More Recent Comments

Thursday, July 14, 2016

The seven biggest problems in science

Here's an interesting article about the biggest problems in (American) science: The 7 biggest problems facing science, according to 270 scientists. Most of them apply to science in other countries.

I've added brief comments under six of the headings. Those are MY opinions, not necessarily those of the authors. The comment under #6 is a direct quote from the article.
  1. Academia has a huge money problem.
    There's not enough money to do high quality science, especially basic science.
  2. Too many studies are poorly designed. Blame bad incentives.
    Some experiments are poorly designed. All scientists are under pressure to make their results seem important.
  3. Replicating results is crucial. But scientists rarely do it.
    Replication is important—especially in medical studies—but I think this problem is exaggerated.
  4. Peer review is broken.
    The system (peer review) isn't working well. That doesn't mean there's a better system.
  5. Too much science is locked behind paywalls.
    This was never a problem in the past when you had to go to the library to read science journals. You could photocopy whatever you wanted. Now it's a problem because we want instant access from our laptops.
  6. Science is poorly communicated to the public.
    "But not everyone blamed the media and publicists alone. Other respondents pointed out that scientists themselves often oversell their work, even if it's preliminary, because funding is competitive and everyone wants to portray their work as big and important and game-changing.

    'You have this toxic dynamic where journalists and scientists enable each other in a way that massively inflates the certainty and generality of how scientific findings are communicated and the promises that are made to the public,' writes Daniel Molden, an associate professor of psychology at Northwestern University. 'When these findings prove to be less certain and the promises are not realized, this just further erodes the respect that scientists get and further fuels scientists desire for appreciation.'
  7. Life as a young academic is incredibly stressful.
    This is not just a problem for my younger colleagues. It affects all of us. It affects morale in an academic department and it affects the way science is done.


  1. "This was never a problem in the past when you had to go to the library to read science journals"

    This is only because your experience has only been in large research universities where the libraries have/had all the journals you wanted. It certainly was a problem for people at smaller universities and research institutes (not to mention institutions in developing nations) that only could subscribe to a handful of journals.

    1. Really? Who knew?

      I didn't realize that small universities and research institutes couldn't afford the main journals in their field. My DEPARTMENT always had subscriptions to the major journals as did many individual PI's.

      My current office used to be the departmental library.

    2. Heck, even some universities that don't exactly qualify as small or new like the one I'm at have issues (one of the first two universities that expanded from just being teaching institutions to doing both teaching and research. Oxford remodeled themselves after us!). Of the top 10 rated journals in paleontology here's what we've got access to:
      1) Paleobiology - Yes, paper.
      2) Journal of systematic palaeontology - No.
      3) Journal of Quaternary Science - No.
      4) Palaeontology - Yes, paper.
      5) Palaeogeography, Palaeoclimatology, Palaeoecology - No.
      6) Journal of Vertebrate Paleontology - Yes, paper.
      7) Marine Micropaleontology. - No.
      8) Cretaceous Research. -Yes, paper (but only up to 1983).
      9) Lethaia. - Yes, paper.
      10) Journal of Paleontology. Yes, paper (but only up to 1986).

      So there is no only access to any of the top 10 Journals in the field, paper copies exist for half of them, but only for 3 a subscription has persisted past the mid 1980s.
      It's even worse for anything with a biological focus - where we simply don't have a copy anywhere (one of the biology institute libraries might, but these are at least 30 minute drives - you just jot down references until there is something you really need and then make copies of anything you just thought was interesting).

  2. Even when I was at TIGR, which was hardly a impoverished institute, we only subscribed to maybe a dozen journals. Maybe you could call those the "major" ones, but you would find references to papers in lots of places other than those, and not having access was a problem. Realize that a typical journal costs several thousand dollars a year to subscribe to. The journals you are thinking of that departments and even individuals could subscribe to themselves are very rare (Science, Nature, and the like that make lots of money off of advertising as nearly every scientist at least skims them).

    1. Look Jonathan, I'm not saying that everyone had access to every journal by going to the library. All I'm saying is that 25 years ago there wasn't a great clamor for free access to scientific papers.

      Times have changed. Now we want access instantly whenever we go online. Even at major research universities nobody goes to the library to photocopy an article.

      I get it. I just don't think that lack of free and open access is one of the seven biggest problems in science.

