11 dreams for the publishing debate — #6 publishing in smaller increments
One more, one more, one more...
Each new post will start from the top, so scroll down a little if you've read the previous one -- but also check out the first post for some motivation.
These are dreams. Some are realistic---perhaps just around the corner; others are way out there---basically crazy. Some will apply to everyone, others only to some. But all have diversity in mind, diversity in our expectations of who researchers are and what they do.
1. write fewer original-research papers
I know what you're going to say. But hear me out. This is at the core: to enable researchers to publish fewer "new result"-papers.
I believe all major problems brought up in the debate are, at the heart, caused by the immense increase in publications -- but not the global increase, the personal one. You have to publish far too much/big these days to get a half-decent position/grant. Increasing publication numbers did not increase the quality of research or, for that matter, the "quality of life" in our communities.
Instead, the massive inflation is killing us, devaluating everything we do as researchers. More papers mean that our papers are worth less. Having to publish more papers means we produce more research of questionable quality (unintentionally and otherwise). Especially young researchers have to publish for metrics instead of quality. Worst of all, evaluating researchers only by this steam-punkesque output means that the jobs go to people with this one singular skill -- writing the right kind of papers to please the editorial boards of the right kind of journals -- leading to an intellectual monoculture instead of diverse, stable, rich communities. In particular, the pressure works against women and minorities as they often start with and continue to face disadvantages in their careers that make it harder to produce the desired "output" in the desired time frame.
(If you're wondering why I'm not bashing "evil publishers". I don't think they are the problem -- we are. If you are happy with the inflation of papers-in-journals, then big publishers are what you need.)
2. get real credit for surveys, reviews and exposition
Surveys, reviews and expositions are research -- nothing more, nothing less. We live in a time where it is actually more important to write expository work. Why? Because we've optimized our production pipeline so well that there is no shortage of new (looking) results. Yes, you can argue about "relevance" (I won't, but I can't stop you). But if you're serious about all that "research for research's-sake"-talk, then you might notice that we've figured out how to educate people in the tens of thousands every year that do nothing else but churn out result after result.
As Felix aptly wrote: we need to move beyond "new" results. And this means to step back from them and, well, review. Surveys are the glue that holds our fields together, holds our communities together. We do it all the time -- if you read a paper thoroughly, you'll most likely write a review anyway. As we aggregate, work on larger projects or grant proposals these aggregate and easily become surveys and expositions. We need to make all of these public so that the community and the original author(!) can benefit from this enormous creative output that otherwise might only show up in a reading seminar or a journal club.
Mathematicians usually do even more. If we read a paper, we create our own version of the results. Mathematics need these just like music needs different interpretations. We need to give people who are "just" re-writing proofs the credit they deserve, the credit for keeping results alive, accessible and understandable to a wider audience than the original author and referee.
And with credit, I don't mean vouchers for some bookstore. Surveys must be first-class citizens! Surveying and reviewing other peoples work belongs on your CV, it is no less original than your "original research" and every department should have people that are exceptional at it, that are better at surveys, reviews or exposition than "new" research and add to the diversity of a department.
3. get real credit for refereeing
Refereeing is research, just like surveys and "new results". It might seem redundant after dream #2 (and in many ways I would say that's the goal), but given the importance of refereeing right now and the way we do it, it's separate.
Currently, good refereeing seems nearly impossible. With the increase of publications we can't even catch up with what's happening in our own narrowly focused interests -- how, then, can we expect to referee properly? Sam tells his own funny but sad story of refereeing honestly but the problem runs deep. Good referees are hard to find and when you find them, they could be writing a paper instead of refereeing -- and why should they not?
There are many ways we could improve refereeing. We can (and should) split up the different stages of refereeing (reference checking, correctness checking on different levels, opinion gathering etc), we should open pre- and post-publication peer review (and in-between peer review), we should use alternative methods to collect these reviews (instead of review journals that have at most one opinion per paper).
One key of refereeing is forming an opinion -- and voicing it. That's very hard, in particular for young and disadvantaged researchers who fear significant repercussions. But we don't need less people sharing their opinions on other researchers' output, we need more (whether they agree or not); it's a responsibility both to our own research communities and to a larger society.
But all of these are very unlikely to work, if we don't find a way to give referees the credit (and criticism) they deserve. Before we can praise referees we need tools to evaluate refereeing, both in current and future form. Above all, we need to be able to put refereeing on the CV, it's part of the research qualities every strong department should strive for.
4. get real credit for communicating
Communicating other people's work through surveys, exposition and reviews is important. Then there's communicating to students aka teaching. This is an especially sore point for mathematics where undergraduate teaching has become a blunt tool to weed out graduates for other disciplines -- our self-respect seems greatly lacking. And then, of course, there's spreading the word to the wider public.
