Knowledge representation for explainable artificial intelligence

Alongside the particular need to explain the behavior of black box artificial intelligence (AI) systems, there is a general need to explain the behavior of any type of AI-based system (the explainable AI, XAI) or complex system that integrates this type of technology, due to the importance of its economic, political or industrial rights impact. The unstoppable development of AI-based applications in sensitive areas has led to what could be seen, from a formal and philosophical point of view, as some sort of crisis in the foundations, for which it is necessary both to provide models of the fundamentals of explainability as well as to discuss the advantages and disadvantages of different proposals. The need for foundations is also linked to the permanent challenge that the notion of explainability represents in Philosophy of Science. The paper aims to elaborate a general theoretical framework to discuss foundational characteristics of explaining, as well as how solutions (events) would be justified (explained). The approach, epistemological in nature, is based on the phenomenological-based approach to complex systems reconstruction (which encompasses complex AI-based systems). The formalized perspective is close to ideas from argumentation and induction (as learning). The soundness and limitations of the approach are addressed from Knowledge representation and reasoning paradigm and, in particular, from Computational Logic point of view. With regard to the latter, the proposal is intertwined with several related notions of explanation coming from the Philosophy of Science.

Nonmonotonic Reasoning, Expectations Orderings, and Conceptual Spaces

In Gärdenfors and Makinson (Artif Intell 65(2):197–245, 1994) and Gärdenfors (Knowledge representation and reasoning under uncertainty, Springer-Verlag, 1992) it was shown that it is possible to model nonmonotonic inference using a classical consequence relation plus an expectation-based ordering of formulas. In this article, we argue that this framework can be significantly enriched by adopting a conceptual spaces-based analysis of the role of expectations in reasoning. In particular, we show that this can solve various epistemological issues that surround nonmonotonic and default logics. We propose some formal criteria for constructing and updating expectation orderings based on conceptual spaces, and we explain how to apply them to nonmonotonic reasoning about objects and properties.

Thirty years of Epistemic Specifications

The language of epistemic specifications and epistemic logic programs extends disjunctive logic programs under the stable model semantics with modal constructs called subjective literals. Using subjective literals, it is possible to check whether a regular literal is true in every or some stable models of the program, those models, in this context also called belief sets, being collected in a set called world view. This allows for representing, within the language, whether some proposition should be understood accordingly to the open or the closed world assumption. Several attempts for capturing the intuitions underlying the language by means of a formal semantics were given, resulting in a multitude of proposals that makes it difficult to understand the current state of the art. In this article, we provide an overview of the inception of the field and the knowledge representation and reasoning tasks it is suitable for. We also provide a detailed analysis of properties of proposed semantics, and an outlook of challenges to be tackled by future research in the area.

Ontologies in human–computer interaction: A systematic literature review

Human–Computer Interaction (HCI) is a multidisciplinary area that involves a diverse body of knowledge and a complex landscape of concepts, which can lead to semantic problems, hampering communication and knowledge transfer. Ontologies have been successfully used to solve semantics and knowledge-related problems in several domains. This paper presents a systematic literature review that investigated the use of ontologies in the HCI domain. The main goal was to find out how HCI ontologies have been used and developed. 35 ontologies were identified. As a result, we noticed that they cover different HCI aspects, such as user interface, interaction phenomenon, pervasive computing, user modeling / profile, HCI design, interaction experience and adaptive interactive system. Although there are overlaps, we did not identify reuse among the 35 analyzed ontologies. The ontologies have been used mainly to support knowledge representation and reasoning. Although ontologies have been used in HCI for more than 25 years, their use became more frequent in the last decade, when ontologies address a higher number of HCI aspects and are represented as both conceptual and computational models. Concerning how ontologies have been developed, we noticed that some good practices of ontology engineering have not been followed. Considering that the quality of an ontology directly influences the quality of the solution built based on it, we believe that there is an opportunity for HCI and ontology engineering professionals to get closer to build better and more effective ontologies, as well as ontology-based solutions.

Quantum Machine Learning Algorithm for Knowledge Graphs

Semantic knowledge graphs are large-scale triple-oriented databases for knowledge representation and reasoning. Implicit knowledge can be inferred by modeling the tensor representations generated from knowledge graphs. However, as the sizes of knowledge graphs continue to grow, classical modeling becomes increasingly computationally resource intensive. This article investigates how to capitalize on quantum resources to accelerate the modeling of knowledge graphs. In particular, we propose the first quantum machine learning algorithm for inference on tensorized data, i.e., on knowledge graphs. Since most tensor problems are NP-hard [18], it is challenging to devise quantum algorithms to support the inference task. We simplify the modeling task by making the plausible assumption that the tensor representation of a knowledge graph can be approximated by its low-rank tensor singular value decomposition, which is verified by our experiments. The proposed sampling-based quantum algorithm achieves speedup with a polylogarithmic runtime in the dimension of knowledge graph tensor.

On the Maximal Number of Complete Extensions in Abstract Argumentation Frameworks

argumentation frameworks are by now a major research area in knowledge representation and reasoning. Various aspects of AFs have been extensively studied over the last 25 years. Contributing to understanding the expressive power of AFs, researchers found lower and upper bounds for the maximal number of extensions, that is, acceptable points of view, in AFs. One of the classical and most important concepts in AFs are so-called complete extensions. Surprisingly, the exact bound for the maximal number of complete extensions in an AF has not yet been formally established, although there is a reasonable conjecture tracing back at least to 2015. Recently the notion of modularization was introduced and it was shown that this concept plays a key role for the understanding of relations between semantics as well as intrinsic properties. In this paper, we will use this property to give a formal proof of the conjecture regarding complete semantics.

On Cycles, Attackers and Supporters --- A Contribution to The Investigation of Dynamics in Abstract Argumentation

argumentation as defined by Dung in his seminal 1995 paper is by now a major research area in knowledge representation and reasoning. Dynamics of abstract argumentation frameworks (AFs) as well as syntactical consequences of semantical facts of them are the central issues of this paper. The first main part is engaged with the systematical study of the influence of attackers and supporters regarding the acceptability status of whole sets and/or single arguments. In particular, we investigate the impact of addition or removal of arguments, a line of research that has been around for more than a decade. Apart from entirely new results, we revisit, generalize and sum up similar results from the literature. To gain a comprehensive formal and intuitive understanding of the behavior of AFs we put special effort in comparing different kind of semantics. We concentrate on classical admissibility-based semantics and also give pointers to semantics based on naivity and weak admissibility, a recently introduced mediating approach. In the second main part we show how to infer syntactical information from semantical one. For instance, it is well-known that if a finite AF possesses no stable extension, then it has to contain an odd-cycle. In this paper, we even present a characterization of this issue. Moreover, we show that the change of the number of extensions if adding or removing an argument allows to conclude the existence of certain even or odd cycles in the considered AF without having further information.