 Today we are going to discuss the topic monotonic and non-monotonic reasoning in AI. At the end of this session, students will be able to demonstrate the difference between monotonic and non-monotonic reasoning. Monotonic reasoning involves techniques for reasoning with a complete, consistent and unchanging model of the world. It is used in conventional reasoning systems. AI reasons to finally deduct certain things which are required in the form of inferences. Information is complete with respect to the domain of interest here. All facts that are necessary to solve a problem are present or can be derived from those that are. The only way to change is to add new facts. For new consistent facts, nothing will ever be retracted. Therefore, this system is consistent. The methods we adopt for monotonic reasoning are represent all your facts in a predicate logic, put them into a particular form which is maybe a clause form and then solve the particular problem with the help of resolution to get inferences. A natural deduction is another form in which we deduce or find out a derived representation to represent our goals of reasoning. Logic programming is another form. Here we may use predicate or propositional logic once again and adopt particular systems so that an adaption is seen in this orientation to provide our goals. We may go for forward and backward reasoning environments wherein we change forward till we find a particular error and then we may backtrack or go backward till we achieve a particular path which is reasonably good. The last form is matching of the particular aspects. It may be matching of predicates, it may be matching of normal forms which we are using in symbolic representations. What do we now expect under non-monotonic reasoning? Non-monotonic reasoning deals with problems with incomplete and uncertain models. They are used to reasonably establish complete, consistent and constant models of the world which is not available. We resolve inconsistencies and this is done by rejection of facts. It is required to observe how revision progresses downwards into a particular environment till reasoning is established. Inconsistent beliefs are singled out and separated. It illustrates problems posed by uncertain, fuzzy and constantly changing knowledge. At a given moment, a statement is believed to be true, false or not believed at all. The key issues for non-monotonic reasoning arise and they are 1. How can knowledge base be extended to allow inferences to be made on the basis of lack of knowledge? Here we need to make a clear distinction between the existing facts which are supplied to us. Any information that depends on the lack of some piece of knowledge is non-monotonic inference. Non-monotonic inferences may be defeated by the addition of new information that violates originally taken assumptions. Therefore, we may have to redo our assumptions if we want to continue these particular aspects. The second assumption is how can knowledge base be updated properly when a new fact is added to the system or when an old fact is removed? Keep track of the proofs or justifications that we have encountered up till our present state. Find all justifications that depend on the absence of a new fact and those proofs that can be marked as invalid. Non-monotonic inferences may be defeated by the addition of new information that violates originally taken assumptions. Now our assumptions have to be modified and we go ahead to reason our particular environment. The third aspect is how can knowledge be used to resolve conflicts when there are several inconsistent non-monotonic inferences that can be drawn? Here we see that normally contradictions are more likely to occur than conventional logic system. This addition adds to more computational power to be used and portions of knowledge base are locally consistent but they seem to be globally inconsistent in many purposes. The methods for non-monotonic reason can be actually felt when we look at the need why we require this to happen and under what conditions this has to happen. We use non-monotonic reasoning to perform default reasoning. We also go to draw conclusions based on what is most likely to be true avoiding the facts that will never be true. The approaches we use are non-monotonic logic. These are systems that provide default reasoning in which the language of the first order logic is augmented with a model operator which is read as inconsistent. When we consider default logic these are new classes of inference rules to be introduced to extend plausible extensions to the knowledge. Non-monotonic reasoning may be of the type abduction that is arriving at a conclusion not licensed by rules of standard logic. This may be wrong but a true guess method may lead us to a near miss to our particular final reasoning result. Inheritance is using values from a prototype to the individual entries that belong to a class. Therefore using inconsistent knowledge or incomplete knowledge accepts these two formulations. A monotonic which will give you a more conventional method of reasoning and a non-monotonic reasoning which will require efforts to develop our particular reasoning. As an example of non-monotonic reasoning let us suppose to take an example that a purchasing agent is investigating purchasing holidays. A resort may be adjacent to the beach or may be away from the beach. This is not symmetric because if the resort was adjacent to the beach the knowledge provider would provide specificness about this particular location. Thus it is reasonable to have a clause if not on the beach implies away from the beach. This clause enables an agent to take an inference that the resort is away from the beach if the agent is not told it is adjacent to a beach. Thus the facts play an important role to show that this combination becomes close to the miss but is not the exact what is required. This is non-monotonic reasoning aspect. For our references we have used the following texts. Thank you.