They pick a subject based on prior studies/theories/knowledge, and then they explain what they're going to study and how they're going to define their constructs. But the important method to avoid bias is to present all of your information in your research paper, and distance yourself from the parts experimentation (with research assistants who don't know the goal of the experiment). The important thing that after you've created your hypothesis, the experiment isn't done with the goal in mind. You go back to the hypothesis after you have all of the data and conduct analysis.
Aren't you biasing things by deciding what you are going to study? You lend sexism, racism, whatever credence by assuming it's an actual thing that can be looked for, no? Or would every test begin assuming these ideas may not exist in real life?
The issue I see is you seem to not want people to talk about the sexism they see until we all wait patiently for a scientist to give us research results. Or we can talk about it until people like to hear us talk, then we have to stop.
There is no such thing as just looking at the world without preconceptions and figuring it out.
This was the point I was hoping to arrive at. I guess the issue may be the specific degree of bias. It's ok to be biased enough to think looking for sexism in games may be fruitful but it's too biased to already think it's there.