Section: New Results
Reasoning with Imperfect Information and Priorities
Participants : Abdallah Arioua, Patrice Buche, Madalina Croitoru, Jérôme Fortin, Souhila Kaci, Namrata Patel, Tjitze Rienstra, Nouredine Tamani, Rallou Thomopoulos.
This work focuses on two main notions, namely argumentation systems, which allow to represent and deal with conflicting information, and formalisms to represent preferences, which allow to compare possible outcomes in decision making and recommender systems.
Fundamental Aspects of Argumentation
A Dung-style argumentation framework aims at representing conflicts among elements called arguments. The basic ingredients of this framework is a set of arguments and a Boolean abstract (i.e., its origin is not known) binary defeat relation on these arguments. This abstract framework can be instantiated in different ways, by representing arguments in a given knowledge representation formalism, which allows to take the semantics of arguments into account in the computation of the defeat relation.
Preference-based argumentation frameworks are instantiations of Dung's framework in which the defeat relation is derived from an attack relation and a preference relation over the arguments. Recently, Dung's framework has been extended in order to consider the strength of the defeat relation i.e., to quantify the degree to which an argument defeats another argument. In this work, we instantiated this extended framework by a preference-based argumentation framework with a valued preference relation. As particular cases, the latter can be derived from a weight function over the arguments or a Boolean preference relation. We showed under some reasonable conditions that there are “less situations” in which a defense between arguments holds with a valued preference relation compared to a Boolean preference relation. Finally, we provided some conditions that the valued preference relation shall satisfy when it is derived from a weight function.
This is a joint work with Christophe Labreuche from Thales and published in 
We also considered an extension to argumentative frameworks based on fuzzy set theory. The knowledge base is fuzzified to allow agents expressing their expertise (facts and rules) attached with grades of importance in the unit interval. Arguments are then attached with a strength score aggregating the importance expressed on their facts and rules. Extensions, corresponding to subsets of consistent arguments, are also attached with forces computed based on their strong arguments. The forces are used then to rank extensions from the strongest to the weakest one, upon which decisions can be made. We have also shown that the strength preference relation defined over arguments is reasonable according to classical rationality postulates and our fuzzy logical argumentation system can be seen as a computationally efficient instantiation of the generic model of structured argumentation framework. We furthered our theoretical research and demonstrate the added value of our approach in the practical setting of the European project EcoBioCap (see Sect. 8.2 ).
One instantiation, among many others, of Dung's framework consists in constructing the arguments from a set of propositional logic formulas. Thus an argument is seen as a reason for or against the truth of a particular statement. Despite its advantages, the argumentation approach for inconsistency handling also has important shortcomings. More precisely, in some applications what one is interested in are not so much only the conclusions supported by the arguments but also the precise explanations of such conclusions. We showed that argumentation framework applied to classical logic formulas is not suitable to deal with this problem. On the other hand, intuitionistic logic appears to be a natural alternative candidate logic (instead of classical logic) to instantiate Dung's framework. We developed constructive argumentation framework. We showed that intuitionistic logic offers nice and desirable properties of the arguments. We also provided a characterization of the arguments in this setting in terms of minimal inconsistent subsets when intuitionistic logic is embedded in the modal logic S4.
This is a joint work with Yakoub Salhi from CRIL and published in 
Lastly, we developed a model of abduction in abstract argumentation, where changes to an argumentation framework act as hypotheses to explain the support of an observation. We presented dialogical proof theories for the main decision problems (i.e., finding hypotheses that explain skeptical/credulous support) and we showed that our model can be instantiated on the basis of abductive logic programs.
This work has been done in Tjitze Rienstra's thesis and published in  .
Use of Argumentation in Explanation, Querying and Decision Making
Besides work on the foundations of argumentation frameworks, we have studied the use of argumentation techniques in various tasks: explanation of query failure, reverse engineering, and decison making. These studies are mainly motivated by agri-food scenarii: bread conception, packaging conception, and durum wheat conception.
We have proposed an argumentation-based explanation for query failure explanation under the inconsistency tolerant semantics ICR in an Ontology-Based Data Access setting with existential rules. We used a rule-based language and we base our work on the equivalence between ICR-based query answering in inconsistent knowledge bases and sceptical acceptance of arguments. We proposed a multilevel explanation that exploits both the inference power of the logical language as well as arguments of dialectical nature. We also investigated an interactive argumentative approach where the process of explanation takes the form of a dialogue between the user and the reasoner.
