Search

In the digital era, recommender systems (RS) have become an integral part of our daily interactions, exerting a significant impact on users and society. However, this also raises ethical challenges related to RS that should be considered. Addressing these challenges requires the application of explainable artificial intelligence (XAI) models to make RS more understandable. Based on the current state-of-the-art literature, this paper aims to provide a comprehensive overview of XAI for RS and its ethical implications, with the aim of proposing a research agenda for ethical RS based on XAI. The findings of the literature review show that neural network-based RS have received much attention in terms of offering explanations, while there is a research gap in explaining context-based RS and in evaluating explanations. In addition, a set of ethical challenges for RS are discussed by exploring how explanations for recommendations can contribute to the ethical use of RS. [ABSTRACT FROM AUTHOR]

The paper examines teacher pragmatic acts (verbal and embodied) in response to children's crying in a Russian preschool. It focuses on crying episodes associated with children's inability or unwillingness to follow the norms and expectations surrounding children's conduct (e.g., crying when not allowed to do something). The paper employs discourse analysis; the data are 40 h of video- and audio-recorded interactions. The analysis shows that teachers used directives and other pragmatic acts (e.g., assessments, explanations) that encouraged children to stop crying. The directives were produced in multi-party participation frameworks where peers were referred to or recruited as active participants. Teachers used positive and negative assessments of crying. Crying was presented in a negative light in contrast with positive non-crying behaviors; positive assessments (praise and affectionate touch) were offered when crying stopped. The consequences of crying were presented as negatively affecting others. It is argued that such discursive treatment of crying contributes to socializing children to conformity. The paper contributes to our understanding of how through the use of pragmatic acts, child crying is discursively treated in ways that invoke norms of pragmatic conduct (e.g., not crying when asked to participate in a planned activity) and social values regarding group membership. • Analysis examines teachers' pragmatic acts (verbal and embodied) in response to crying in a Russian preschool. • Methods utilize discourse analysis. • Findings explore teachers' use of directive trajectories and other pragmatic acts that encouraged children to stop crying. • Conclusions are discussed in relation to norms of pragmatic conduct and social values. [ABSTRACT FROM AUTHOR]

• Derivation of an alternative method for the interpretation of deep learning models. • Design a human-expert level deep learning diabetic retinopathy classification model. • Design a diabetic retinopathy interpretable classifier. • Visualization of typical feature and input space score map samples. In this paper we present a diabetic retinopathy deep learning interpretable classifier. On one hand, it classifies retina images into different levels of severity with good performance. On the other hand, this classifier is able of explaining the classification results by assigning a score for each point in the hidden and input spaces. These scores indicate the pixel contribution to the final classification. To obtain these scores, we propose a new pixel-wise score propagation model that for every neuron, divides the observed output score into two components. With this method, the generated visual maps can be easily interpreted by an ophthalmologist in order to find the underlying statistical regularities that help to the diagnosis of this eye disease. [ABSTRACT FROM AUTHOR]

In this paper, we present the formalization and implementation of a decision framework in which a group of agents can express their preferences among a set of alternatives by using several comparison criteria. We define special kind of argument to support agents' preferences, and the set of generated arguments is used both to obtain the set of selected alternatives and to explain why they were selected. We introduce a software application that implements our proposed approach and provides interactive tools for obtaining different kinds of explanations for the computed results. Also, we formally study the proposed argumentation process and we prove that the set of warranted conclusions, which support the obtained results, is consistent. We also show that our framework works properly on choice scenarios with incomplete evidence, and we propose how to optimize the computation process for dynamic environments. [ABSTRACT FROM AUTHOR]

Dung's Argumentation Framework (AF) has been extended in several directions, including the possibility of representing uncertainty about the existence of arguments and attacks. In this regard, two main proposals have been introduced in the literature: Probabilistic Argumentation Framework (PrAF) and Incomplete Argumentation Framework (iAF). PrAF is an extension of AF with probability theory, thus representing quantified uncertainty. In contrast, iAF represents unquantified uncertainty, that is it can be seen as a special case where we only know that some elements (arguments or attacks) are uncertain. In this paper, we first address the problem of computing the probability that a given argument is accepted in PrAF. This is carried out by introducing the concept of probabilistic explanation for any given (probabilistic) extension. We show that the complexity of the problem is FP # P -hard and propose polynomial approximation algorithms with bounded additive error for PrAFs where odd-length cycles are forbidden. We investigate the approximate complexity of the related FP # P -hard problems of credulous and skeptical acceptance in PrAF, showing that they are generally harder than the problem of computing the probability that a given argument is accepted. Next we consider iAF and, after showing some equivalence properties among classes of iAFs, we study iAF as a special case of PrAF where uncertain elements have associated a probability equal to 1/2. Finally, given this result, we investigate the relationships between iAF acceptance problems and probabilistic acceptance in PrAF. [ABSTRACT FROM AUTHOR]

