https://scitechdaily.com/fake-science-is-growing-faster-than-legitimate-research-new-study-warns/
2 Likes
This trend is very troubling to me as well @John_Hemming. The headline is a bit misleading in that legitimate (even if conceptually or methodologically weak) science constitutes the majority of scientific research and publication thus comparing rates of change alone is a bit unscientific. Of course, we created the burgeoning fake science industry by attaching so much money and job security to research and publication. Had we listened to economists along the way, we might have prevented at least some of this “fakeness.” The problem is compounded by the organized systems that gin up and disseminate anti-science.
The larger problem is not that marginal human beings are inventing and disseminating harmful lies for god only knows what purposes. The problem is that the people who gather and report the news – and to a point, those who consume it – are inadequately trained in the foundations of science. There are several lesser problems as well. As one example: those who report on pre-publication legitimate research, not understanding how that system works and that the published findings could easily be reversed of redefined, assuming it is ever published (many are not).
It is not difficult for a reasonably well-trained scientist to discern at least some publications that are likely based on falsified data or methods or otherwise intentionally flawed. Not that long ago, I was intrigued by the novel findings of a particular study and requested the dataset, offering to pay expenses of course. I was turned down on all points. In my opinion, that study – essentially a secret – should not have been published.
In the longer term, I think a combination of heightened standards for transparency as a condition of publication and AI review systems will comprise the most effective solution. Solutions that assign forensic roles to editors or reviewers are unlikely to work. And we should not lose sight of the opposite problem. In some areas reviewers form an unscientific cabal, effectively preventing the publication of breakthrough findings. This weakness is eventually corrected but it can take time.
Practically speaking, when someone asks me for guidance based on weakly reported news of a single study with novel/divergent findings, I tell them to wait until the experts weigh in and perhaps to wait for a fewer more confirmatory studies, conducted by different research teams.
There are lot of issues not least replication problems. I think we will have to move on to a form of post publication review.
1 Like
Right. Most editors exhibit a demonstrated empirical bias against replication studies. The result is that no one want to do them because they don’t get published and/or don’t figure well in promotions. This is but one of quite a few unscientific views embedded in the science culture. A particularly irksome one for me in the area we are discussing is that the RCT and similar should be the goal of all research designs. I have no hard data but I would estimate that well more than half of the medical and health researchers hold this simplistic and demonstrably false view.
I agree with the sentiment but it fails to take account that, once published, a large and intractable dissemination system grants the findings a life of their own that becomes difficult, sometimes practically impossible, to extinguish. The remedy must be a change in publication standards and processes, including those for pre-publication systems.
In that case you absolutely should at least write to the Editor of whichever journal it was published in. Usually the authors have agreed to make data available as part of the publication agreement. But I agree with you that if they are simply unwilling to share it means the published study is probably bullshit and they have something to hide (or the original data does not actually exist).
Another massive problem now will be LLMs and their role in scientific writing. I know for sure that many studies nowadays are written by LLMs, and a good number are peer-reviewed by LLMs too. And the problem is that once accepted and published, they become the “truth” and training material for further LLM learning. So we’ll be building bullshit on top of the assumptions of other bullshit unfortunately.
2 Likes
Elisabeth Bik has become a full-time science sleuth detecting fraud in research publications through the analysis of datasets and images. She deserves our support but, of course, we need much broader initiatives.
Most disappointing to me is the fact that we are now detecting fraud in decades old research that took place during my formative period that had become touchstones for scientific progress. Not a great deal of it, but enough to be sad that some in our community could not uphold standards in their practice. It fuels public distrust beyond what is statistically justified.
A borderline type of fraud in medical research is exemplified by the distortions, and selective omissions and emphasis in research on female HRT. A combination of paternalism and deeply held unscientific biases widespread in the medical community committed countless women to suffering and shortened lives – led without vigor – that science could have granted them. It has taken decades do dig out from under these scientific misrepresentations. Even today, I doubt that half of the practicing OB/GYN docs can correctly interpret the science and prescribe HRT correctly and many are still influenced by the intentionally distorted and thereafter uncritically misinterpreted 2002 Women’s Health Initiative estrogen-progestin trial.
Cognitive psychology (incluidng fMRIs) is especially prone to this effect BOLD signal changes can oppose oxygen metabolism across the human cortex | Nature Neuroscience
I know a scienceblogger who doesnt believe in most cogpsych research