January 30, 2014

Co-authorship and careerism in academia

This article by Stan Liebowitz (free download) takes a look at the phenomenon of co-authorship in academic journal articles, concluding that there is too much of it. Why?

Authors are rewarded for the number of items in the publication list on their C.V., and this number is scarcely devalued when the author was not the sole author. When you write an article with one partner, strict prorating would give you 50% of the credit. In practice, each co-author gets about 70-90% of the rewards, esteem, and credit from their department.

So, what's to lose? You can pump out more papers, and you get something close to full rewards in most cases. You may not even have to contribute too much to each one -- if you're really lucky, your name is tacked on spuriously to the author list.

Liebowitz notes that prorating the authorship would solve much of the problem. If you're thinking of co-writing an article with 9 others, and you expect to only get 1/10 of the rewards from your employers that a sole author would, you start to find better uses for your time, instead of trying to maximize the accretion of lines onto your C.V.

He has also measured the rates of co-authorship among the top-ranked economics journals back to the 1940s. The rate is low and fairly constant during the '40s and '50s, and some increase is already apparent by the early '70s, after which it steadily increases.

This fits the phenomenon into the status-striving and inequality cycle. The initial push toward the over-production of elites began circa 1970, as Peter Turchin has estimated by looking at the growth in law school enrollments. Intensified status-striving appears to have been a chain reaction more than a unified wave, beginning with more aspiring elites seeking credentials, and then spreading out toward the lower tiers of the social pyramid (e.g., the higher education bubble that began about 10 years later, circa 1980).

Fitting it into this broader cycle also supports the interpretation of excessive co-authorship as something that is individually beneficial yet socially corrosive. Liebowitz provides a simple model to show how assigning more credit than the person's fair share leads to a level of co-authorship beyond the optimum. The department is now churning out too many co-authored articles, amounting to a lower total contribution to knowledge than if those authors had focused more on their own work (with some co-authorship too).

Before the shift in the social norms toward dog-eat-dog, the prevailing norm was making-do and reining-it-in. Hence the low (but non-zero) levels of co-authorship during the '40s and '50s.

Liebowitz focuses on economics because that's where he feels most comfortable sifting through the old journal articles. But he cites a growing literature on co-authorship that details its broad practice across disciplines.

Most of that literature assumes that the growth in co-authorship is due to increasing specialization, greater complexity of subject matter or mathematical techniques, or something similar that requires more researchers today to form teams and write articles as co-authors. Liebowitz instead draws the natural conclusion from the theory of self-interest that the growth in co-authorship is a form of rising careerism (although he doesn't delve too much into the possible causes for this).

I've heard these kinds of explanations informally -- that because of the fancy-schmancy statistical toolkits out there now, researchers need to write articles jointly with someone who knows what they're doing. Y'know, instead of learning how to do something yourself, or asking for clarification or help when you don't, but not abdicating responsibility and outsourcing the quantitative stuff to someone else entirely.

In fact, those who study co-authorship among legal scholars, such as this article, have reached the opposite conclusion (and no less self-assuredly) -- the growth is due to a greater empirical focus, whereby you need area experts who know the local terrain and can interpret the local language for you. Not that the abstract theory, formal models, mathematics, etc., are becoming bewilderingly complex.

Two firm conclusions about the same phenomenon that directly contradict each other means that something else is going on. Unless we ignore the distinction between abstract vs. empirical complexity, and lump them under a single thing called "complexity," which makes co-authorship more necessary.

The idea that things are just so much more complex than they used to be, is nothing more than a self-serving rationalization. Like, sure, back in the old days when Isaac Newton invented calculus and classical mechanics, when Maxwell invented electro-magnetism, and when Einstein wrote his four Annus Mirabilis papers, it was still possible to write up your ideas on your own. Add on Darwin, Mendel, Fisher, Wright, Hamilton, Trivers, and others from biology. And whoever else you prefer from your own discipline.

