Inductive reasoning is strictly probability-based, which has immediate evolutionary utility "This looks like 40% probability of a Snake in a bush -> Its a snake => RUN away", the brain neural networks evolved to quickly build inductive reasoning chains for speedy judgement of unclear situations with a strong stimulus of increasing chances of survival. If inductive reasoning fails, its failure had likely much less of a negative cost than ignoring the association:
Our pattern recognition engine is essentially biased in favor of finding less probable connections and fast associations from memory(stereotypes/folk beliefs/tradition/religion infused by society), as long as we can build a reduced case justification like the composition fallacy.
Beliefs that involve incomplete picture of reality, are especially prone to it: a flimsy connection chain where very little data exists can bring forth a strong belief system, since it "fills the gaps" without much competion and "feels like solving the problem"( a specific form of
http://changingminds.org/disciplines/argument/fallacies/hasty_generalization.htm is the perfect example of "fast, inductive, instinctual reasoning" that evolution rewards)
E.g."This video has suspicious artifacts" -> "NASA is hiding something because this video is faked"-> "Space is fake" -> "Earth is flat"