A Bad Look For Science

Science is facing an image problem

Think about the last time somebody said something false about you. Maybe you had an argument with a loved one. Maybe you had a dispute at work. How did it make you feel? Hurt? Upset? Angry? Now imagine that instead of a person it’s a powerful institution. An entity you’ve never interacted with that knows nothing about you. And they’re not speaking to you or to a person who knows you. They’re telling the whole world. How do you think that would make you feel? In a recent article for National Review Andrew Follett shows one reaction. Follett argues that social science is being weaponised against conservatives. He claims that various studies are driven more by ideology than by science which causes them to conduct research that is not reproducible. This means the results cannot be obtained by another researcher following the same process.

I disagree with several statements made in Follett’s article. For example that social science has a unique problem with reproducibility. I worked in computer science research for close to 6 years and can confirm the reproducibility issue is substantial there. But I agree with the article’s core point. I’ve written several articles about studies that claim to have used statistics to prove that conservatives are inferior to liberals. I’ve also highlighted the impact this kind of research can have on society at large. Follett’s article is a prominent example of this impact.

This is a problem for a society as a whole as it undermines public confidence in science as a discipline. If conservative folks feel that science is a tool to make them look foolish they’ll be less likely to trust it. The field of AI is particularly susceptible to this problem for two reasons. First, the methods used in these studies (regression, p-values etc.) are often used in AI work. Undermining confidence in those methods risks undermining confidence in the AI solutions that apply them. Second, there’s a strong relationship between academia and the AI community. The Bureau of Labor Statistics lists a Bachelor’s degree as the minimum entry requirements for various AI-related roles like data scientist, statistician, web developer, web and digital interface designer. There’s also substantial collaboration between industry and academia on AI projects. A negative perception of science and scientists will translate to a negative perception of AI. If it hasn’t already.

The perception of science as a weapon against conservatives will have harmful consequences for AI in particular, for science in general, and for society as a whole.??