Why I do not Want to be a Co-author on Your Paper

I recently discovered that someone had put my name on a conference submission without first asking my permission.   This is inappropriate, and I asked that my name be talken off the submission.   Below is a “letter of explanation”, with general thoughts about authorship.

Dear XXX

Please do not add me as a co-author on a paper without getting my permission first.  I realise your intentions were good, but you need to ask first!  I also expect to see a draft of the paper in question.

In general, when I am asked to be a co-author on a paper, I look at scientific quality and my level of contribution, and sometimes (to a lesser extent) at scientific impact.

Scientific Quality

Scientific papers need to be factually correct, with good evaluation and awareness of previous work; papers based on ML need to use appropriate training data.   This is not rocket science, but unfortunately many NLP papers, even at “prestiguous” venues, do not meet these criteria.    From my personal perspective, I do not want my name on a paper which does not meet the above criteria.

In your case, you based your evaluation purely on metrics such as BLEU.   I do not believe that this is a good way of evaluating NLG systems (and have recently written a paper about this, as well as many blog posts), and I will not be a co-author on such a paper unless there are clearly-stated sound reasons for doing a metric-only evaluation (which was not the case in your paper).

My Contribution

Paper authors should have made a substantial contribution to the work described in the paper. In medicine, authors are expected to have made contributions to both the research work and the writing of the paper, and also to accept responsibility for properly investigating mistakes (ICMJE guidelines).   I think the same should apply to CS research papers: authors should have made substantial contributions to the novel ideas and/or evaluation of these ideas; authors should have helped write or revise the paper; and authors should accept responsibility for mistakes in this work.

In your case, we had a good face-to-face discussion about your research, and also exchanged some emails.   I hope this helped you; certainly I try to be helpful when people ask for my advice on their research.  However, I do not consider this level of involvement to be sufficient to warrant being a co-author.  It would be more appropriate to thank me in your “Acknowledgements” section.

Scientific Impact

When I am the first or primary author of a paper, I only submit it if I think it will have a real scientific impact.   To put this differently, writing a paper is a lot of work, and I’m not going to spend my time on a paper unless I think it is saying something worthwhile.

When I am asked to be a co-author, this is less of a concern (since other people are doing most of the writing), but I still want to see that the main author is excited about the work.  Which of course is true for you and your paper!

I also appreciate that junior researchers who are trying to establish themselves may need to be less choosy than established researchers such as myself.  In  very crude terms, I think a paper is useful from a CV perspective if it has the potential to increase your h-index.  Which means that a paper which gets 10 citations will be useful to your CV but not to mine.

Final Comments

Although I do not wish to be a co-author on your paper as it currently stands, I am happy to keep on discussing your research with you, and making suggestions as to how to improve it.   Perhaps at some point in the future this will lead to a proper collaboration and hence co-authorship.  But in the meantime, if you think my comments are helpful, just thank me in your Acknowledgements section.

Best wishes