Chances are, as a computer scientist, you will write very few papers alone. Of all of the refereed conference papers that I have published, I have been sole author for only two. Instead, as the chart below shows, I am much more likely to publish papers that have three to five authors. Writing collaboratively creates interesting challenges and opportunities. This post focuses on your responsibilities as an author and how to work effectively with your co-authors.
- Contributed to the research,
- Contributed to the writing,
- Verified the content of the paper, and
- Are willing and able to present the work.
When you are the lead author for a publication, it can sometimes be hard to decide who else should be listed as an author. I am a fan of a direct and generous approach in such cases: Ask your collaborators! People will decline if they do not think they have (or will be able to) meet the criteria listed above, and if they accept they are explicitly agreeing to help you with the research and writing. The earlier in the process you ask the better, because that explicit commitment takes place up front. But do revisit the conversation if circumstances change. Never include anyone as an author that you have not had an explicit authorship conversation with, because that person has not agreed to vouch for the content of the paper.
Including additional authors is almost always a good thing, as long as each author genuinely contributes. As the graph below shows, in the eight years since I finished my Ph.D. I have progressively chosen to write papers with more authors. Credit does not seem to divide by the number of authors, but rather multiply. Co-authors provide diverse insight and perspective, additional research and writing support, and evangelization post-publication. There are long-term benefits as well; I have written refereed conference papers with 64 other researchers, and each is someone who will now speak well of my research and cite my papers.
You should also be generous with your acknowledgements. Include an Acknowledgement section in your paper even if it means you have to trim elsewhere, so that you can recognize ideas and feedback from colleagues. Keep track of the conversations you have, because it can be easy to forget who to mention.
One challenge with multiple authors is that you must decide the order in which to list them. Typically this is done by the size of the contribution, with the author who contributed the most being listed first.
However, specific communities have specific conventions (e.g., alphabetical, or student first and advisor last) which you should be aware of and follow.
It can be difficult to decide whether one author contributed more than another. Fortunately, however, the only position that really matters is that of first author, and this role is often fairly obvious. The first author is important because the paper will be referenced using the name of the first author (e.g., "Teevan et al. found that..") and the first author will present the paper. For students, in particular, it can be important to be first author on a number of early papers, since that will indicate an ability to drive research to prospective employers and provide visibility. The remaining positions do not matter very much, and nobody is likely to care whether they are listed third or fourth - although they may notice consistent patterns over multiple papers. For collaborators that work together regularly but contribute comparably, it can be a good idea to swap ordering occasionally. It is also common to list the most senior authors later, as they already have strong publication records.
Many conferences require that submissions be anonymous to avoid biases in the review process. This means the authors and their affiliations should not be easily discernible from the submission. If you are working with a proprietary system or dataset, for example, you should refer to it by a generic name. Do not, however, remove relevant citations to your past work from the paper, because their absence can make it hard to judge the contribution in context. Instead, continue to cite past work but do so in the third person. Change a statement like, "In previous work we found that..," to, "A study by Teevan et al. revealed that..." If you are not sure whether you have anonymized a submission appropriately, check with the Program Chairs for the conference.
Successful co-authorship means being a good collaborator during the writing process. Because you are not working alone, you must externalize your thinking. Nobody can build on something that is just in your head. Get your ideas onto paper as quickly as you can without worrying about perfection, and share outlines and drafts early and often. Logistically, while many computer scientists write their papers in LaTeX, I strongly prefer to collaborate using Word. Word’s change tracking and commenting functionality make collaboration fundamentally more straight forward, and I prefer to write in a WYSIWYG environment. Word supports seamlessly supports collaborative editing, so rather than emailing drafts around I like to create a share on SkyDrive.
When writing, be respectful of your collaborators’ time. Many deadlines fall at strange times, such as Sunday evening or three in the morning (i.e., midnight in Samoa), and you should not expect thoughtful feedback on a draft shared mere hours before the deadline. Share your time constraints and solicit other’s (e.g., with four children I jealously guard dinner time). Then create a schedule that works for everyone and keep to it.
Finally, don’t forget to speak up. If you are an author on a paper, you should feel ownership of the text. If one of your co-authors makes a point that you disagree with, raise it for discussion. If someone edits a paragraph in a way that doesn’t make sense to you, clarify it. It can be intimidating to write a paper with someone who is more senior in the field, but they are collaborating with you because they respect what you can contribute.
This post is part of a series of posts about the formula for academic papers. The components being discussed are: