This article presents an approach to solve the problem of having an action-oriented system such as a robot with eﬀectorial capabilities and a goal to complete. The robots are controlled by software agents based on a BDI architecture that reason via a defeasibleargumentation module (DeLP). This module uses a defeasible logic program in the form of rules that combines desires and beliefs. Therefore, the agent will query about its desires and will derive a single intention, depending on the answers obtained. Finally, the robot will perform an action trying to satisfy the selected intention.
Besides abstract argumentation approaches, differ- ent more concrete argumentation systems exists, spec- ifying a knowledge representation language, and how arguments are built. One of those systems is De- feasible Logic Programming (DeLP - ), a formal- ism that combines results of Logic Programming and DefeasibleArgumentation. DeLP allows representing information in the form of weak rules in a declara- tive way, from which arguments supporting conclu- sions are constructed, and provides a defeasible argu- mentation inference mechanism for determining war- ranted conclusions. The defeasibleargumentation ba- sis of DeLP allows to build applications that deal with incomplete and contradictory information in dynamic domains. Thus, the resulting approach is suitable for representing agent’s knowledge and for providing an ar- gumentation based reasoning mechanism to agents.
Although the World Wide Web is a vast repository of information, its utility is restricted by limited facilities for searching and integrating different kinds of data, as search for queries is mostly syntax- based (e.g., using keywords). The Semantic Web  has emerged as a project intended to create a universal medium for information exchange by giving semantics to the content of documents on the Web. A common way to provide semantics to documents on the web is through the use of ontology definitions. Common problems from common-sense reasoning (e.g., reasoning with uncertainty or with incomplete and potentially inconsistent information) are present when defining ontologies. In recent years, defeasibleargumentation has succeeded as approach to formalize such common-sense reasoning [6, 16]. In this preliminary report, we explore different alternatives for defining an ontology algebra whose semantics is based on defeasibleargumentation and belief revision. We suggest how different aspects of ontology integration can be defined in terms of defeasibleargumentation and belief revision.
Although the World Wide Web is a vast repository of information, its utility is restricted by limited facilities for searching and integrating different kinds of data, as search for queries is mostly syntax- based (e.g., using keywords). The Semantic Web  has emerged as a project intended to create a universal medium for information exchange by giving semantics to the content of documents on the Web. A common way to provide semantics to documents on the web is through the use of ontology de- finitions. Common problems from common-sense reasoning (e.g., reasoning with uncertainty or with incomplete and potentially inconsistent information) are present when defining ontologies. To cope with such problems, during the last decade defeasibleargumentation has developed as a successful approach to formalize such common-sense reasoning [6, 16].
We have shown how DeLP can be combined with DataBase technologies in order to achieve argumentation over a large amount of data. This approach is more efficient than explicit codification of facts because other systems may “give” data to ours without the need of complex interfaces. This can lead to definitions of new architectures for Argument-based Recommender Systems, as well as Desi- cion Support Systems (DSS). Formalism was introduced, an architecture for a framework that allows argumentation over databases has been presented, and the process of obtaining data from databases and using them in argument building was shown. As for future work, there are several lines identified: Semantic-level traduction between predicates and database schemas research is desirable, effi- ciency test with several huge databases needs to be performed, reasoning over databases schemas, mechanisms for resolving non-ground queries have to be de- veloped, and new rules’s learning based on obtained data.
will adopt an extensional approach as it can be easily integrated in the existing ontology, as we will see in the next section. Semantical issues are not discussed in this paper. However, it must be noted that Gabbay's LDS provide a sound basis for de¯ning formal semantics associated to arbitrary logical systems using labelled deduction. 2
Arti¯cial Intelligence (AI) has long dealt with the issue of ¯nding a suitable formalization for commonsense reasoning. Defeasibleargumentation has proven to be a successful approach in many respects, proving to be a con°uence point for many alternative logical frameworks. Di®erent formalisms have been developed, most of them sharing the common notions of ar- gument and warrant. In defeasibleargumentation, an argument is a tentative (defeasible) proof for reaching a conclusion. An argument is warranted when it ultimately prevails over other con°icting arguments. In this context, defeasible consequence relationships for modeling argument and warrant as well as their logical properties have gained particular attention.
Possibilistic Defeasible Logic Programming (P-DeLP)  is a logic program- ming language which combines features from argumentation theory and logic programming, incorporating as well the treatment of possibilistic uncertainty and fuzzy knowledge at object-language level. In this article, we show a prelim- inary approach to reason with possibly inconsistent DL ontologies in P-DeLP. For this we deﬁne the concept of weighted DL ontology which is an ontology whose axioms have given numerical weights indicating their degree of certainty, then the ontology can be interpreted as a P-DeLP program.
