Non-monotonic logic is a family of formal logical systems designed to capture forms of reasoning where conclusions are tentative and can be retracted upon the acquisition of new information. As a research field and methodological approach within artificial intelligence and logic, it investigates how intelligent agents can reason effectively with incomplete knowledge, default assumptions, and changing environments, contrasting sharply with classical monotonic logic where the set of valid conclusions only grows with added premises. Its key characteristic lies in the invalidation of prior inferences by new data, making it significant for modeling commonsense reasoning, planning, diagnosis, and other domains requiring belief revision and reasoning under uncertainty.