Author: Patrick Chi Ket Toh
Artist: Fion Lam
Editor: Caleb Scutt
In 1942, an American Sociologist, Robert Merton introduced the 4 values for ethical Scientific Research:
- Communalism: Scientific research should be shared freely among the scientific community and the wider public.
- Universalism: Any race, gender, or ethnicity is allowed to participate in science
- Disinterestedness: Scientists shouldn’t use science to pursue fame, money, social and political reasons.
- Organised Skepticism: New scientific knowledge should be initially viewed through a sceptical lens.
Unfortunately, the scientific community today hasn’t demonstrated that. I will address more than just fraud; I’m going deeper into flaws that the scientific community wants to be swept under the rug.
Many Scientists from Jan Henrik Schon, who manipulated data on his work on superconductors and Woo Suk Hwang, who manipulated images of stem cell lines for cloning to Andrew Wakefield, were incentivised by a lawyer to fake data linking MMR vaccines with autism in children. All three broke the value of disinterestedness. Science is based on trust, and they broke that trust, affecting the lives of PhD students and patients for selfish reasons as well as having dire consequences, for example, Andrew Wakefield’s work set back the eradication of measles around the world.
Publication bias is the weird uncle of the family–everyone knows they exist but don’t want to address it. Franco et al found that 65% of studies with null results didn’t get written up. The competitive nature of scientific publishing pushes scientists to the limit. Not helped by the selective nature of high-impact journals; a study looked at 20,538 manuscripts from 2013 to 2018 published in the journals Cell and Cell Report found more novel studies were likely to be accepted compared to those with less novelty This can create a cascade of data manipulation, ‘legal’ of course. P-hacking is when scientists re-run the analysis of their experiments, taking existing data and running statistical tests without a hypothesis until a p-value of 0.05 is found, meaning a significant result.
Furthermore, P-hacking can contribute to sloppy studies being published. In 2015, a survey looking at animal studies for basic study designs by Macleod et al found only 25% had randomised their groups, P-hacking isn’t the only reason though, another reason is statistical power, how likely the effects found in the study were true. Many studies have been found to have low statistical power; a 2013 review found many neuroscience studies had used a smaller sample size than required leading to exaggerated effects. Compounded by scientists’ reluctance to share data, this breaks the rule of communalism and organised scepticism leading to thousands of underpowered studies.
Why is this all important? Why should we care? Because many of these studies are a lifeline for many patients, especially in biomedical research. Florian Prinz et al. at the pharmaceutical company Bayer, found only 20% of biomedical studies were successfully replicated. Many patients and their respective families are dependent on the scientific content, hoping one day a breakthrough in science can cure their loved ones. Additionally, doctors rely on studies to prescribe drugs, if studies are not proper and a drug isn’t found to have any actual effect, this can be detrimental to both the patient and the doctor. Even the unaffected are caught in the crossfire, $28 billion of tax poured down the drain.
How could we fix this? AI algorithms are being tested to spot fake data or identify errors and plagiarism in papers before they are even published, which can lead to a more scrutinising view during the editing process. Journals are becoming more open to publishing papers with null results. For P-hacking, it’s a tricky situation; solutions have been proposed, like changing the p-value from 0.05 to 0.005, increasing the hurdle for significance, or taking analysis out of scientists’ hands entirely. Finally, an innovative approach of setting up a database to publish drafts of papers to allow open access for scientists and the wider public to prevent any backdoor dealings.
It’s high time we change the way we view science, not as a flawless beacon of truth, but as a system filled with flaws. Millions worldwide rely on us, so we must look inward for the betterment of the community and science.
kinesismagazine.com 