Modeling the epistemic state of a rational agent is the most difficult enterprise that must be addressed in its design process. Pereira et al. have endorsed the use of Extended Logic Programming (ELP) for accomplishing this task . In this setting, the knowledge of the agent is codified by a logic program extended with explicit negation, and its beliefs are set by the well-founded semantics of this program. This approach has a clear advantage: it admits a seamless transition between theory and practice. Nevertheless, ELP cannot deal with incomplete and potentially contradictory information, an essential capability in practical agents. Several argumentation formalisms [12,10,8,4,13] have been proposed as knowledge representation and reasoning tools able to handle uncertain information. This property elicits argument-based theories as proper tools for modeling the epistemic state of rational agents. Unfortunately, argumentative systems lack the implementability of ELP.
with clustering techniques, making them more attractive and suitable for solving real-world applications. Argumen- tation provides a sound qualitative setting for common- sense reasoning, complementing thus the pattern classifi- cation process, which relies on quantitative aspects of the data involved (such as numeric attributes or probabilities). Recent research in information technology is focused on developing argument assistance systems , i.e. systems that can assist users along the argumentation process. Such systems provide visual tools which help to keep track of the different issues that are raised and the conclusions that are drawn. We think that such assistance systems could be integrated with the approach outlined in this paper, com- plementing existing visual tools for clustering and pattern classification .
Let us delve a bit into this apparent contradiction (that is, having a sensible framework for KR&R which fails to uphold cut, cautious monotonicity, or both). It can be argued that defeasible argumen- tation has evolved in the last twenty years as a successful approach for modelling commonsense rea- soning. With several formalism reported in the literature reaching a mature state [16, 14, 13, 5, 7, 2], research in this field has played a major role in the development and deployment of solutions to tough, complex real world problems involving varying degrees of common-sense reasoning, such as mediation framework for supporting decision making in groups , intelligent systems used in med- ical research , or providing proactive assistance for natural language usage assessment . We refer the interested reader to the comprehensive survey of applications built around argumentation frameworks compiled by Carbogim et al. .
The importance of using intelligent agents based on mental components like Beliefs, Desires, Commitments and Intentions to solve complex problems is well known in the literature , espe- cially those agents based on BDI theory . Nowadays, tools are needed to specify and programs agents in terms of these components. Several programming Languages and architectures based on BDI are proposed in the literature. However, only some of them allow to specify agents in a declar- ative way. In particular, we are interested in agent development tools that provide an argumentative mechanism for agent reasoning, besides a declarative way to specify its mental components.
become available or information we used to count on with is no longer available or valid). Usually incom- plete information appears in any way of reasoning be- cause its very difficult to represent absolutely all the information related to the objects we count on. As a matter of fact there are systems such as Situation Calculus  where this problem is clear. Any time information about a new entity becomes available we must revise all the axioms on the representation. Ar- gumentative Systems’s devolvement is based on previ- ous research on Logic Programming, Nonmonotonic Reasoning. Argumentation has obtained important re- sults, providing powerful tools for knowledge repre- sentation and some aspects of Commonsense reason- ing. In this sense DeLP  was developed. DeLP is a formalism that combines results of Logic Program- ming and Defeasiblereasoning.
aspects of a generic argument-based framework can be integrated with other ML-based ap- proaches. The paper is structured as follows. First, we briefly introduce the components of most argument-based framework and then outline possible directions for the integration of ML techniques and defeasibleargumentation. Next, we describe a particular setting suitable for the application of such approach, namely text mining problems. Finally, we discuss some promising research lines that are currently being pursued.
Computing warrant, on the other hand, can also be better understood in the light of some logical properties of |∼ T ∗ . Restricted inclusion ensures that any non-defeasible fact in a theory Γ can be considered as warranted. Idempotence indicates that successive applications of |∼ T ∗ . on a the set S of warranted literals returns exactly the same set. From Horn supraclassicality it follows that every conclusion obtained via Sld is a particular case of warranted literal, whereas Horn right weakening indicates that non-defeasible rules behave as such in the meta-level (a strong rule y ← x ensures that every warrant A for a literal x is also a warrant for y). From subclassical cummulativity it follows that two theories Γ and Γ ′
argument in the argumentation line. To see whether B is a circular argument, the set of rules in B must be checked to ensure that it is not a subset of the rules of a previous argument in its argumentative line. Once that B is deemed as valid, it is added into the dialectical tree, and the links to its defeaters are used to determine which hypothetical arguments must be analyzed in the future. This process continues until there is no more defeaters to consider and the construction of the dialectical tree is finished. Then the marking of the tree is made as usual.
The study and development of argumentative frameworks has deserved special atten- tion in this regard, since argumentation constitutes a con°uence point for characterizing traditional approaches to non-monotonic reasoning systems, such as Gelfond's extended logic programming and Reiter's default logic [BDKT97]. In that context, Labeled Deduc- tive Systems (LDS) [Gab96] emerged as an interesting alternative that provides a °exible methodology to formalize complex logical systems.
As the reader has noticed, when an action is executed new literals can be added or deleted from Φ, and therefore, new defeaters could appear or disappear interfering somehow with the assumed warrants. As it was shown in the previous examples, this could cause that the planner selects an improper sequence of actions that can not be used as a plan. In traditional planning the solution is to protect the literals. However, since in this approach we are using a deductive knowledge base, we need to protect the warrant of the literals. A solution is proposed in the following section.