Many people claim to love science ,with and without the f-word intensifier, and lament the fact that some large part of the population disagrees with and does not accept ‘science’. I have a real problem with this meme, because it does not acknowledge the enormous range in predictive capability and actual knowledge produced across the many fields of study that claim to be science produced by the scientific method. A perfect example of the trap occurred at the time of the total solar eclipse in the USA earlier this year. Famous media scientist Neil Degrasse Tyson opined that people who trusted science to predict solar eclipse should also trust predictions of global warming and associated bad effects there from.
It is a bogus comparison and a bogus argument and Tyson should know better than to make it. Eclipses can be predicted with models based on linear differential equations that are solvable, and accurate to the limits of observational measurements. Climate models also use differential equations for some of the processes they model. But many critical climate processes are non-linear, chaotic, not well understood and not measured or only recently measured in detail. Examples include precipitation, troposphere hot spot, and cloud formation. Climate models are not validated and will fail a rigorous validation process, just on the issues around tropical hot spot and cloud formation.
But there are much bigger problems in science than bogus claims about the capabilities of climate science. The website five thirty eight has an excellent article on one of the most prevalent forms of bad science, studies based on probability of a null hypothesis being less than 0.05. Most of the social sciences, psychology, nutrition studies, and social justice, political studies are based on null hypothesis p <0.5. In the article they coin a useful term, p-hacking, selective partition of input data can produces spurious correlations. p-hacking can also be called cherry-picking.
At a step below null hypothesis p <0.05 are qualitative studies, which should not be regarded as science, but as speculation and opinion. It would be one thing if qualitative research were confined the social sciences, but it is growing in medicine. When it gets to physics and chemistry, it may be too late to stop the contagion. Action is needed.
Given what is at stake here, I would propose we can not leave science to the scientists, let regulate it like we do other industries. A great low-cost, easy-to-enforce regulation is warning labels. They worked really well with cigarettes. I have helpfully provide a couple of samples. The proposed regulation is:
- All published studies using the null hypothesis p , 0.05 method should carry the warning label saying the study uses methods know to produce spurious correlations.
- All qualitative studies should carry the warning label noting that this study is options and anecdote.
These regulation are low cost for scientists, scientific journals to implement. The benefits will large in comparison. So it you f-ing love science, get behind me here and let’s do this.