Event minimization in the fluent calculus

When drawing conclusions about narratives, minimizing - to a reasonable extent - the occurrence of events is crucial. We argue that unguided minimization is insufficient in case events are causally connected, for it easily fails to distinguish unmotivated event occurrence from those that have a cause. Two solutions are offered, the first of which has the advantage of being straightforwardly realized but on the other hand has a restricted range of applicability. Our second solution overcomes these restrictions but requires two uncommon and novel features. First, event occurrences are identifed as fluents, which allows to adapt a recent causality-oriented solution to the Ramification Problem so that if an event is caused by another event then the former is obtained as indirect effect of what caused the latter. Second, volitional actions and natural events which have no cause inside the reasoning context, are furnished with a special cause, namely, the reaching of the time-point at which they take place. We present both a highlevel narrative description language and an axiomatization based on a novel Fluent Calculus in which is realized this solution to the event minimization problem. This report is an extended version of a paper, entitled 'How (not) to minimize events', which the author has presented at the Sixth International Conference on Principles of Knowledge Representation and Reasoning and which has been published as [Thielscher, 1998a]. Besides some minor corrections, this report additionally contains the general translation of a formal narrative into the Fluent Calculus and a correctness proof. (orig.)SIGLEAvailable from TIB Hannover: RR 7739(98-11)+a / FIZ - Fachinformationszzentrum Karlsruhe / TIB - Technische InformationsbibliothekDEGerman