Regarding what is predictable, there’s no stopping an academic with a theory. The more “data” which is collected, the easier it will be to find things in it, including those things which support or generate new theories. The larger a dataset grows the closer a theory—almost any theory—can be matched (especially using frequentist methods with its cult of the p-value). Scientists are adept at uncovering supportive evidence, and nearly anything can be said to have been the result of a theory. Scientists are poor at finding contradictory evidence, however, or when finding it, they are slow to acknowledge it, or when they acknowledge it, they are clever at showing how the uncooperative data is an aberration or the exception which proves the rule.