Have you ever noticed that many of the great researchers are excellent communicators (and teachers and surveyors and referees)? I would go as far as to say that a truly great researcher will be great in at least one of these ways to communicate. Without this, you're only a great something-else. We should cherish not just one ability of our truly great minds, but all of them. Right now we promote researchers according to how their "new result"-output compares to the "new result"-output of the truly great ones. Why are we so one dimensional?
Also, communication goes both ways, so we must listen. Can you imagine a graduate-student-run, "Law Review"-like journal for mathematics? Graduate students are perfect for forcing you to reflect on your research -- we should encourage them to make their thoughts public (including anonymously and pseudonomously). But we mathematicians need to go further. When Bob Moses was speaking at Michigan earlier this year he argued that history will judge us as math literacy becomes a civil rights issue in the 21st century. Are we listening?
Without communication, we risk the longevity of our own research areas because we won't be understood by the next generation, by other areas and by society as a whole. But this means something's gotta give and we need to accept that by giving real credit.
5. sharing all our work every way we can
One battle that most scientists are still fighting -- full self-archival rights -- mathematics has long won. We need to make this happen for everyone. We must use the arXiv, our homepages or (if you must) walled gardens such as academia.edu and researchgate. But we should also embrace more recent alternatives like figshare and github to post all our research publicly -- preprints, notes, lecture notes, expository notes, simply everything. Open notebook science is the key, but it's in its infancy. We need to find ways (many different ones) that work for a larger part of our communities so it becomes easier for people to experiment with it and to make it something even better. It might not be for everyone, but it's something everybody will benefit from.
However, the truth is that even among mathematicians a large group doesn't use the arXiv, let alone keep professional homepages deserving the name. I know that especially older researchers often hesitate because technical issues "are more trouble than it's worth". This is a challenge and we must argue against this and more importantly help to sort out problems, within departments, within small research communities etc to overcome these obstacles. Approach people, ask them why their papers aren't availabe and help them put them properly online. In all other cases: Don't be a Grothendieck.
6. publishing in smaller increments
Have you ever read a paper that seemed to hide its true goal because the author wasn't finished but had to publish something out of the pressure of not having a job otherwise? Have you ever read a paper that made a small but reasonable result look much more than it really was just so that it will make the least publishable unit? Have you ever read a paper that was so badly written that you couldn't make sense of it?
Paradoxically, one way to publish fewer "new results" papers might be to publish more but differently. For scientists this might seem easier, publishing as the data comes in. But even for mathematics we have all those little results -- the small gems, the one-line-proof, the clever counterexample, the good reformulation, the useful shortcut -- all those could be published quickly and openly instead of waiting to find enough to "make it a paper". Just like data, these could be reviewed publicly much more easily and we should get proper credit for doing so (both author and reviewer).
Longer results could (depending on their nature) be published incrementally, with multiple, public revisions. Take the preprint idea one step further and make your writing process public. Use a revision system like github to expose the process. Allow for outside input in your writing process, in your working process. The internet makes us an intricately connected community and we can work together in one big virtual seminar. There are already excellent examples in that area. In mathematics in particular, we have the Polymath project or Ryan O'Donnell's Analysis of Boolean Functions, a text book he's working out as a blog.
But as I've argued Polymath doesn't work for very many people so, again, we need many, many more projects like these so that more people have the opportunity to find a way that works for them.
There's of course a risk -- this could create a lot of noise as incrementally published results implode when data turns out to be flawed, proofs disintegrate and general anarchy rears its head! But I think it's worth the risk. Search technology is constantly improving and good scientific standards should ensure that failed research is marked accordingly. And we have so much to gain! We might be able to finally give credit for failing productively -- the main thing researchers do anyway, we fail and fail and fail until we understand what's going on. Sometimes we have to give up, but why shouldn't somebody else continue for us?
Even if your research implodes, you should get credit and, much, much more importantly, you will help others not to repeat your mistake. Fred once worked on a nice old problem which, after a few months, clearly didn't get anywhere. But he realized that all his attacks had probably been attempted before and so he wrote a pre-print on all the ways to fail, published his code along with it so that people in the future might benefit from his failure. Or as Doron Zeilberger wrote: if only John von Neumann's maid had saved all the notes he'd thrown away each day!
Or to put it differently: the most exciting result in 2011 was having no result about Con(PA).
- Andreas Blass, 2012/06/16
Perhaps even more useful than publishing one’s own failures is publishing a survey that also includes other people’s failures on the same problem. The instance that I know about is a paper by Frank Drake, explaining why the approaches people had tried to solve the singular cardinal problem, by extending Easton’s proof for the regular case, couldn’t work. (This was written just before Silver gave an even better explanation, namely that the result they were trying to prove was false.)
- Peter, 2012/06/16
Thank you for the comment, Andreas. I fully agree, of course, that all errors are useful. It’s a great example. I wished we had a culture of mistakes, now that we have tools that would allow easy dissemination of these.
- Peter, 2012/06/16