Explainable AI and Reinforcement Learning—A Systematic Review of Current Approaches and Trends

Research into Explainable Artificial Intelligence (XAI) has been increasing in recent years as a response to the need for increased transparency and trust in AI. This is particularly important as AI is used in sensitive domains with societal, ethical, and safety implications. Work in XAI has primarily focused on Machine Learning (ML) for classification, decision, or action, with detailed systematic reviews already undertaken. This review looks to explore current approaches and limitations for XAI in the area of Reinforcement Learning (RL). From 520 search results, 25 studies (including 5 snowball sampled) are reviewed, highlighting visualization, query-based explanations, policy summarization, human-in-the-loop collaboration, and verification as trends in this area. Limitations in the studies are presented, particularly a lack of user studies, and the prevalence of toy-examples and difficulties providing understandable explanations. Areas for future study are identified, including immersive visualization, and symbolic representation.

Download Full-text

Reinforcement Learning's Contribution to the Cyber Security of Distributed Systems

International Journal of Distributed Artificial Intelligence ◽

10.4018/ijdai.2020070103 ◽

2020 ◽

Vol 12 (2) ◽

pp. 35-55

Author(s):

Christophe Feltus

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Systematic Review ◽

Information System ◽

Reinforcement Learning ◽

Distributed Systems ◽

Unsupervised Learning ◽

Cyber Security ◽

Analytic Framework ◽

Learning Paradigm

Reinforcement learning (RL) is a machine learning paradigm, like supervised or unsupervised learning, which learns the best actions an agent needs to perform to maximize its rewards in a particular environment. Research into RL has been proven to have made a real contribution to the protection of cyberphysical distributed systems. In this paper, the authors propose an analytic framework constituted of five security fields and eight industrial areas. This framework allows structuring a systematic review of the research in artificial intelligence that contributes to cybersecurity. In this contribution, the framework is used to analyse the trends and future fields of interest for the RL-based research in information system security.

Download Full-text

Towards Trustable Explainable AI

Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence ◽

10.24963/ijcai.2020/726 ◽

2020 ◽

Author(s):

Alexey Ignatiev

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Recent Work ◽

Abductive Reasoning ◽

Rigorous Approach ◽

Explainable Artificial Intelligence ◽

Explainable Ai ◽

Close Relationship

Explainable artificial intelligence (XAI) represents arguably one of the most crucial challenges being faced by the area of AI these days. Although the majority of approaches to XAI are of heuristic nature, recent work proposed the use of abductive reasoning to computing provably correct explanations for machine learning (ML) predictions. The proposed rigorous approach was shown to be useful not only for computing trustable explanations but also for validating explanations computed heuristically. It was also applied to uncover a close relationship between XAI and verification of ML models. This paper overviews the advances of the rigorous logic-based approach to XAI and argues that it is indispensable if trustable XAI is of concern.

Download Full-text

Explainable artificial intelligence for education and training

The Journal of Defense Modeling and Simulation Applications Methodology Technology ◽

10.1177/15485129211028651 ◽

2021 ◽

pp. 154851292110286

Author(s):

Krzysztof Fiok ◽

Farzad V Farahani ◽

Waldemar Karwowski ◽

Tareq Ahram

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Black Box ◽

Target User ◽

Explainable Artificial Intelligence ◽

Explainable Ai ◽

History Of ◽

User Groups ◽

And Training ◽

Intelligent Action

Researchers and software users benefit from the rapid growth of artificial intelligence (AI) to an unprecedented extent in various domains where automated intelligent action is required. However, as they continue to engage with AI, they also begin to understand the limitations and risks associated with ceding control and decision-making to not always transparent artificial computer agents. Understanding of “what is happening in the black box” becomes feasible with explainable AI (XAI) methods designed to mitigate these risks and introduce trust into human-AI interactions. Our study reviews the essential capabilities, limitations, and desiderata of XAI tools developed over recent years and reviews the history of XAI and AI in education (AIED). We present different approaches to AI and XAI from the viewpoint of researchers focused on AIED in comparison with researchers focused on AI and machine learning (ML). We conclude that both groups of interest desire increased efforts to obtain improved XAI tools; however, these groups formulate different target user groups and expectations regarding XAI features and provide different examples of possible achievements. We summarize these viewpoints and provide guidelines for scientists looking to incorporate XAI into their own work.

Download Full-text

Explainable AI

Cognitive Skills You Need for the 21st Century ◽

10.1093/oso/9780197529003.003.0015 ◽

2020 ◽

pp. 170-179

Author(s):

Stephen K. Reed

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Biological Knowledge ◽

Types Of Knowledge ◽

Explainable Artificial Intelligence ◽

Explainable Ai ◽

Degree Of Confidence ◽

Connectionist Learning

Deep connectionist learning has resulted in very impressive accomplishments, but it is unclear how it achieves its results. A dilemma in using the output of machine learning is that the best performing methods are the least explainable. Explainable artificial intelligence seeks to develop systems that can explain their reasoning to a human user. The application of IBM’s WatsonPaths to medicine includes a diagnostic network that infers a diagnosis from symptoms with a degree of confidence associated with each diagnosis. The Semanticscience Integrated Ontology uses categories such as objects, processes, attributes, and relations to create networks of biological knowledge. The same categories are fundamental in representing other types of knowledge such as cognition. Extending an ontology requires a consistent use of semantic terms across different domains of knowledge.

