Scientific studies are a first-class source of knowledge. However, the studies are plagued by biases, both when published and when disseminated and disseminated, which slows down progress.

In the field of social sciences we have witnessed major scandals, such as Sokal's. Or more recently, the so-called ya Sokal squared, regarding gender studies. But these problems, to a greater or lesser extent, have been detected in all areas of scientific literature.

Loss of trust in science

In short, the problem of disseminating good quality scientific knowledge is fueled by:

• Publication bias: publishing only positive studies
• Citation bias: citing only positive studies. In 2012, Anne-Sophie Jannot and a team examined 242 meta-analyses published in the Cochrane Database of Systematic Reviews between January and March 2010 that confirmed this.
• Distortion quotes. In 2010, Andreas Stang posted a review of the Newcastle-Ottawa scale, a scale used in meta-analysis to evaluate the quality of observational studies. Sometimes references are simply copied from one document to another. It's hard to know how common this is, but Pieter Kroonenberg, a Dutch statistician, discovered a nonexistent study that had been cited more than 400 times.
• Underutilization of evidence: not citing existing studies. The researchers Karen Robinson and Steven Goodman examined the frequency with which subsequent clinical trial authors reported similar clinical trials. They identified 1,523 essays and tracked how they had cited others on the same topic. Only about a quarter of the relevant trials were cited, which also constituted only about a quarter of the subjects enrolled in the relevant trials.

Within publication bias, in turn, different forms of bias have been identified:

• Time lag bias: in which trials with impressive positive results (large size, statistical significance) are published more quickly than trials with negative or equivocal results.
• Results reporting bias: reporting only statistically significant results or results that favor a particular claim, while other results have been measured but not reported.
• Location bias: publication of non-significant, equivocal, or unsupported findings in less prestigious journals, while studies reporting positive, statistically significant findings tend to be submitted to better-known journals.

In Xataka

Popular studies are less reliable (says a scientific study that is becoming quite popular)

A significant case was the one analyzed in 2015 by Michal Kicinski and colleagues, who examined 1,106 meta-analyses published by the Cochrane Collaboration on the effectiveness or safety of particular treatments. For meta-analyses that focused on efficacy, positive and significant trials were more likely to be included in the meta-analyses than other trials. In contrast, for meta-analyses that focused on safety, "results that provided no evidence of adverse effects were on average 78 percent more likely to enter the meta-analysis sample than statistically significant results that showed they existed." Adverse effects".

The existence of these biases is not only a problem that affects the quality of scientific literature, but is undermining the integrity of science, allowing myths or half-truths to flourish more easily. Once we have diagnosed the problem, it is imperative to get to work to mitigate it and, by extension, reinforce the foundations on which we build the edifice of science.

–
The news

The dissemination of scientific knowledge is plagued by biases and that is a problem

was originally published in

Xataka Science

by
Sergio Parra

.

A milestone has been achieved for planetary scientists and artificial intelligence researchers at NASA's Jet Propulsion Laboratory (JPL): identify craters on other planets thanks to machine learning algorithms.

The image had been captured by the HiRISE camera team aboard NASA's Mars orbiter MRO. AND, thanks to these algorithms, scientists could save the hours they spend each day studying images captured by the MRO.

Mars Reconnaissance Orbiter

In the Mars orbiter's 14 years, scientists have relied on MRO data to find more than 1,000 new craters. But it is hard work. The process requires patience, requiring approximately 40 minutes for a researcher to carefully scan a single image from the context camera.

In Xataka Science

Mars was originally covered in ice sheets.

To save time, we now have the so-called automated fresh impact crater classifier, as part of a broader JPL effort called COSMIC (Capturing Onboard Summarization to Monitor Image Change) that develops technologies for future generations of Mars orbiters.

What takes a human 40 minutes, this classifier did it in 5 seconds.

In Xataka Science

Clear view of rocks formed by running water on a cliff on Mars

To train the crater classifier, the researchers fed it 6,830 context camera images, including those from locations with previously discovered impacts that had already been confirmed through HiRISE. The tool was also fed images with no new hits to show the classifier what not to look for.

The first discovery of a crater made by artificial intelligence took place by exploring around 112,000 images.

–
The news

This is the first discovery of a crater on Mars with the help of Artificial Intelligence

was originally published in

Xataka Science

by
Sergio Parra

.