But inventing all of modern mathematics and science was just baby stuff. I mean, who hasn't independently invented calculus before dozing off on the couch at night? And quantum physics is one of those things that most of us think up while we're staring at the refrigerator waiting for our tea kettle to boil in the morning. And I've lost count by now of how many times I've integrated Darwinian biometrics and Mendelian inheritance while being put on hold by the power company.

The bogosity of these arguments comes through more clearly when we look at the actual output in the supposedly more brain-bending era of the past 30 or so years. If research these days requires three Einsteins rather than just one, shouldn't the results be even more mind-blowing than the original work was back in its day? Yet how many Earth-shattering theories or empirical patterns have come out in the past 30-odd years that are several times more profound and awe-inspiring than any of those listed above by single authors?

From what I've been exposed to, the closest thing to a Big Deal is the recent body of work on human evolution that has used human genome sequencing. It throws light on our murky early origins, in particular the genetic influence on homo sapiens from related species like the Neanderthals (and bringing to light a species that we didn't even know existed before, the Denisovans). It also shows us how different groups have evolved in recent times, say since the dawn of agriculture, adapting to their local conditions and new ways of life (starchy diets, centralized states).

At the same time, I wouldn't rank this sub-sub-sub-field up there with Fisher's formalization of natural selection and adaptation (oh yeah, plus inventing modern statistics), or Hamilton's theory of kin selection and inclusive fitness.

And I can't include this work in a larger category called "human genome research," because outside of evolutionary results, looking at genomes hasn't taught us anything, when it promised to locate what genes were responsible for cancer, schizophrenia, homosexuality, and other fitness-depressing traits that would have been weeded out a long time ago. It also promised to deliver gene therapy that would help the many people suffering from these conditions. All that work hasn't amounted to diddly squat.

It is hard to escape the conclusion that researchers are too concerned with bloating their publication list in their grab for greater status, sacrificing the focus and originality that comes from serving as the captain of their own ship. If you can put in a decent effort and get fully rewarded, why give it your all?

Soon the norm becomes design by committee, pass the buck, cover your ass, and plausible deniability. Fewer folks follow the norm of stewardship and of tending your flock of ideas until they're good and ready to be sheared or slaughtered.

16 comments:

  1. You're mixing up complexity and difficulty. The great discoveries of the past were usually not complex, just very difficult and thus required one great mind to solve. A modern complex problem may need a collaborative team.

    ReplyDelete
  2. Andrew Gelman had a discussion with Liebowitz over the issue. Gelman was skeptical.

    On the greats of the past, I'm reminded of Tyler Cowen on why there will never be another Milton Friedman.

    ReplyDelete
  3. " The great discoveries of the past were usually not complex, just very difficult"

    Right, nothing complex about mechanics near the speed of light.

    "A modern complex problem may need a collaborative team."

    Be more specific. Remember that the co-authorship trend is as recent as the 1970s.

    ReplyDelete
  4. Also, when you read the typical co-authored paper, you are not struck by how wide-ranging it is. You're like, "Why are there five authors on a simple experimental psychology paper?"

    Random example, not meant as an ad hominem, but to emphasize how ridiculous the world has gotten. Do you remember that evolutionary psychology study from several years ago, showing that strippers who were ovulating earned more in tips than those who were farther away from the peak of their fertility cycle? It was taken as evidence for estrus in humans.

    I'm trying to think of a study that was high-profile so readers will recognize it. That could not have been a simpler study -- one person could have done the lit review, set up the theoretical expectations, conducted the experiment (with some grad student slaves perhaps thanked in the acknowledgements), run the very simple statistical tests, and written it up for submission.

    Instead there are THREE authors on that paper. ("Ovulatory cycle effects on tip earnings by lap dancers: economic evidence for human estrus?" by Miller, Tybur, & Jordan 2007.)

    ReplyDelete
  5. How many authors ought to be on a literature review? One, or more than one?

    Searching Google Scholar for " 'literature review' psychology " gives an endless list of multi-authored lit reviews. The second result has FOUR co-authors, and it's only 25 pages! (Common method biases in behavioral research: A critical review of the literature and recommended remedies, from 2003.)