Abstract: Our interest is in developing embodied cognitive systems. In the majority of work on cognitive modeling, the focus is on generating models that can perform specific tasks in order to understand specific reasoning processes. This approach has traditionally been exceptionally successful at accomplishing its goal. The approach encounters limitations, however, when the cognitive models are going to be used in an embodied way (e.g., on a robot). Namely, the models are too narrow to operate in the real world due to its unpredictability. In this paper, we argue that one key way for cognitive agents to better operate in real-world environments is to be able to identify and explain unexpected situations in the world; in other words, to perform explanatory reasoning. In this paper, we introduce a framework for explanatory reasoning that describes a way for cognitive agents to achieve this capability. [Copyright &y& Elsevier]

Highlights • Decision support system for applications with multiple comparison criteria. • Users' preferences configurable via conditional rules. • Decision framework based on an argumentative system. • The system provides detailed explanations for the selected alternatives. • Empirical analysis to evaluate the performance of an implemented version. Abstract In our current society, it is very common to have access to a huge amount of information and scenarios in which it is possible to choose among a great number of alternatives. Therefore, systems that automate or assist with the decision-making process are becoming crucial. Those systems should allow users to represent their preferences accurately, even in scenarios where the user has many forms of comparing the available alternatives. In this paper, we propose an argument-based framework for multi-criteria decision-making based on conditional rules, that allows users to encode their preferences among the comparison criteria in terms of requirement and tolerance values. From these rules, arguments that are involved in the comparison process are generated. In order to comply with users' demands for explanations of the given answers, our proposed framework can provide qualitative explanations about the argumentation process, and also an explanatory graph that shows how all the alternatives were ranked and the connections with the user's rules that were used. We also show the results of an empirical analysis of the performance of a developed prototype. [ABSTRACT FROM AUTHOR]

In this paper, we present a fundamental framework for defining different types of explanations of AI systems and the criteria for evaluating their quality. Starting from a structural view of how explanations can be constructed, i.e., in terms of an explanandum (what needs to be explained), multiple explanantia (explanations, clues, or parts of information that explain), and a relationship linking explanandum and explanantia, we propose an explanandum-based typology and point to other possible typologies based on how explanantia are presented and how they relate to explanandia. We also highlight two broad and complementary perspectives for defining possible quality criteria for assessing explainability: epistemological and psychological (cognitive). These definition attempts aim to support the three main functions that we believe should attract the interest and further research of XAI scholars: clear inventories, clear verification criteria, and clear validation methods. • We propose a framework for defining different types of explanations of AI systems. • We contextualize current XAI discourses within the proposed framework. • We highlight two broad perspectives for defining quality criteria for explainability. • We discuss the relevance of our framework in light of current and upcoming AI regulation. • We confer fundamental aspects for future research of XAI scholars. [ABSTRACT FROM AUTHOR]

This paper provides an extensive overview of the use of knowledge graphs in the context of Explainable Machine Learning. As of late, explainable AI has become a very active field of research by addressing the limitations of the latest machine learning solutions that often provide highly accurate, but hardly scrutable and interpretable decisions. An increasing interest has also been shown in the integration of Knowledge Representation techniques in Machine Learning applications, mostly motivated by the complementary strengths and weaknesses that could lead to a new generation of hybrid intelligent systems. Following this idea, we hypothesise that knowledge graphs, which naturally provide domain background knowledge in a machine-readable format, could be integrated in Explainable Machine Learning approaches to help them provide more meaningful, insightful and trustworthy explanations. Using a systematic literature review methodology we designed an analytical framework to explore the current landscape of Explainable Machine Learning. We focus particularly on the integration with structured knowledge at large scale, and use our framework to analyse a variety of Machine Learning domains, identifying the main characteristics of such knowledge-based, explainable systems from different perspectives. We then summarise the strengths of such hybrid systems, such as improved understandability, reactivity, and accuracy, as well as their limitations, e.g. in handling noise or extracting knowledge efficiently. We conclude by discussing a list of open challenges left for future research. [ABSTRACT FROM AUTHOR]

