Preprint
AI Safety & Alignment

Explainable Artificial Intelligence (XAI): What we know and what is left to attain Trustworthy Artificial Intelligence

Sajid Ali(Sungkyunkwan University), Tamer Abuhmed(Sungkyunkwan University), Shaker El–Sappagh(Suez University), Khan Muhammad(Sungkyunkwan University), José M. Alonso(Universidade de Santiago de Compostela), Roberto Confalonieri(University of Padua), Riccardo Guidotti(University of Pisa), Javier Del Ser(University of the Basque Country), Natalia Díaz-Rodríguez(Universidad de Granada), Francisco Herrera(Universidad de Granada)
April 18, 2023Information Fusion1,573 citations
PDFDOI

1.6k

Citations

54

Influential Citations

Information Fusion

Venue

2023

Year

Abstract

Artificial intelligence (AI) is currently being utilized in a wide range of sophisticated applications, but the outcomes of many AI models are challenging to comprehend and trust due to their black-box nature. Usually, it is essential to understand the reasoning behind an AI model’s decision-making. Thus, the need for eXplainable AI (XAI) methods for improving trust in AI models has arisen. XAI has become a popular research subject within the AI field in recent years. Existing survey papers have tackled the concepts of XAI, its general terms, and post-hoc explainability methods but there have not been any reviews that have looked at the assessment methods, available tools, XAI datasets, and other related aspects. Therefore, in this comprehensive study, we provide readers with an overview of the current research and trends in this rapidly emerging area with a case study example. The study starts by explaining the background of XAI, common definitions, and summarizing recently proposed techniques in XAI for supervised machine learning. The review divides XAI techniques into four axes using a hierarchical categorization system: (i) data explainability, (ii) model explainability, (iii) post-hoc explainability, and (iv) assessment of explanations. We also introduce available evaluation metrics as well as open-source packages and datasets with future research directions. Then, the significance of explainability in terms of legal demands, user viewpoints, and application orientation is outlined, termed as XAI concerns. This paper advocates for tailoring explanation content to specific user types. An examination of XAI techniques and evaluation was conducted by looking at 410 critical articles, published between January 2016 and October 2022, in reputed journals and using a wide range of research databases as a source of information. The article is aimed at XAI researchers who are interested in making their AI models more trustworthy, as well as towards researchers from other disciplines who are looking for effective XAI methods to complete tasks with confidence while communicating meaning from data.

Analysis

Why This Paper Matters

This paper addresses a critical gap in the XAI literature by providing a holistic survey that goes beyond typical taxonomies of post-hoc methods. It systematically covers evaluation metrics, open-source tools, and datasets—areas often neglected in prior reviews. As AI systems are deployed in high-stakes domains like healthcare, finance, and law, the need for trustworthy and interpretable models is paramount. By consolidating knowledge on how to assess explanations and tailor them to different user types, this work directly supports practitioners aiming to comply with emerging AI regulations (e.g., GDPR) and build user trust.

Technical Contributions

  • Four-axis hierarchical categorization: (i) data explainability, (ii) model explainability, (iii) post-hoc explainability, and (iv) assessment of explanations. This provides a clear framework for researchers to situate their work.
  • Comprehensive review of evaluation metrics: The paper catalogs metrics for fidelity, stability, comprehensibility, and other explanation quality dimensions, enabling objective comparison of XAI methods.
  • Open-source packages and datasets: Lists available tools (e.g., LIME, SHAP, InterpretML) and benchmark datasets, lowering the barrier for entry into XAI research.
  • User-centric explanation design: Emphasizes that explanations must be tailored to the audience (e.g., domain experts vs. end-users) and discusses legal and ethical implications.

Results

The survey analyzed 410 articles from 2016–2022, revealing a rapidly growing field. It identifies that most XAI research focuses on post-hoc methods for deep learning models, while data explainability and evaluation remain underexplored. The paper does not present new experimental results but synthesizes existing knowledge into a structured taxonomy. Key findings include the lack of standardized evaluation protocols and the need for more user studies to validate explanation effectiveness.

Significance

This paper is a landmark reference for the XAI community, offering a one-stop resource for understanding the state of the art. By highlighting gaps in evaluation and user-centric design, it sets a research agenda for the next generation of XAI work. Its emphasis on legal and ethical concerns aligns with global regulatory trends, making it valuable for both academic researchers and industry practitioners seeking to deploy trustworthy AI systems.

explainable-aitrustworthy-aisurveysupervised-learningpost-hoc-explainabilityevaluation-metricslegal-requirements
Reasoning

Related Papers