Monday, January 16, 2017

Science, Perfectionism, and Power

Science needs a lesson on communication. Science is such a perfectionist -- it is so afraid of being wrong or irrelevant -- that it stifles its own progress.

Science's Fears of being wrong and of being irrelevant directly oppose one another. As acknowledged by Jared Horvath, scientists fear that "funding will diminish each time a researcher fails to deliver on grandiose (and ultimately unjustified) claims of efficacy and translatability." Science is afraid of being wrong, but Science is sometimes even more afraid of being lost in the huge crowd of scientists. The result is that false science is published.

That's why sites like PubPeer are VITAL. Science is under huge pressure. But, as stated in an interview with the anonymous developers or PubPeer:

The most important metric is publication in top journals, which determines jobs, grants, everything. This distorts the scientific process toward mostly illusory "breakthroughs" and "high-impact research" at the expense of careful work…
PubPeer is helping scientists retake control of their lives, work, and careers by providing a collective judgment that is independent of and ultimately more important than acceptance by the top journals. That judgment is the expert opinion of your peers.
Now this false science isn't always maliciously published. Unfortunately, statistical error is an intricate part of science. Jeremy Burg points out that, based on a few assumptions, 36% of positive data is actually false. However, scientists frequently get so caught up in the excitement of a novel discovery that they overlook important controls or fail to perform the correct analyses of their data.

Another result of Science's Fears, though, is that negative data is never published. You know what that leads to? That leads to a huge number of scientists stumbling down the same, wrong path, wasting time and resources -- and all because no one wants to admit that their big idea failed.

Listen Science, we are all friends here. I know that there are a good number of very shifty and selfish scientists out there that have hurt you in the past, but the only way that we can phase out those people is to stand together in support and collaboration of each other, through the good times and the bad, in sickness and in health.

If you try something and fail, publish it! I would love to see what you did. Maybe I have insight into a certain aspect of your research that needs to be changed to get positive results. Are you really going to throw in the towel because you were too prideful to ask for help? Maybe I was contemplating pursuing the research that you never published for my PhD dissertation. Are you really going to let me waste 3 years in grad school fumbling around with a project that you already know is doomed? I'm don't think you're a failure because your idea failed. I'm think that you are a productive member of Science if you start a conversation about your concept -- and conversations are about truths, not successes. 

And the irony is that based on the same Bayesian statistics that showed that 36% of positive data is false, publishing negative data is more accurate, as explained by The Economist.

We should work on removing the stigma that you're a failure of PhD if you're not a professor. We could train PhDs specifically to facilitate the statistical accuracy of experiments before and after they are performed. We could benefit from more consultants and specialists of various techniques. We could use science writers who are good at accurately portraying results. Let's diversify our people and their skills and stop looking down at people from "other" disciplines and with "alternative" routes.

The irony of this post, though, is that I gathered these data from the articles provided to me by my statistics professor. I did not look at the original studies, and I did not verify that their published results were accurately controlled or analyzed, like the true Millennial that I am. Read "Identifying and Avoiding Bias in Research" to find out how many issues that can cause.

No comments:

Post a Comment