Abstract: In decision-making problems under uncertainty, a decision table consists of a set of attributes indicating what is the optimal decision (response) within the different scenarios defined by the attributes. We recently introduced a method to give explanations of these responses. In this paper, the method is extended. To do this, it is combined with a query system to answer expert questions about the preferred action for a given instantiation of decision table attributes. The main difficulty is to accurately answer queries associated with incomplete instantiations. Incomplete instantiations are the result of the evaluation of a partial model outputting decision tables that only include a subset of the whole problem, leading to uncertain responses. Our proposal establishes an automatic and interactive dialogue between the decision-support system and the expert to elicit information from the expert to reduce uncertainty. Typically, the process involves learning a Bayesian network structure from a relevant part of the decision table and computing some interesting conditional probabilities that are revised accordingly. [ABSTRACT FROM AUTHOR]

Abstract: This paper aims to explain previously described increases in self-reported psychological distress between 1987 and 2006 among samples identical in respect of age (15 years), school year and geographical location (West of Scotland). Such increases might be explained by changes in exposure (changes in levels of risk or protective factors) and/or by changes in vulnerability (changes in the relationship between risk/protective factors and psychological distress). Key areas of social change over this time period allow identification of potential explanatory factors, categorised as economic, family, educational, values and lifestyle and represented by variables common to each study. Psychological distress was measured via the 12-item General Health Questionnaire, Likert scored. Analyses were conducted on those with complete data on all variables (N = 3276 of 3929), and separately for males and females. Between 1987 and 2006, levels of almost every potential explanatory factor changed in line with general societal trends. Associations between explanatory factors and GHQ tended to be stronger among females, and at the later date. The strongest associations were with worries, arguments with parents, and, at the later date, school disengagement. The factors which best accounted for the increase in mean GHQ between 1987 and 2006 were arguments with parents, school disengagement, worry about school and, for females, worry about family relationships, reflecting both increasing exposure and vulnerability to these risk factors. A number of limitations to our analysis can be identified. However, our results reinforce the conclusions of others in highlighting the role of family and educational factors as plausible explanations for increases in young people’s psychological distress. [Copyright &y& Elsevier]

Abstract: In this paper, we describe scalable highly expressive reasoner (SHER), a breakthrough technology that provides semantic querying of large relational datasets using OWL ontologies. SHER relies on a unique algorithm based on ontology summarization and combines a traditional in-memory description logic reasoner with a database backed RDF Store to scale reasoning to very large Aboxes. In our latest experiments, SHER is able to do sound and complete conjunctive query answering up to 7 million triples in seconds, and scales to datasets with 60 million triples, responding to queries in minutes. We describe the SHER system architecture, discuss the underlying components and their functionality, and briefly highlight two concrete use-cases of scalable OWL reasoning based on SHER in the Health Care and Life Science space. The SHER system, with the source code, is available for download (free for academic use) at: http://www.alphaworks.ibm.com/tech/sher. [Copyright &y& Elsevier]

Abstract: Network security tests should be periodically conducted to detect vulnerabilities before they are exploited. However, analysis of testing results is resource intensive with many data and requires expertise because it is an unsupervised domain. This paper presents how to automate and improve this analysis through the identification and explanation of device groups with similar vulnerabilities. Clustering is used for discovering hidden patterns and abnormal behaviors. Self-organizing maps are preferred due to their soft computing capabilities. Explanations based on anti-unification give comprehensive descriptions of clustering results to analysts. This approach is integrated in Consensus, a computer-aided system to detect network vulnerabilities. [Copyright &y& Elsevier]

