which itself is not trivial, because the required data is "hidden". Therefore, one can use this tool to see the average response times of journals (JRT) and compare them , which consequently may help one choose the right journal based on one's constraints. For more, try it yourself: JRT calculator. Good luck (which you will need a lot :))!

Besim and Rana Faisal