Guest post by Felix Brandt
We recently had an internal discussion about author ordering within our group and agreed to retain alphabetical ordering (sort of renewing a decision we made some years earlier when the group was composed differently). Author ordering is a surprisingly tricky issue and I think it’s particularly difficult in algorithmic economics where the cultures of many different disciplines clash.
The problem usually starts with the fact that nobody makes a conscious decision at the beginning of one’s career which author ordering policy to use. In most cases, PhD students just publish their first papers using whatever convention is used in their group. Once they have published papers, they are reluctant to change the convention for the sake of consistency and, perhaps more importantly, because changing the convention can have negative effects on previous coauthors (by making their contribution appear less). Whenever authors with different authorship conventions write a joint paper, they need to decide which convention to use (and if it’s not the alphabetical convention, they also need to agree on a particular ordering for the paper).
The two predominant conventions are alphabetical ordering and ordering by contribution. (Another interesting convention I heard about, but that I have rarely seen, is to list authors by age from youngest to oldest.) Alphabetical ordering is the standard in mathematics, economics, and theoretical computer science. In most other disciplines, including AI, it is not. Sometimes, the head of a group is listed as an author even though he did not contribute anything and, in some areas, there is a special meaning to being the first, the last, or even the second-to-last author (like here). Let’s assume for simplicity that in theoretical areas such as algorithmic economics, only people who significantly contributed to a paper are actually listed as authors (even though that might not always be the case).
At first glance, ordering by contribution seems to be the fairest solution. Measuring relative contribution, however, is a source of great dispute that can hamper productivity. Studies have shown that authors almost always disagree with respect to their relative contributions. Another problem arises when using a default ordering in case the contribution of all or some authors is roughly equivalent (which is mostly the case in our group). Even if one uses non-alphabetical ordering only if the contribution of one author was significantly larger than those of the others, the asymmetry remains: If the authors are listed alphabetically, the first author may have contributed more. If the authors are listed non-alphabetically, the first must have contributed more.
About two years ago, I had the privilege to write a joint paper with Paul Seymour and asked him to move my name further back because I had contributed less. His reply was that “it’s quite standard to be alphabetical and no-one reads anything else into it (and that is a treasure worth preserving)”. I think that’s a very good point because once non-alphabetical ordering is used, it indicates that some information can be deduced from the author orderings in general. The fact that authors have to be ordered in some way can be seen as an unwanted by-product of the sequential nature of language. Whenever I read papers myself, I never put any meaning into the ordering of authors. This is, however, not the case in general and sometimes researchers are denied awards, fellowships, or tenure because the committee in charge expects them to be “first authors”. In order to avoid this problem, some authors put a disclaimer on their webpage or on their papers, explaining which ordering convention is used and sometimes even listing the contribution of individual authors. An alternative, perhaps more elegant, solution to neutralize the author ordering in CVs or on personal homepages (places where it may actually matter) is to list publications as “<paper title> (with coauthor x and coauthor y)”. Despite all these efforts, author ordering remains a controversial issue simply because the cultures of fields, the principles of authors, and the expectations of readers differ so much.
Other articles on author ordering:
- A statement by the AMS about author ordering. In mathematics, alphabetical author ordering is also known as the Hardy-Littlewood rule.
- A paper in the Journal of Politicial Economy by Engers, Gans, Grant, and King (First-Author Conditions) finds that “it is an equilibrium for papers to use alphabetical ordering whereas it is never an equilibrium for authors always to be listed in order of relative contribution”.
- A recent arxiv paper by Ackerman and Brânzei (Research Quality, Fairness, and Authorship Order) analyzes the “phenomenon that alphabetical ordering is correlated with higher research quality”. They cite several studies providing empirical evidence for this claim.
- Blog posts by Michael Trick and Michael Mitzenmacher.
Turing's Invisible Hand 
I try to use alphabetical author ordering whenever possible. But, in addition to the issues mentioned by Felix, there are issues that arise when collaborating with researchers who use contribution-based author ordering (which happens quite often in my case). If the primary author is me or someone from my group, the contribution-based orderers typically feel uncomfortable about being listed before the primary author. And if the primary author is a contribution-based orderer, then contribution-based ordering is used, which causes inconsistencies.
What I don’t understand is related to something you mention: why does the set of authors have to be indicated sequentially?
A paper could have a set of first authors listed (in arbitrary order), and a group of second authors, etc. In fields where it is normal to use alphabetical ordering, everyong is in Group 1.
I am in TCS where it is normal to use alphabetical ordering. However, there are cases when the author ordering is not alphabetical. In fact, a noticable percentage of papers that go on to win awards are not alphabetical (RSA comes to mind), while in a typical conference accept list, all papers appear to be listed alphabetically. This leads me to believe that people are for the most part ok with alphabetical ordering even in situations where the contributions are clearly not equal, but when the result is very important, they are less happy with this convention.
