This requires both (1) a belief (I sound pretty Bayesian here, no?) and (2) a clear understanding of how the results of applying a specific analytical tool might refute that belief.

FORBES: Meeting the Big Data challenge: DON'T be objective