NPR - Scientific findings often fail to be replicated
How our botched understanding of 'science' ruins everything
The first one talks about a team who tried to replicate the results of 100 published psychology studies. They only got similar results in 39 of them. It talks about how what it most likely means is not that they were "false" but that they didn't totally understand the causes, or conditions needed to be present to get the results.
The second one is an opinion piece/screed. He says "Science is the process through which we derive reliable predictive rules through controlled experimentation" and says that a ton of what is presented as "science", really isn't. He's got a point. A ton of what we do today is math models and use of data, not technically "science". Think Nate Silver. He's using a math model to try to predict an outcome based on data. That's what climate studies are trying to do, too. But neither of these things are technically "science" since they aren't controlled experiments. But I disagree with this guy that this means using data and math in this way has no value.
Anyway, I thought this paragraph was pretty damning:
Take the debate about the impact of raising the minimum wage, or lowering taxes, or any other controversial piece of public policy. We have tons of data that you can find supporting whatever position you want. We could put the debate to rest by doing controlled experiments (like he says they did with welfare reform in the 90s).Our approach to public policy is also astonishingly pre-scientific. There have been almost no large-scale truly scientific experiments on public policy since the welfare randomized field trials of the 1990s, and nobody seems to realize how barbaric this is. We have people at Brookings who can run spreadsheets, and Ezra Klein can write about it and say it proves things, we have all the science we need, thank you very much. But that is not science.