Abstract: The goal of predictive toxicology is the automatic construction of carcinogenecity models. Most common artificial intelligence techniques used to construct these models are inductive learning methods. In a previous work we presented an approach that uses lazy learning methods for solving the problem of predicting carcinogenecity. Lazy learning methods solve new problems based on their similarity to already solved problems. Nevertheless, a weakness of these kind of methods is that sometimes the result is not completely understandable by the user. In this paper we propose an explanation scheme for a concrete lazy learning method. This scheme is particularly interesting to justify the predictions about the carcinogenesis of chemical compounds. In addition we propose that these explanations could be used to build a partial domain knowledge. In our particular case, we use the explanations for building general knowledge about carcinogenesis. [Copyright &y& Elsevier]

Abstract: This paper describes an explanation-based tools for constraint programming system. These tools provide to the user the conflicts when it raise during solving process. Our tools simulate constraint additions and/or constraint relaxations without any propagation; it also determine if a given constraint belongs to a conflict and it provide diagnosis tool (e.g. why variable cannot take value val?). With more user-friendly representation of conflicts and explanations, our proposed tools give better problem understanding to the user. We prooved that the proposed tools are efficient, and while there is no debugging system that allow the user to interact with the solver, our explanation-based tools could be used for many other applications. [Copyright &y& Elsevier]

Most of the algorithms developed within the Constraint Satisfaction Problem (CSP) framework cannot be used as such to solve interactive decision support problems, like product configuration. Indeed, in such problems, the user is in charge of assigning values to variables. Global consistency maintaining is only one among several functionalities that should be offered by a CSP-based platform in order to help the user in her task; other important functionalities include providing explanations for some user''s choices and ways to restore consistency.This paper presents an extension of the CSP framework in this direction. The key idea consists in considering and handling the user''s choices as assumptions.  From a theoretical point of view, the complexity issues of various computational tasks involved in interactive decision support problems are investigated. The results cohere with what is known when Boolean constraints are considered and show all the tasks intractable in the worst case. Since interactivity requires short response times, intractability must be circumvented some way. To this end, we present a new method for compiling configuration problems, that can be generalized to valued CSPs. Specifically, an automaton representing the set of solutions of the CSP is first computed off-line, then this data structure is exploited so as to ensure both consistency maintenance and computation of maximal consistent subsets of user''s choices in an efficient way. [Copyright &y& Elsevier]

Research in the social sciences has shown that expectations are an important factor in explanations as used between humans: rather than explaining the cause of an event per se, the explainer will often address another event that did not occur but that the explainee might have expected. For AI-powered systems, this finding suggests that explanation-generating systems may need to identify such end user expectations. In general, this is a challenging task, not the least because users often keep them implicit; there is thus a need to investigate the importance of such an ability. In this paper, we report an empirical study with 181 participants who were shown outputs from a text classifier system along with an explanation of why the system chose a particular class for each text. Explanations were both factual , explaining why the system produced a certain output or counterfactual , explaining why the system produced one output instead of another. Our main hypothesis was explanations should align with end user expectations; that is, a factual explanation should be given when the system's output is in line with end user expectations, and a counterfactual explanation when it is not. We find that factual explanations are indeed appropriate when expectations and output match. When they do not, neither factual nor counterfactual explanations appear appropriate, although we do find indications that our counterfactual explanations contained at least some necessary elements. Overall, this suggests that it is important for systems that create explanations of AI systems to infer what outputs the end user expected so that factual explanations can be generated at the appropriate moments. At the same time, this information is, by itself, not sufficient to also create appropriate explanations when the output and user expectations do not match. This is somewhat surprising given investigations of explanations in the social sciences, and will need more scrutiny in future studies. [ABSTRACT FROM AUTHOR]

Previous research in Explainable Artificial Intelligence (XAI) suggests that a main aim of explainability approaches is to satisfy specific interests, goals, expectations, needs, and demands regarding artificial systems (we call these " stakeholders' desiderata ") in a variety of contexts. However, the literature on XAI is vast, spreads out across multiple largely disconnected disciplines, and it often remains unclear how explainability approaches are supposed to achieve the goal of satisfying stakeholders' desiderata. This paper discusses the main classes of stakeholders calling for explainability of artificial systems and reviews their desiderata. We provide a model that explicitly spells out the main concepts and relations necessary to consider and investigate when evaluating, adjusting, choosing, and developing explainability approaches that aim to satisfy stakeholders' desiderata. This model can serve researchers from the variety of different disciplines involved in XAI as a common ground. It emphasizes where there is interdisciplinary potential in the evaluation and the development of explainability approaches. [ABSTRACT FROM AUTHOR]