Some puzzle games have a unique solution that can be solved by deduction rather than guesswork (e.g. Slitherlink, Masyu). Using a computer to solve this puzzle it's pretty easy, we can use a backtracking method to find the best solution in second (in general, the puzzle size is not too big).
Is it possible to train a bot to solve this kind of puzzle by deduction?
I think by train it to watch a previous step-by-step solution several times the bot can find some implicit rules/patterns to solve a specific puzzle. Is this possible? are there any references for this method?