Download Full-text

Prediction Models for Public Health Containment Measures on COVID-19 Using Artificial Intelligence and Machine Learning: A Systematic Review

International Journal of Environmental Research and Public Health ◽

10.3390/ijerph18094499 ◽

2021 ◽

Vol 18 (9) ◽

pp. 4499

Author(s):

Anil Babu Payedimarri ◽

Diego Concina ◽

Luigi Portinale ◽

Massimo Canonico ◽

Deborah Seys ◽

...

Keyword(s):

Public Health ◽

Artificial Intelligence ◽

Machine Learning ◽

Systematic Review ◽

Public Transportation ◽

Prediction Models ◽

Positive Impact ◽

Health Interventions ◽

Public Health Interventions ◽

Commercial Activities

Artificial Intelligence (AI) and Machine Learning (ML) have expanded their utilization in different fields of medicine. During the SARS-CoV-2 outbreak, AI and ML were also applied for the evaluation and/or implementation of public health interventions aimed to flatten the epidemiological curve. This systematic review aims to evaluate the effectiveness of the use of AI and ML when applied to public health interventions to contain the spread of SARS-CoV-2. Our findings showed that quarantine should be the best strategy for containing COVID-19. Nationwide lockdown also showed positive impact, whereas social distancing should be considered to be effective only in combination with other interventions including the closure of schools and commercial activities and the limitation of public transportation. Our findings also showed that all the interventions should be initiated early in the pandemic and continued for a sustained period. Despite the study limitation, we concluded that AI and ML could be of help for policy makers to define the strategies for containing the COVID-19 pandemic.

Download Full-text

Artificial Cognition: How Experimental Psychology Can Help Generate Explainable Artificial Intelligence

10.31234/osf.io/ygr4c ◽

2021 ◽

Author(s):

J. Eric T. Taylor ◽

Graham Taylor

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Deep Neural Networks ◽

Experimental Approach ◽

Well Being ◽

Human Mind ◽

Black Box ◽

Psychological Science ◽

Black Boxes ◽

Explainable Artificial Intelligence

Artificial intelligence powered by deep neural networks has reached a levelof complexity where it can be difficult or impossible to express how a modelmakes its decisions. This black-box problem is especially concerning when themodel makes decisions with consequences for human well-being. In response,an emerging field called explainable artificial intelligence (XAI) aims to increasethe interpretability, fairness, and transparency of machine learning. In thispaper, we describe how cognitive psychologists can make contributions to XAI.The human mind is also a black box, and cognitive psychologists have overone hundred and fifty years of experience modeling it through experimentation.We ought to translate the methods and rigour of cognitive psychology to thestudy of artificial black boxes in the service of explainability. We provide areview of XAI for psychologists, arguing that current methods possess a blindspot that can be complemented by the experimental cognitive tradition. Wealso provide a framework for research in XAI, highlight exemplary cases ofexperimentation within XAI inspired by psychological science, and provide atutorial on experimenting with machines. We end by noting the advantages ofan experimental approach and invite other psychologists to conduct research inthis exciting new field.

Download Full-text

Machine learning applications to neuroimaging for glioma detection and classification: An artificial intelligence augmented systematic review

Journal of Clinical Neuroscience ◽

10.1016/j.jocn.2021.04.043 ◽

2021 ◽

Vol 89 ◽

pp. 177-198

Author(s):

Quinlan D. Buchlak ◽

Nazanin Esmaili ◽

Jean-Christophe Leveque ◽

Christine Bennett ◽

Farrokh Farrokhi ◽

...

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Systematic Review ◽

Machine Learning Applications

Download Full-text

Recommender Systems in Digital Libraries Using Artificial Intelligence and Machine Learning

Advances in Systems Analysis, Software Engineering, and High Performance Computing - Handbook of Research on Methodologies and Applications of Supercomputing ◽

10.4018/978-1-7998-7156-9.ch012 ◽

2021 ◽

pp. 162-178

Author(s):

Namik Delilovic

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Digital Libraries ◽

Full Text ◽

Text Search ◽

Full Text Search ◽

Artificial Intelligence Techniques ◽

Search Results ◽

Advanced Search ◽

Search Field

Searching for contents in present digital libraries is still very primitive; most websites provide a search field where users can enter information such as book title, author name, or terms they expect to be found in the book. Some platforms provide advanced search options, which allow the users to narrow the search results by specific parameters such as year, author name, publisher, and similar. Currently, when users find a book which might be of interest to them, this search process ends; only a full-text search or references at the end of the book may provide some additional pointers. In this chapter, the author is going to give an example of how a user could permanently get recommendations for additional contents even while reading the article, using present machine learning and artificial intelligence techniques.

Download Full-text