There's a controversy brewing between philosophy twitter and epidemiology twitter about how to do good science. The question boils down to this:

Is it a good idea for scientists to show bias in how they ask scientific questions?

It's an fun debate so lets dig in...

What I think @EpiEllie means is you should construct experiments and collect data in ways that make it possible to falsify your hypothesis. Otherwise, you can become trapped in prison of your own beliefs. https://twitter.com/EpiEllie/status/1341205977127923715

I think @KevinZollman has a different concern. There might be situations where all your data so far says A is true but in fact B is true. He thinks the only way out of this situation is to allow people to test irrational ideas just in case they are right. https://twitter.com/KevinZollman/status/1341431076825276416?s=20

So to summarize:

1. Don't let your data limit your hypothesizing ( @KevinZollman )

2. Don't let your hypothesis completely define your data collection and analysis ( @EpiEllie)

I think principle 1 needs a caveat. First. The hypothesis should probably come from some reasonable place like human intuition.

I think principle 2 should have the caveat that sometimes you're collecting data in order to measure something (i.e. fit a parameter in a model). It's not always about refuting the hypothesis.

In conclusion, I think both principles are true and don't conflict with each other. They are both important ideas that scientists need to keep in mind.

Addendum: A second point I think @EpiEllie is making is that scientists should be clear about their assumptions and they should be open about when they are smuggling their assumptions into the analysis through biased study design!