I'd like to make a crowd-sourced Bayes net of all evidence, hypotheses, and policy proposals. Let people vote on the connections between nodes and their influence, but then let the Bayes net decide what is consequentially true. Most people are able and willing to follow local arguments about individual things, but discard the whole thing when the argument expands to have actual consequences for policy. The goal would be to let people keep making those local judgement calls on evidence and truth, but let the machine handle the consequent implications.