    The only sole-author lit review on the first page of results (in article, not book form) was written in 1973, before careerism got so far out of hand. (Self-disclosure: a literature review.)

    ReplyDelete
  6. To prepare children for the real world, I suppose they'll have to start teaching kindergarteners how to co-author their book reports.

    ReplyDelete
  7. Of course near light travel isn't complex, we can understand all we need with one particle. Are you sure you know what complexity is?

    "Be more specific."

    Supercollider physics. Your other points were fair though, this does seem to be a trend driven by more than the need for more authors.

    ReplyDelete
  8. That's just quibbling on the meaning of "complexity." But if you're going into complex systems, non-linear dynamics, "uncoordinated coordination," and the like, let's look at non-linear differential equation models in biology.

    Most of the pioneering work was done by sole or at most duo authors. I'd say around half did not co-author the papers, but have been given joint credit for roughly equal contributions to our understanding.

    Lotka and Volterra (predator-prey), separate authors
    Kermack and McKendrick (S-I-R epidemic model)
    Hodgkin and Huxley (neuron firing)
    FitzHugh and Nagumo (simplified H-H, excitable systems), separate authors
    Nicholson and Bailey (host-parasitoid)
    Keller and Segel (self-org of cellular slime molds)

    Then the sole authors:

    Fisher's wave of advance / reaction-diffusion (yep, same Fisher again)
    May's logistic maps and chaos in pop. dynamics
    Turing's chemical basis of morphogenesis
    Conway's "Game of Life" and cellular automata
    Murray's pattern formation of animal coat colors

    Von Neumann and Maynard Smith wrote sole as well as duo works in game theory.

    ...At any rate, it took one or at most two authors to create all of this. And what's happened in the period of rising co-authorship? The same with intellectual life in general -- a great big silence.

    The theoretical ecology and pattern formation stuff from the '70s and early '80s seems to be the last two bodies of work impressive enough to be included in textbooks. It's 30 years later -- where is the new stuff?

    ReplyDelete
  9. Don't say "quibbling", it has a meaning and it's quite relevant. As for your examples, that is theoretical research into complexity. Is the work itself terribly complex? Large scale experiments to test those or find practical results are what I'm thinking of.

    ReplyDelete
  10. You are quibbling when you use an ordinary word in a technical sense without saying so, and expect everyone else to be on board with the technical sense.

    I think you're confusing the scale of teams needed to build the tech that can carry out the experiment, along with technicians to run it, with the scale needed to ask a question, investigate it, and report their results. It may take a huge international team to build the LHC, but it doesn't take hundreds of people to pose a research question that use of the LHC could shed light on, conduct that experiment, and then write it up.

    Same with the human genome field. That took large, international, interdisciplinary teams to build the tech, technicians to make sure it runs smoothly, and so on. But you don't need 10 PhD's to ask, investigate, and report on a question that uses human genome data.

    People want too much credit for "authorship," implying some kind of creative role, when they played a role more like a technician, consultant, etc. They can be gratefully thanked in the acknowledgements section, just like in a book.

    Looking at it that way, I don't think the earlier empirical work on epidemic diseases, population cycles, neuroscience, and so on was less complex than today. It took large teams over long periods of time to build up the necessary infrastructure and technology that the authors of a paper would make use of (research universities, adding machines, electrical instruments, calculus, etc.). Not to mention the huge number of technicians, lab assistants, consultants, and so on who would have to be recruited to study the spread of an epidemic disease in real time, to crunch the numbers (rooms full of secretaries / grad students using adding machines, slide rules, or whatever else).

    But back then, they didn't recognize the "authorship" of everyone but the main authors. They got into the acknowledgements if lucky, and if they were staff or grad students, perhaps not mentioned by name. Now, if an associate professor plays the role of a technician, he wants (shared) author credit.

    ReplyDelete
  11. A sensible solution would be to formalize the different roles and assign credit like they do for movies or recorded music -- director, screenwriter, editor, etc., but where only the director's name appears in the "author" credit ("A film by David Lynch").

