Wednesday, October 23, 2013

Science is broken

Science is all about creating reliable and reproducible knowledge.
The Economist has a cover story How science goes wrong.
It is worth reading, pondering, and discussing.

I agree with the general observations of the article. Unfortunately, some of my worst fears are confirmed. Some of the problematic issues that are highlighted have been discussed on this blog before. Problems discussed include:
  • the career pressure to publish leading to a lot of low quality work
  • the pre-occupation with "sexy"new results that can be published in high profile journals
  • poor quality of refereeing, meaning many erroneous papers get published
  • there are few papers about negative results because they are hard to get published
  • there are few papers testing/confirming the results in other papers because they attract little attention
I like the article because it is constructive in proposing reform, particularly from within science, and does discuss various initiatives, including some funded by private foundations to address the problems. The article is not "anti-science", does not lead to postmodern conclusions, or suggesting cutting science funding.

I welcome discussion about the scope of these problems and ways we can address them.

6 comments:

  1. Its easy to point out problems, but hard to come up with solutions.

    I suppose it might be useful to take a Freakonomic viewpoint and ask where the incentives are.

    Oversupply is nothing new. When the space race began and science research & education started gearing up, these downsides of a heavily educated baby boom cohort were probably not appreciated. Their children appreciate it, though.

    Of course, it is easy to be glib about oversupply. Oversupply relative to what? It is not clear that raising science education does not also expand the market for academics also, because more technology also means more posts created for experts. So, the demand and the supply are not really independent.

    One way to correct the problem is probably emerging by itself in the US: oversupply of educated academics will probably be solved by tuition price inflation. Educated people make more money, have more capacity to pay, and thus allow for the price to rise so that more money is needed to educate one's children and reducing the number of people who can pay. Presumably the problem is solved when equilibrium is reached.

    In Australia, who knows? The link between the University as service provider and the student (student's parents?) as customer is much more complicated here, as the system is public. Probably the answer is going to be that the level of education just gets worse while the student volume is fixed by policy. Equilibrium is reached when the science quality (at fixed volume) becomes so bad that no one wants to train as a scientist anymore.

    ReplyDelete
    Replies
    1. I agree that it is a lot easier to point out the problems than to propose constructive viable solutions. But, the articles does mention a few modest ones. In a future post I will make a radical suggestion.

      Delete
  2. I wonder how science worked in the "old good days", like in 1920's and 1930's?

    ReplyDelete
    Replies
    1. Actually, the "good old days" were as recent as the 1960s. There was a reasonable success rate on grant applications. People just published a few papers per year. Physicists and chemists ignored Nature and Science. There were no metrics. There was less hype. There were fewer commercial journals.

      Delete
  3. It seems the nature of the tenure-track system for most well-established researchers of our time is exactly the opposite of what it should be. Tenure should correspond to freedom and the ability to take risks in new research fields without much consequence in the case of failure. Unfortunately, it seems most professors do not see it this way. The majority have straight research paths with very small room for surprise and discovery. Obviously, the safer directions are those which lead to faster, but in most instances mediocre reward. It is a feature of human instinctive nature. In the "good old days" science was perhaps mostly immune to this disease. Now it is as much a part of the scientific enterprise as it is of most corporations.

    b) It is hard to see change in the way. Even those who start with the idea of changing things are in large part drowned by the current system. Most PhD students are familiar with the advice on publishing and finding ways to get cited as most as they can. Those who think they can play the system in exchange for a future with tenure and freedom to pursue research the way they dreamed become disoriented by the time they get to their so-called dream. Several lose faith in science or change their values. It is probably hard not to. As a consequence the ill state of affairs is perpetuated.

    I agree it is too easy to mention problems, but give no solutions, so here is one: from my experience, most graduate students and even post-docs are not where they are because they love science, asking questions or understanding aspects of the universe we live in. Some are where they are because it is not hard to join a graduate program anymore, and while it is not the most financially rewarding job, a research assistant or teaching assistant job is still quite stable. One can argue about the pressures one faces in a post-doc position, but it is not uncommon now to have people being post-docs for 5-10 years. In any case, why not just reduce the number of graduate students? Improvements would be obvious in the quality of mentoring as a consequence of a reduction in the number of people in a group. Another effect would be the dilution of research grants, i.e., there would not be groups with 20 people requiring millions of dollars per year. Perhaps the agencies would distribute more grants for ambitious projects for which it is impossible to give an outline of everything that is going to happen. Scientific quality could also be consistently improved: a smaller number of motivated students would correspond to more inspirational environments. In the "good old days", science was deeply respected by most of its practitioners, who loved it, and seemed to defend its moral values under all circumstances. When a larger fraction of those in science are not interested in its core, in what it represents, the result is what we are seeing now. Now, the suggestion of solution given above is at present nothing more than utopia. Universities will not reduce the number of graduate students entering every year anytime soon. Schools need cheap teaching assistants, advisors with a large number of grants need a large number of students to execute their projects to keep their status quo, their benefits, and so on. Perhaps some order will emerge from this chaos. I sure hope so.

    Now, a question for the author of the blog: the first point raised regards career pressure to publish, but in a previous post I recall a comment about an expectation on your behalf of postdocs to publish at least a single author paper in 6 months. It is hard for me not to see a contradiction in your critique of publication pressure if you have such a view. Since I enjoy very much reading this blog I would like to hear your opinion on this matter.

    ReplyDelete
  4. I read the article in question. Like almost every essay on these issues, it equates "Science" with "Biomedical Research". This seems to be almost universal among STS (ScienceandTechnology Studies) sociologists, because they find so much rich material to complain about in Pharma-funded research; but it is not a fair appraisal of Science in general.

    That being said, I acknowledge that Physics is not immune to similar criticism; just less so.

    The Big Journals are on their last legs. Democratic peer review via the Internet is coming to eat their lunch. But it is vitally important to prevent the latter from degenerating into vanity presses and "Like" contests. The same sort of nuanced, multifaceted evaluations used in responsible peer review must be made available in the democratic version(s). (See e.g. http://opeer.org :-)

    ReplyDelete