  3. Amen to your seven points. The New York Times has a piece this morning about how hard it is for Ph.D.s to get a tenure-track job. I was lucky to come along in a different era when it was much easier, but now even in my own highly-rated genomics department only about 1/6 of our graduates end up with tenure-track jobs. One problem is that advisors want to have many postdocs to make their lab productive, and don't concern themselves with what happens to the people afterwards.

    A bit over 10 years ago biomedicine succeeded in getting a near-doubling of the NIH research budget. That led to a wave of people training. But it was for positions that mostly wouldn't exist, because there was no subsequent further doubling. Time was, when a postdoc was an opportunity to change your work into a new field, or to a new organism or molecule. And 50 years ago, you got only one year of postdoc work -- more was simply unheard of. Then you went off to your faculty position. It is sad to see the postdoc used as a kind of lab tech, and not even well paid.

    1. Some 80% of post-docs in Germany make less than the federal minimum wage, thanks to a loophole in the labor laws for academic jobs. Most Ph.D. positions are paid in such a way that after the contract runs out students end up significantly below the poverty line since unemployment benefits depend on the last wage (mine just ran out and since I can probably defend in November-December I've got a bit of an issue). To quote from the official at the job agency: "Quit science! There's no bloody future in science", "I would urge you to abandon your thesis. Unfinished graduate work looks better on your CV than a completed PhD. It tells employers you are not crazy" and "If you want employment at the university, consider the custodial services. You'll actually make a living".
      I did leave a minimum wage job, which was 75%, to go to grad school. My income went down. There's one post-doc position in paleontology currently advertised: It's less than what I made in grad school.
      I do not think this is an issue of money. It's an issue of distributing money between materials and people. There are grant proposals that include some equipment and lab consumables that dwarf the personnel costs. But funding agencies know that there's no wiggle room for that X-ray defractometer or that sequencing machine. But if you paid a post doc for 10 hours a week, they would surely still work full time to get enough publications to eventually maybe get a better position. So that's the part of the grant you can cut. And that's the kind of jobs that get offered.

    2. I often wish I was still a student (and a first- or a second-year one too). The uncertainty about the future is absolutely soul crushing and demoralizing.

      I also wish I never knew anything about the overall situation in science. Because the best adaptive strategy is to work as if the future is bright and certain -- then you are optimistic and focused on your work and as a result a lot more productive than you are if what will happen in the future is constantly weighing on your mind. Unfortunately what has been seen cannot be unseen.

      I am quite certain I am not the only feeling that way and this has to be having a very negative effect on science as a whole.

      It probably generates another very unfortunate overall effect too -- the people who rise to the top are precisely the people who we would not want to be there if the system was to be reformed one day, because the people who worry about the system are selected against as worrying about it automatically disadvantages them, while the people who can't be bothered to spend any of their time thinking about such issues are selected for...

  4. On two slightly different tacks...

    Blame bad incentives for poor experimental design? What about poor "designers"? You've often commented on poorly written papers. Why should the authors think more clearly about how to set up experiments than they do when they write up the results?

    Can you explain a bit more about why you think the replication problem in medical studies is exaggerated? I've read some writing for a lay audience that criticizes many studies for insufficient statistical power and says less dramatic effects as further data are accumulated is the inevitable result. Do you feel the problem is exaggerated due to over-dramatization of relatively few bad studies, or for some other reason?

    1. What about poor "designers"?

      That is also a result of bad incentives, just a more indirect one.

      Poor "designers" are a consequence of poor education (at the high school, and especially the undergraduate and the graduate level).

      Now why is science education poor? Because what dominates it is learning facts and recipes, with things like experimental design, data interpretation, and other epistemological issues, i.e. the stuff that separates the real scientist from the technician, being pushed aside.

      Why is that? One reason is that at the lower levels nobody sees teaching these things as important, as the emphasis is on teaching "useful" things, "real-life" "applied" skills, etc.

      And at the higher levels, graduate education in particular, the incentives are stacked directly against developing those skills.

      First, the professors don't really have time to spend teaching graduate students, because they have to write papers and grants (mostly the latter) to keep their labs going.

