A Study On Principles Under Which The Preferred Extensions Of A Dynamic Argumentation System Are Stored

Abstract argumentation theory is an emerging research area of logic and artificial intelligence.It can be used to model different kinds of nonmonotonic reasoning,helping agents reason in some inconsistent situations.Because of the dynamic environment of agents,the dynamic problem of argumentation is very important.When an agent is in some dynamic environment,the update of underlying knowledge will change the upper abstract argumentation framework,so the semantics of argumentation system will also be updated.So far,how to characterize this dynamic argumentation system is still a challenging problem.The main reason is that when an agent is updated(e.g.,by adding or removing an argument or an attack relation),it is difficult to determine the arguments whose status is changed and how to reevaluate the status of each argument is also very hard.Meanwhile,because there is no tractable algorithm to compute the semantics of an argumentation system in most situations,when it is changing,reevaluating the status of each argument is not only inefficient,but also difficult.To cope with this problem,this paper studies the principles under which the semantics of a dynamic argumentation system is stored,based on some existing works.It can reduce much unnecessary process of reevaluating,and therefore is significant.So far,there are some classical semantics(grounded semantics,preferred semantics,stable semantics etc.),and foreign researchers only study the principles under which the grounded semantics is stored.Since the computation of grounded semantics can be performed in polynomial time,while the computation of preferred semantics is an NP hard problem,in this paper we focus on the principles under which the preferred semantics of a dynamic argumentation system are stored,and only considers the situation of adding or removing an attack relation.Through such a research,we can get some main conclusions as follows:when we add an attack relation,there are 20 principles under which the preferred extensions are stored,among the other 16 principles,there are 10 principles which are absolutely not,but there are 6 principles which could be if we add some additional constraints;when we remove an attack relation,there are 29 principles under which the preferred extensions are stored,the rest 7 principles could be if we add some additional constraints.Of course,in this article we only make a preliminary research,the much further work can be done.