Within the framework of the European project EcoBioCap http://www.ecobiocap.eu about the design of next generation packagings using advanced composite structures based on constituents derived from the food industry, we have been developing a Decision Support System (DSS) for packaging material selection.  ,  . The DSS consists of two steps: (1) aggregating possibly conflicting needs expressed by several parties involved in the considered field and (2) querying a database of packagings with the resulting aggregation obtained at point (1). We instantiate for each need, called viewpoint or aspect, an argumentation system to reason about arguments solely expressed on it  . This will then be used to generate the query on the packaging database. To this aim we show how to instantiate ASPIC with the DLR-Lite logic modeling expert ontologies in this real world scenario  .
Evaluating food quality is a complex process since it relies on numerous criteria historically grouped into four main types: nutritional, sensorial, practical and hygienic qualities. They may be completed by other emerging preoccupations such as the environmental impact, economic phenomena, etc. However, all these aspects of quality and their various components are not always compatible and their simultaneous improvement is a problem that sometimes has no obvious solution, which corresponds to a real issue for decision making. We propose a decision support method guided by the objectives defined for the end products of an agrifood chain. It is materialized by a backward chaining approach based on argumentation  . An extended version of this paper reporting on experimental results and expert evaluation has been published in Ecological Informatics  .
Knowledge elicitation, representation and reasoning explanation by / to non-computing experts has always been considered as a crafty task due to difficulty of expressing logical statements by non-logicians. We use the COGUI editor in order to elicit and represent argumentation schemes expressed using existantial rules within an inconsistent knowledge base. COGUI is a visual, graph based knowledge representation editor compatible with main Semantic Web languages (see Section 5.1 ). COGUI allows for default reasoning on top of ontologies. We investigate its use for modelling and reasoning using argumentation schemes and discuss the advantages of such representation. We show how this approach can be useful in the practical setting of EcoBioCap where the different argumentation schemes can be used to lead reasoning.
Work published in SGAI 2014  where it received the best technical paper award. In collaboration with Wael Hamdan, Rady Khazem and Ghaisa Rebdawi from the Higher Institute of Applied Science and Technology (HIAST), Syria.
Qualitative and comparative preference statements of the form “prefer to ” are useful components of many applications. This statement leads to the comparison of two sets of alternatives: the set of alternatives in which is true and the set of alternatives in which is true. Different ways are possible to compare two sets of objects leading to what is commonly known as preference semantics. The choice of the semantics to employ is important as they differently rank-order alternatives. Existing semantics are based on philosophical and non-monotonic reasoning grounds. In the meanwhile, they have been widely and mainly investigated by AI researchers from algorithmic point of view. We came to this problem from a new angle and completed existing theoretical investigations of the semantics. In particular, we provided a comparison of the semantics on the basis of their psychological plausibility by evaluating their closeness to human behavior.
This is a joint work with Eric Raufaste from CLLE and published in 
There has been a growing interest in the study of preferences for their utility in solving problems related with decision making. Most of the preference representation languages developed in the literature are based on comparative preference statements since they offer a simple and intuitive way for expressing preferences. They can be further interpreted following different semantics, imparting a greater flexibility on how outcomes can be compared. So far the main objective has been to rank-order the set of outcomes given a set of comparative preference statements and one or several semantics. Tackling this problem from a different angle, we looked into the behavioral aspects of the preference semantics and statements by attempting to formalise the intuition behind them using postulates studied in preference logics and non-monotonic reasoning. We selected the postulates w.r.t. three criteria: coherence, syntax independence and inference. Thus, our analysis provided a means to determine those properties that are satisfied for a given preference semantics.
This work has been done in Namrata Patel's thesis and published in 
Intelligent `services' are increasingly used on e-commerce platforms to provide assistance to customers. Numerous preference elicitation methods developed in the literature are now employed for this purpose. However, it is commonly known that there is a real bottleneck in preference handling as concerns the elicitation of preferences because it does not cater to the wide range of preference representation languages available. Thus, as a first step in developing a decision-support tool using an AI based on such languages, this paper describes a preliminary study conducted to address this issue. We proposed a method of eliciting real-time user preferences expressed in natural language (NL) which can be formally represented using comparative preference statements complying with different semantics, and provided a proof of concept to demonstrate its feasibility. Since we developed NL resources to detect preference semantics, we also made a comparative study with existing resources to underline the peculiarities of our model.
This work has been done in Namrata Patel's thesis and published in