Well, you couldn't test against nothing without running into those issues, but you could compare strategies against each other, which can be more robust anyway.

So, you tell group A to try strategy 1 for a type of problem, and you tell group B to try strategy 2. You measure as best as possible the relationship between commitment to the strategy, effectiveness, and akrasia and later you compare and see who accomplished the most. You could tell each group that "(1 or 2) is the best method we've found for accomplishing (X/Y/Z) and we'd like you to track your results with it." which would act as a fairly effective blind as long as group A and B don't know about each other.

If you really want to test against a baseline, you could first try to discern what the "common" strategy is for most people and make that your control.