Chain of thought is basically reasoning as humans do it, the only difference is that unlike humans the model can't see that its output is wrong, abandon a line of reasoning and re-prompt itself (yet).