3 ms·

Publishers unite to tackle doctored images in research papers

43 points by sohkamyung 5y ago

Ambolia 5y ago
Current science seems like a real mess of wrong incentives living from past times prestige and being used as a opinion-laundering scheme in the worst cases. I don't really trust any "studies" anymore unless they are successfully actualized in the real world, either through accurate predictions, or through engineering projects that actually build something.
- inglor_cz 5y ago
  At the end of the day, trust attaches to people more than to processes / institutions, especially if the institutions are too large. I trust articles co-signed by Katalin Karikó, for example. She seems to be a very dedicated researcher. But this kind of trust cannot be easily extended to any random professor of Acme U.
bongoman37 5y ago
This is way needed particularly in Biological Sciences and Medicine. It might be a good project to build a tool to go through a lab's papers in the past and see if images have been reused, or relabeled and identify anomalies.
yummypaint 5y ago
I was very surprised only 20% of papers had any image alterations. I suspect the vast majority of this is necessary due to including photographs. It's essentially impossible to make photos appear properly without messing with the levels. I also wonder about false positives from things like changing a chart legend label with GIMP at the last minute instead of regenerating the figure.
einpoklum 5y ago
I find this article to be problematic: 1. It claims 20% of images in articles are digitally altered, but immediately that this figure is essentially irrelevant - they count things like increasing contrast for better print visibility. There is no information in the article about regarding the fraction of papers in which photos were altered illegitimately. 2. The article seems to offer conflicting "levels" of alteration. First, level 1 is said to be the harmless alterations, but then it is described as splicing images from different experiments. 3. The article claims that there is a "the growing problem of image manipulation in scientific papers". Is there? We've already observed no information is provided about the breadth of the problem at the moment (point (1.)); and - there is no hint of indication of this having somehow gotten worse recently.

43 points by sohkamyung 5y ago

Current science seems like a real mess of wrong incentives living from past times prestige and being used as a opinion-laundering scheme in the worst cases. I don't really trust any "studies" anymore unless they are successfully actualized in the real world, either through accurate predictions, or through engineering projects that actually build something.

inglor_cz 5y ago
At the end of the day, trust attaches to people more than to processes / institutions, especially if the institutions are too large. I trust articles co-signed by Katalin Karikó, for example. She seems to be a very dedicated researcher. But this kind of trust cannot be easily extended to any random professor of Acme U.

bongoman37 5y ago

This is way needed particularly in Biological Sciences and Medicine. It might be a good project to build a tool to go through a lab's papers in the past and see if images have been reused, or relabeled and identify anomalies.

yummypaint 5y ago

I was very surprised only 20% of papers had any image alterations. I suspect the vast majority of this is necessary due to including photographs. It's essentially impossible to make photos appear properly without messing with the levels. I also wonder about false positives from things like changing a chart legend label with GIMP at the last minute instead of regenerating the figure.

einpoklum 5y ago

I find this article to be problematic: 1. It claims 20% of images in articles are digitally altered, but immediately that this figure is essentially irrelevant - they count things like increasing contrast for better print visibility. There is no information in the article about regarding the fraction of papers in which photos were altered illegitimately. 2. The article seems to offer conflicting "levels" of alteration. First, level 1 is said to be the harmless alterations, but then it is described as splicing images from different experiments. 3. The article claims that there is a "the growing problem of image manipulation in scientific papers". Is there? We've already observed no information is provided about the breadth of the problem at the moment (point (1.)); and - there is no hint of indication of this having somehow gotten worse recently.