Non-well founded semantics for belief revision Cecilia Chávez Aguilera. Abstract: Among the literature in belief revision, we can roughly classify it between two main approaches. The classical approach represented in AGM theory, based on a first order logic, suitable for static revision of factual information. And the DEL approach, appropriate for multi-agent learning actions and revision of higher order beliefs. A fusion of the above mentioned theories, can be found in the Baltag and Smets approach. The advantages of the previous approaches, is taken into account here, systematizing several fine-grained distinctions into a unified framework, and the changes induced by the learning actions are emphasized. One of the advantages of having a unified systematic framework is that it sheds light over the specific needs of the logic we want to work with, both at the syntactic and the semantic level. Two of them are the need of taking into account infinitary logics, and the consideration of non-well founded orders in the models used. Infinitary examples of belief revision are not unusual. The Consecutive numbers puzzle and others are a sample of this. Belief revision seen as a learning method requires non- well founded orders. However, non of these features have received enough attention. A non-well founded set semantics seems to us a suitable mean to create a framework which take care of these aspects. Moreover, the links between modal logics and non-well founded sets have received few attention, notwithstanding, the research in this field has shown it is an area worthy to keep studying. In this thesis we give a non-well founded set semantics for the logic L_K◻ and L∞_K◻ developed by Baltag and Smets.