    That seems to be what the phrase "corresponding author" is about -- like, "Oh, you mean the real author?" The other roles could be spelled out in the acknowledgements, or listed without mentioning their role if their role doesn't fall into a simple box.

    BTW, what's happened to the frequency of the phrase "corresponding author" in the books in Google's digital library?

    https://books.google.com/ngrams/graph?content=corresponding+author&year_start=1900&year_end=2008&corpus=15&smoothing=3&share=&direct_url=t1%3B%2Ccorresponding%20author%3B%2Cc0

    Flat from 1920 to 1980, and takes off like a rocket afterward. More or less parallels the co-authorship trend, and is a partial solution to distinguishing who the real author is on a typical paper these days.

    ReplyDelete
  12. Tying it all back to widening inequality, these days some technicians are receiving authorship and its rewards, while other technicians are not. The difference? Whether the technician has a PhD, tenure, etc.

    I remember putting in a summer's worth of work coding raw data so that it could be analyzed by my prof, who would use it to write papers. There were several others doing this job, too, writing up weekly lab reports, discussing them at weekly meetings, etc.

    I wouldn't appear in the author line, though, because I was an undergrad and had no status, power, or influence to get my name there (not that I thought of doing so anyway). If an associate prof had done the exact same work that we grunts did, he would've gotten authorship, citations, another line on the publication list, greater ease getting grants, higher chances of getting a raise, etc.

    Liebowitz found that full profs are more likely to co-author than upstarts. It's a way of defending their spot at the top from the intensifying competition from below.

    Back in 1920, neither the student nor the professor would've received authorship for serving as a technician. Nowadays, the students and staff continue to go unmentioned, while the profs have climbed their way out of the acknowledgements and into the author line. The top's getting heavier, while the bottom is absorbing more of the former middle layers.

    ReplyDelete
  13. I don't have any data, but that seems fair enough. Any trends you can see in other status striving times?

    ReplyDelete
  14. The only new work that could really be done would be in new fields, but that is discouraged in this conformist age.
    The most important research would be in interface design, finding ways to make computers easier to use, and a better way to rapidly enter data than hunt and peck typing (slow) or touch typing (impossible).
    Another completely neglected field where progress would be possible is education, which is still stuck in the bronze age.
    Human motivation and self control has also been neglected, as seen by the obesity levels.

    PS: An example of what I mean is that it is INCREDIBLY difficult to comment here, it took almost twenty minutes to log in using Livejournal. It kept saying: Your OpenID credentials could not be verified. I logged into Livejournal earlier, so the cookies should still have been active. It also kept duplicating the text in the box. Is it possible to send comments through email?

    ReplyDelete
  15. I'm not sure if you can send comments through email. If it's that much trouble, I may go back to allowing anonymous comments, but then I'll have to turn moderation on again -- otherwise it'll get over-run by spam. Maybe I'll just include a note under the comment form that says you must choose a nickname if commenting anonymously.

    ReplyDelete
  16. "And I can't include this work in a larger category called "human genome research," because outside of evolutionary results, looking at genomes hasn't taught us anything, when it promised to locate what genes were responsible for cancer, schizophrenia, homosexuality, and other fitness-depressing traits that would have been weeded out a long time ago. It also promised to deliver gene therapy that would help the many people suffering from these conditions. All that work hasn't amounted to diddly squat."

    Not entirely true. It requires statistical correlation studies on large groups to find promising alleles contributing to polygenic traits. Susceptibility to mental disorders like schizophrenia tends to belong to that category. Even if one finds promising candidate genes (like DISC1, for example), how is possible to perform controlled experiments in order to establish mechanisms of causation? Various animal models of schizophrenia are not very useful as a substitute, and research on humans would be clearly unethical. Natural experiments have only limited usefulness.

    ReplyDelete

You MUST enter a nickname with the "Name/URL" option if you're not signed in. We can't follow who is saying what if everyone is "Anonymous."