Also, you say that not using alphabetical ordering leads to “dispute that can hamper productivity”. I have also seen it to be the case that a paper is written with very unequal contributions, the authors are listed alphabetically, a less contributing author later on acts publically as if they were a totally equal contributer and this offends the person who did the bulk of the work, who then does not want to continue the collaboration. If the greater contribution were at least acknowledged even privately, then this person would be more sure they are going to get credit they deserve/need.
I wonder if alphabetical ordering really raises the quality of research for the following reason: Suppose a senior professor works on many problems and gives high level ideas, but does not make specific contributions to a paper. However, due to the convention of alphabetical ordering, they can publish 20 papers per year where it is not possible to see if they made a significant contribution or not. Now, there may be some people who can really write/contribute significantly to that many good papers per year. But most people can not. So then a junior person who only gets his name (alphabetically) on papers for which he did the bulk of the work is scared because he does not see how it is possible to write 20 papers a year and indeed, with a significant contribution, it is not possible even for most good researchers. The point is that this leads to young researchers comparing themselves to research profiles that are not real and getting discouraged.
So while non-alphabetical ordering might lead to immediate disputes, I am not sure it leads to fewer disputes in the long run. Also the actual information about how much a researcher can actually achieve is more accurate with non alphabetical ordering. A senior researcher who has the choice to work on something important vs. become last author on 20 papers may be more motivated to do the former (and the latter, which are possible simultaneously) and produce better work.
If article-level metrics are going to take over as our main measure of researcher productivity, more information on the contributions of each author will have to be available. This may include non-alphabetical ordering, but simply describing what people did may be enough (X had the main idea, Y did experiments, Z did data analysis is seen in some fields). In more mathematical areas, this may be trickier.
I agree that we will lose something in terms of collegiality if we go down this route, which is sad, but I am not sure we can avoid it. Usually I use alphabetical ordering but sometimes I have made a point to list my name first in cases where it was obvious that I have done more (of course, having a surname starting with W, it is easier for me to make such a statement than for Felix).
Maybe we need an extension of value theory – it is hard to see how we could measure whether a particular author is pivotal to a project, but if such a metric worked well, it would presumably encourage further specialization among researchers.
I was very happy to see this entry 🙂 I’m also wondering about the interaction between researchers that use alphabetical and contribution-based ordering (whether they collaborate at a lower rate and in general, about the effects of using both conventions in parallel (such as in the AI community)).
I think that nowadays we should stop using alphabetical order. People often remember the first author only, or the first two. Instead of alpha order we should move to random order. The authors flip a few coins and decide the order accordingly. Then, in a footnote, it is explained that the order of the authors is random. Much more fair…
There are certainly cases where alphabetical ordering seems very unfair, in particular when there are authors who did not contribute anything (and sometimes did not even *read* the paper). However, I see this as more of a problem with authorship rather than with the actual author ordering.
Alphabetical ordering seems to work quite well for us because most of our papers originate from some of us standing in front of a whiteboard and jointly proving a statement. Of course, some people contribute more and some less, but I think it would seriously impair the atmosphere if people were constantly emphasizing how important *their* proof step was or firing out half-baked ideas just to make they were the first to mention it. (We also jointly write down papers using a version control system, often with several people working on the write-up of the same proof.)
In my view, one of the biggest drawbacks of alphabetical ordering is the Harvard “et al.” citation style, which I otherwise like a lot. I am also quite sympathetic towards random ordering and ordering by reverse seniority. However, I’d be reluctant to change the policy for the sake of consistency and because I have the hope that people don’t read anything into the ordering if it is alphabetical.
I like anti-alpha’s proposal. I.e., group the authors into several “blocks,” then use some arbitrary order (e.g., alphabetical) within the blocks.
I think Felix makes a great point in his comment that the “et al” citation style does not work well with alphabetical name ordering. I have been trying to avoid the style, using phrases such as “[15] shows that…”. It’s awkward English, but it does not have the unintended consequence of only associating one person’s name with work done by four people (for example, no one calls the five-author linear time deterministic median algorithm “Blum’s algorithm”).
I think in math and other “theorem-based” disciplines, ranking contributions is often just infeasible. Sure, sometimes someone clearly contributed less. But in many cases even if author A proved just one lemma, that might be the difference between having a result and not having it. That one lemma might be as hard as anything else in the paper. Or should we compare my lemma vs. your lemma? By contrast, in disciplines like biology, the roles played by different authors are usually very clear: e.g. some postdoc came with the plan of attack, she and a graduate student did most work on the experiments, some beginning grad student or undergrad did the more routine experiments, someone did the stats and drew the figures, and the PI blessed all of the above. Any of these roles comes with a well-undersood place in the order; when the major contributors (i.e. those who designed the experiments and were most involved in doing them) are several, they’re often listed alphabetically, with a note that “both authors contributed equally”.
I believe this is among the most vital information for me. And I am glad reading your post. But wanna comment on few general things, The web site style is ideal, the articles is really excellent : D. Good job, cheers
Nice blog. Is it https://wordpress.com platform?