two different kinds of scientists
Last night, I was talking with a neuroscientist who wrote a popular framework for quantifying animal behavior. He observed that other scientists tend to use his framework in two different ways.
One group of scientists has a specific metric in mind, like the length of a tendon, and the scientists track a few points and quantify that. Another group tracks a lot of points on the animal, then figures out what patterns are present afterwards. According to the neuroscientist, the groups were split 50/50.
I think this is a good illustration of two different ways of doing science: hypothesis-based or data-driven.
The hypothesis way is the classic "scientific method" commonly taught in classrooms. Form a hypothesis then test it. It's powerful in distinguishing competing theories from few observations. However, it's hard to apply in places where theories are incomplete or the phenomenon is too complicated to make predictions.
The data-driven approach flips the order of the steps. Collect the data then find hypotheses. It shines when there is a lot of data for a complex or even relatively unexplored phenomenon. It can generate new theories that would have been hard to imagine. However, it's easy to over interpret some pattern that doesn't in fact generalize beyond your data, particularly in the case of small or biased datasets.
I don't think there's a correct way to do science between these two approaches. The data driven approach often leads the way in theory crafting and then the hypothesis approach takes over to test and refine existing theories.