      Second, it is against the departments' interest to have the students taking classes -- these are not undergraduate students paying tuition, these are graduate students being paid (at least for the first or the first several years) a salary by the department. Time spent taking classes is time not spent at the bench. Which is the real reason the students are there. And the truth is that you don't really need to know that much to be productive at the bench in biology. This is not the case in fields like math and physics, where the level of what is being taught to undergraduates is very low compared to what the cutting edge research deals with. So students have to spend several years taking hardcore classes just so that they can begin to do research. The pyramid of knowledge is tall and has a lot of layers there. In biology it is not very tall but rather very broad. Thus what happens in practice is that grad students in biology take half a dozen classes at most (while doing rotations working at the bench), and are done with it, and those classes are not exactly difficult. If they have been taught experimental design and interpretation as undergraduates, great, if not, well, they can pick it up on their own if they are so inclined. The process of getting their PhD is not likely to help with that directly because:

      Third, students have to learn to navigate the peer-review system to get their papers published and get their PhD. Which is another set of incentives in the direction of creating poor "designers", no need to go into details, we all know what I am talking about.

    2. In biological sciences, at least in evolutionary biology, the technical skills you need to master have increased. I used to claim, maybe with only a bit of exaggeration, that you could clear a room of systematists (taxonomists) by using the word "logarithm". John Harshman will remember far enough back to say how much of an exaggeration this is -- I think not much.

      Nowadays a systematist needs to have at least a rough working knowledge of Markov Chain Monte Carlo methods.

      I have been teaching theoretical population genetics and the theory of inferring phylogenies for many years. I am about to retire and stop doing that. It will be interesting to see whether there is a demand at my university to have someone teach those subjects.

      Even at universities that never taught them, there are now ways people can go off to summer workshops and learn that material. The summer workshops such as the Workshop on Molecular Evolution were founded with the idea that they were stopgap measures, and would be obsoleted as most universities hired people to teach those methods. But they are still going strong now after almost 30 years. Ultimately they might disappear, or change their focus to teach more advanced stuff that is not yet in the university curricula.

    3. It should also be noted that, at least in the U.S.A., that there is now a great emphasis in new biology curricula on teaching scientific methods rather than just asking for regurgitation of facts. I think that the Common Core standards that have come in recently emphasize this.

      This should lead not only to better scientists, but among nonscientists to a better understanding of how science works. So things are not quite as bleak as seems.

    4. Joe, you mistake my age, at least as a systematist. I didn't enter grad school until 1990, and numerical taxonomy was already ascendant. It was the tail end of the war between cladists and pheneticists, both of them accustomed to algorithmic analyses and complicated character coding schemes. And I was at UC. While I was largely successful in avoiding population genetics, I still had to deal with Dave Raup, Mike Foote, and Jack Sepkoski.

    5. @Georgi,

      I suspect that lab meetings in Michael Lynch's group are very difference than the lab meetings you attended as a graduate student. That's got nothing to do with whether the PI teaches formal courses or not and everything to do with day-to-day mentoring.

      I have a colleague at my university who runs a very "successful" operation but nobody in his group ever questions whether alternative splicing is real or not because the PI doesn't encourage such critical thinking.

      As you might imagine, the design of their experiments and the design of their papers reflects this bias.

    6. It depends on the person, of course.

      But I maintain that overall the incentives are stacked in the direction of getting away with as little mentoring and as much exploitation of trainees as possible.

      It's up to the PI to resist and individual PIs may indeed do that (for the record, my graduate advisor was doing a very good job at dedicating time for mentoring and paying attention to her students), but the incentives are what they are and doing that puts them at disadvantage. Thus there are a number of labs where the PI spends all his time writing grants, going to meetings, sitting on committees and consulting, is completely clueless about fundamental aspects of the research that goes on in his lab, sometimes doesn't even know the names of his postdocs, especially if they are not from an English-speaking country, etc. Note that none of this is exaggerated, real-life examples for each of those things do exist.

      Trainees see that and very quickly learn that they are disposable (if they didn't know it already). As a result an environment is created that is precisely the opposite of the stable and nurturing one that produces great people and great discoveries.

    7. @John: I misunderstood how far back you could recollect.

      Anyway, be assured that systematists of the 1960s were not real happy when they enountered a logarithm.

    8. I fully accept that possibility. But I spent the 1960s in grade school through high school, and systematics was not in the curriculum.