If you've ever published a paper, then you know authorship order can lead to tricky situations. The arxiv PDF has both the best and most hilarious solution to this problem I've witnessed.
Nah. The best method is to have a standard for the entire discipline. For instance, math papers almost always have the authors listed in alphabetical order. I say "almost always," because, although I've read more math papers than the average bear, and I've never seen one deviate from the alphabetical order protocol, the Math Overflow link below does mention a couple:
Lol, if that's a joke, it's pretty funny. :-) Also, sorry about being Mr. Joke Explainer below.
If not, you're kinda missing the point. All mathematicians know that this is how it works. If they did not already know it going into grad school, at some point, they will find out.
That means that department-level processes all more or less take into account that this is done. That might mean that you'd need a few solo or at most 2 author papers to have a good tenure case, but nonetheless, it's accounted for.
At the faculty-wide level, math departments do have to communicate this standard to faculty of other disciplines. But, OTOH, it does facilitate a joke I know: at one point, it was widely held that the three best British mathematicians alive were Hardy, Littlewood, and Hardy-Littlewood.
You might know how it works, and be conditioned by years in in the environment into accepting it, but that may not be how it feels if your name is Zappacosta and most of the ideas in the paper are yours.
The trouble there is that the first paper to be published about some research carries more credibility than any following. So, when scientists collaborate, who publishes first?
This looks like a non-problem to me. You have one name on the paper, so any ideas in it which are not credited to someone else via some citation are assumed to be that author's original ideas.
The three authors can publish three papers, each with their own angle on the subject matter, highlighting their ideas, and citing the ideas from the collaborators.
The papers could mutually cite each other, creating a graph, and be published with the same date.
One issue is that the meaning of author order differs depending on field. In some fields the order is already meaningless and authors are just listed alphabetically. In other fields the order is decreasing from most important contributor to least. And in yet other fields there is a system where both ends of the list are important where junior authors (like grad students) are at the start and senior scientists who mentored them at the end. But people who have only worked in one field often think that the rules of authorship in their field are universal when they aren't.
They are trying to solve a systemic issue of how credit is given/shared in academia/research with bad typography.
My gripe with their proposal is, if I print a research paper, then the author information is just useless for looking up a reference. There is information loss/degradation.
Listing names in alphabetical order and including a statement that all authors share equal credit ought to be sufficient, in theory. Changing centuries of practice across thousands of institutions around the globe --- nearly impossible.
A better solution would be a standard like DOI but for authorship. That should be a simple microformat that has the contributor list, role and who the primary author(s) are. Over a period of time, all citations can be migrated to that format with an authorship URL embedded.
Randomize it. Every time a digital copy is loaded the authors should appear in a different order. Then if they're cited, hopefully each citation will have a different ordering and the 'weight' will be evenly distributed.
As an academic, I don't think this would stick as it makes it hard to refer to papers in a common way (if you're chatting with your team).
Referring to something like the "Lin transformers paper" or the "Vaswani attention paper" is easy to get people onto a common ground. If you randomise the authors, you'd need to fall back to full paper titles, and long paper titles aren't particularly evocative.
Randomisation may be fine for citation (particularly numbered citation formats like[0], where the names are only in the bibliography), but even then, you are breaking the flow of thought. If someone references an article like "blah blah(mayford 2020)", then for those familiar with the field you can more easily continue reading without having to break your chain of thought to jump to the bibliography and figure out what paper that was.
edit ----
a further thought... referring by TEAM might be alright, but falls down when you're multidisciplinary. e.g. you might refer to the "Deepmind" paper, or the "Facebook llm paper"...but my area of research is agritech so we work with animal scientists from a different institution. Do we refer to both teams? 'The google/meta paper'? Again, you're likely to just fall back to whatever is the first author or institution in your mind.
> Again, you're likely to just fall back to whatever is the first author or institution in your mind.
And that's exactly the problem we're trying to solve :-) If only the first author is in your mind, then the rest of them aren't getting enough credit.
Next time you're reading a new paper and it's written by the 4th author, are you going to recognize their name and go "Oh, this was written by that author that wrote that other great paper, I should read it!"? Or will you not have a clue who they are because only the first author stuck out?
Disrupting the thought process so that you either need to refer to all of them or their team or the paper by name is the point. If it's written by 2 teams, it's the same problem. You would have to randomize Google/Meta and Meta/Google. Although..well maybe not if it's short enough that it's always referred to in full.
With regards to paper names being too long, we have that problem at my office sometimes too. I just give them cute shorter code names so they're easier to refer to.
How about author's proposing a hashtag for their paper that could be used citing them? For instance something like #transformers_nuerips17 for the attention is all you need paper. It might be also coherent with the social media like twitter authors are using these days.
What's your solution for letting a citation be easily parsed within the flow of text?
E.g. if I was reading an article and saw (Jmole et al 2020)--and was familiar with the field--I could easily continue on with an idea of what your paper contained.
If you cite using numbers (as is common in my field, engineering), then the current paper becomes a bit easier to read, but you lose the flow if you need to jump back and forth to the references (e.g. what is [0]? oh, it's jmole 2020, oh where was I in the methods?) and need to rebuild your mental model.
unique ids and titles. If you're familiar with the field, you've certainly seen paper 24601 cited many times and probably read it yourself. If you are unfamiliar, you look at the title "Effects of harsh prison sentences on recidivism". Works well enough for industry standards, laws, product SKUs, engineering drawings, patents, etc. Hell even libraries work this way.
The only advantage of citing by author is if people are actually completely unfamiliar with the paper but are familiar with other works by the author and just guess that the work is related and of similar quality, but this is problematic both because it's incredibly unscientific to just accept something based on authority, and if your lead author is a grad student with few publications to their name you probably aren't familiar with them anyways.
Looks like it was submitted to the joke conference SIGTBD: sigtbd.csail.mit.edu/#home. The conference was April 7, so they were in fact four days early. (Although not publishing on April 1 does seem like a missed opportunity.)
https://arxiv.org/pdf/2304.01393.pdf