Friday, February 3, 2023

Physics Needs Better Literature Reviews

One of my favorite physicists, Stacy McGaugh, reacting to a tweet expressing the same opinion by another of my favorite physicists, Sabine Hossenfelder, bemoans a cultural and institutional problem with the fundamental physics community that I agree is a serious one. 

What is it?

Physicists routinely publish papers that fail to review the literature sufficiently to identify the fact that previous published work already rules out, disproves, or contradicts the hypotheses that they are advancing in their papers.

It is a standard and almost universal practice that pretty much every thesis, dissertation, and published physics paper (other than a very short letter preliminarily reporting a very narrow measurement or result before a full length analysis of the results can be published) contains some review of the literature that brings the reader to the point of scientific knowledge where the matters being addressed by the authors in the new thesis, dissertation, or paper begins.

But, in many cases, this literature review is half-hearted and perfunctory, and misses key prior work relevant to the new paper.

For example, one of my pet peeves is when a paper says that their proposal is "well motivated" by concepts developed decades earlier that have later been found to be deeply flawed.

This isn't a "mortal sin". The physics literature is vast and it grows every week. Not everyone in the discipline can devote the time that I do to reading every abstract in a whole range of related fundamental physics categories every day when it comes out on arXiv. And, there are multiple ways of looking at a problem that can make identifying relevant papers challenging (the same issue comes up in doing patent and trademark searches, or searching for precedents related to a legal issue).

But, if you are going to be advancing a new hypothesis in this field, you really should do a proper literature review (and more generally, you should really know the literature relevant to your work from multiple perspectives) before advancing theories that are contradicted by other observational evidence or theoretical considerations that you don't mention or engage with in your paper.

You don't have to agree with everything else that has ever been published. Sometimes previously published papers are incorrect and you are right. But when that happens, rather than ignoring what previously published papers have to say, you really should engage with prior contradictory papers and explain why you think that their observations or analysis is flawed or inapplicable, and thus doesn't actually contradict your work.

You don't necessarily have to spell out the contradictions or flaws of the prior work in full in every new paper in a series of papers developing an idea. It is sufficient to do it once in your first paper identifying what you believe is a flaw in prior work and then to cite that that discussion, incorporating it by reference and with a brief mention, in later papers. But that is very different from ignoring contradictory prior work entirely.

If the authors of physics papers did more diligent and comprehensive literature reviews (and peer reviewers did a better job of insisting on better quality reviews of the literature which would catch both many innocent omissions and many cases where prior contradictory work is willfully ignored), the quality of the papers that did get published would be greater. This is because a lot of speculative garbage papers that ignore known insurmountable obstacles to their work would be dropped before they were presented.


Tom Bridgeland said...

I would expect the peer reviewers to correct some of this. Perhaps I expect too much. Being a peer reviewer is an annoying task, I have been told.

neo said...

Physics Needs Better Literature Reviews

i know of 2 that applied to