I think about this one a lot, how easy it is for pattern recognition to misfire and synthesize conclusions and connections that are partially rational and correct, but with one bad piece of input data skewing them into fantasy. Sometimes the conclusion may be right, but its magnitude, impact, or applicability are grossly overestimated. Part of what makes such gifted people is the ability to imagine, hypothesize, and use hypotheticals. This giftedness often coincides with the ability to make many parallel what if thoughts and identify the most probable. At some point, hypotheticals morph into faith and what was once a thought experiment becomes belief. What ifs break down into belief "this is actually happening, this is reality." It becomes its own compounding feedback loop, as faulty conclusions layer on top of each other over time.