Non-monotonic Reasoning: Why Can More Information Retract an Earlier Conclusion?

Non-monotonic reasoning allows new information to defeat a conclusion that was previously reasonable. If we know that Tweety is a bird, we may tentatively conclude that it flies. If we later learn that it is a penguin, we should withdraw that conclusion. This is not a contradiction: the earlier inference relied on the default that birds normally fly, and the new fact defeats that default.

This differs from strict deduction. When a conclusion follows necessarily from a set of premises, adding another premise does not invalidate the original derivation. Everyday judgement, however, often relies on typical cases, the best available explanation or the current absence of counter-evidence. Non-monotonic reasoning is not licence to change one's mind arbitrarily: a retraction must identify the new information that altered the support.

The concept shows why marking a conclusion as revisable is not weakness. When AI systems, diagnoses and ordinary decisions operate with incomplete information, reliability lies not only in producing an answer, but also in knowing what evidence should make that answer be withdrawn.

https://plato.stanford.edu/entries/logic-nonmonotonic/