I'm most interested in the last bit you wrote, because it deals with a few different things.

How much is enough? What is the right kind of evidence?

Going with the useful thing, who decides what is useful? If the number of times I blink per day (or something else, if you prefer) is not useful, does that mean it is impossible to form a scientific theory about it?

I do like the idea that some sort of cause and effect is necessary - that seems to me to fit with the predictive power requirement we discussed earlier. The real trick is in determining what is the cause and what is the effect - this is what many theories try to do (such as the stars being the cause for our personalities and actions or bodies of mass attracting each other). Each has a cause and effect, yet one is deemed pseudoscience and the other science. How do you know individual stars don't have any effect on you, other than that it seems improbable because you see no causal relation? How do you know there is no causal relation?

I don't want to let the psychology part go, though. What if we had machines that could tell us every factor about a human before we collected data from that human. What if we new every factor about every human and took data from every human. With enough data and data analysis, could we ever create a science?