I recently saw a thread on Twitter about an advisor and/or institution which required PhD students to have three papers in high-prestige conferences before they can graduate. I was surprised about this, this is certainly not my perspective. To me, the core goal of a PhD is learning to be a good researcher. As part of this, students do need to publish as evidence that they have acquired research skills, but I personally would never say that I expect N publications in a small set of selected venues.

I have had some really excellent students who published extensively as PhD students and afterwards became research leaders and valued colleagues. I have also had students who struggled to publish even one paper. Plus of course many students in between these extremes. I try to help all of my students, that is my job and role as a PhD supervisor. Unless my students decide that a PhD is not the right path for them, which has happened on a few occasions.

My expectations

My view is that ultimately a PhD is a sign that the individual knows how to do research (and how to write it up). I do insist that my students understand how to formulate hypotheses, conduct good quality experiments, honestly analyse results, and present their findings in a well-written and understandable paper; I also insist that they attempt something novel, interesting, and important.

Once my students have achieved these basics, I dont hold them to an expectation of N papers in high-prestige venues. In part this is because many (most?) of the papers I see in “high-prestige” venues such as ACL dont have clear hypotheses, good evaluations, honest analysis of results, and/or understandable presentation. Ie, there is often little correlation between what I regard as high-quality science and what gets published at ACL. There is also a huge element of chance and luck in reviewing and paper acceptance, which again makes my dubious about using this to assess whether a student has become a good researcher.

We also need to keep in mind that students have different career goals. Some students want academic careers, and publication record is of course very important for such careers. My personal view is that quality is more important than quantity; ie a single high-impact paper which people remember is more useful (for getting a job as a post-doc) than several papers which are quickly forgotten (whatever venue they are published in).

However, many of my students want to get jobs in industry, and here publications may not matter much (depends on the company). For example, some of my students have worked for Arria after getting a PhD, and Arria doesnt put much emphasis on publications when deciding who to hire. Potential employees need to be knowledgeable and capable; at the PhD level they should also be able to propose, present, and evaluate ideas for improving Arria’s technology. Having good research skills is important; having 3 ACL papers is of marginal relevance, especially considering (as above) the low scientific quality of many ACL papers.

Admitting and selecting students

I guess a related issue is my expectations for admitting PhD students. Some institutions only accept PhD students who have already published papers, this definitely reduces the risk of accepting a candidate who will struggle to publish. But I think this is unfair, especially to people who have unusual backgrounds that don’t fit the normal BSc-MSc-PhD path. We need a wider variety of people doing research, and I personally am happy to consider applicants with all sorts of different backgrounds provided that I think they are smart, hard-working, and have the “spark”.

I see a lot of talk in the academic community about increasing diversity in researchers. I think this is a great goal, and will help us do better research which is more relevant to society. But it will be easier to achieve this goal if we consider PhD applicants as individuals when deciding who to accept.

My experience

To put this in a personal context, I had my PhD oral examination in March 1990, and at this point I had only published workshop papers. I did have papers under review at INLG, ACL, and AAAI, all of which were accepted and presented later in 1990, but as far as I can remember (this was 31 years ago) I didnt know whether these papers had been accepted when I had by PhD oral. I certainly didnt have any publications when I was accepted as a PhD student!