assuming the truism that intelligence is imperfect, what is the best way to minimize the obstacles to achieving a high degree of fidelity in the products of intelligence?