Some researchers have analyzed what the summaries of scientific papers say. There is enough “clickbait”

Academic articles, Papers reviewed by pairs published in scientific journals, they are One of the pillars of science nowadays. These articles usually have a more or less defined structure, with introduction, results, conclusions and discussion, in addition to a section dedicated to the methodology used.

An element that never (or practically ever) is missing in this type of articles is abstract.

Abstract It is the term with which a kind of summary of the content of the article is known. It is a key piece that has the objective of serving as a bibliographic guide to those who are looking for a study, so this short text must answer properly to the question What is this article going?

But beyond this basic function, the abstract Often fulfills the function of Summary of the articleincluding information on methods, results and conclusions of the experiment or study.

Many of the scientific articles are limited access, protected by a Paywallthe price of a single article can be several tens of eurosbut summaries are available in open.

Scientific articles, including this short introductory text, are subjected to several editorial and scientific reviews, so it would be expected that the abstracts be faithful representations of what the article and the study carried out. The problem is that sometimes, They are not so much.

In the late 90s, a group of researchers analyzed the existence of discrepancies between the summaries of the articles and their content. The team analyzed more than 260 articles (44 pieces by six scientific relief journals) published in 1996 and 1997. They studied two ways in which these summaries could be incorrect, or by inconsistencies with the body of the article, or by the omission of relevant information.

The results showed variation in the results according to the magazine (they found that between 18% and 68% of the articles presented problems). They concluded, in their own abstractthat the inconsistent or absent data in these summaries were “common, even in the medical magazines of great circulation.” The study was published in 1999 the magazine Jamaone of the publications analyzed in it.

25 years have passed since the publication of the magazine’s study Jama and almost 30 since the publication of some of the articles analyzed. Science has changed a lot in those 25 years. However some subsequent studies They indicate that this problem persists.

In 2016, a group of researchers made a compilation and analysis of the studies carried out in this field. This literature review, published in the magazine BMC Medical Research Methodologyhe found that the median “level of inconsistency found these studies was In 39%although the variability was high: it ranged between 4% and 78%.

Since not all errors are equally severe, this review was fixed on the studies that discriminated against the serious inconsistencies of the milder. They observed that the median in this case was somewhat lower, but still considerable, of 19%.

Subsequent studies, like one posted this year In The magazine American Journal of SurgeryThey continue to show the existence of this trend in scientific literature.

What happens then? Are scientists falsifying your data? Or are we simply witnessing an important accumulation of errors? We know that the summaries of the articles are determinants when receiving quotes of other academic articles and that This metric is key for the evaluation of scientific work for the authors. But the publication of an article may sometimes depend on its results being novel.

That is why there is an incentive to emphasize some results and clarify them later. A non -significant result can cause the editors of the magazine or future readers to lose interest in the article, regardless of the real quality of the study. The call publication bias (which refers to the fact that studies with different results are overrepresented in scientific literature) is also the result of this interest in the novelty.

Clickbait academic

The titles of the articles have also been subject to scrutiny in recent years. Consciously or unconsciously, a striking holder can be decisive when we are more or less interested in a study.

In 2016, A study Posted in the magazine Frontiers in Psychology It echoed this phenomenon. The analysis observed how the way in which the headlines affected were affected within reach of the study.

Gwilym Lockwood, author of the study, analyzed More than 2,000 academic articles And he observed that the titles that enunciated something in a positive frame had better metrics than the average. On the other hand, he also found that the works that resorted to speech games showed a worse performance. The titles containing questions, meanwhile, did not deviate from the average significantly.

The problem of abstractsIt is one of many to which the scientific publishers. Some publishers pressed By scandals of various typesfrom the “Mills of scientific articles”Even the problems with rates charged by publication or access to their contents.

The artificial intelligence It is one of these problems, but perhaps also a potential solution. In recent months, and after the occasional scandal, scientific publishers They have been integrating The artificial intelligence tools in the scientific publication, beyond the work that these tools may have developed in the development of research itself. Artificial intelligence has the capacity, among other things, to generate more “objectives” summaries or to detect and correct possible errors and discrepancies between texts and summaries.

In Xataka | This is how bad science infiltrates the international scientific debate: they are not just the great scandals, more than 50,000 questionable articles are incorporated every year

Image | Sonia Radosz