There's a problem in classic combinatoric artificial intelligence, where an automated reasoning program (such as the General Problem Solver)[1] gets easily lost among the huge volume of all possible logical paths.
To avoid that pitfall and reach some conclusions, the programmer needs to provide some guiding heuristic that limits the problem subspace where solutions are being searched. But if the provided heuristic is the wrong one, the logic-based solver will still get lost.
I make no claim that the rational brain works in any way similar to an automated solver, but I find it plausible that, when you withdraw your conscious attention, the brain keeps working on the problem using multiple parallel, non-inference-based brain processes that continue to do pattern matching on the problem, without being limited to intermediate consistent steps.
Such non-focused strategy could very well expand the search space considered, and find potential solution matches that weren't near the focus of the previous conscious thinking.
To avoid that pitfall and reach some conclusions, the programmer needs to provide some guiding heuristic that limits the problem subspace where solutions are being searched. But if the provided heuristic is the wrong one, the logic-based solver will still get lost.
I make no claim that the rational brain works in any way similar to an automated solver, but I find it plausible that, when you withdraw your conscious attention, the brain keeps working on the problem using multiple parallel, non-inference-based brain processes that continue to do pattern matching on the problem, without being limited to intermediate consistent steps.
Such non-focused strategy could very well expand the search space considered, and find potential solution matches that weren't near the focus of the previous conscious thinking.
[1] https://en.wikipedia.org/wiki/General_Problem_Solver