Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{208760,
        author = {Aishwarya Kolhe and Geeta Yadav and Dnyaneshwari Chalak and Dr. Sunayana Kundan Shivthare},
        title = {A Review of Explainable AI Techniques in Machine Learning},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {no},
        pages = {678-682},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=208760},
        abstract = {Machine learning (ML) is increasingly used in healthcare, finance, education, cybersecurity and other decision support applications. As predictive systems become more complex, it is often difficult for users to understand why a model produces a particular output. This issue is commonly described as the black box problem. Explainable Artificial Intelligence (XAI) addresses this concern by providing information that helps humans understand model predictions and behaviour. This paper reviews major XAI approaches and organizes them according to explanation scope, model dependence and the stage at which interpretability is introduced. Prominent techniques including SHAP, LIME, saliency maps, attention-based approaches, counterfactual explanations and surrogate models are discussed. Their working principles, strengths, limitations and typical application areas are compared. The paper also examines important criteria for judging explanation quality, including fidelity, stability, robustness, completeness and human interpretability.
A key observation from the reviewed literature is that an explanation that appears convincing to a human is not necessarily a faithful representation of the underlying model. The review therefore distinguishes plausibility from faithfulness and highlights the need for systematic evaluation. Finally, research gaps involving standardized benchmarks, causal reasoning, fairness, privacy, robustness and domain-specific explanations are discussed. The review concludes that there is no single XAI method that is optimal for every model or application; the choice should depend on the model, data, decision context and needs of the intended user.},
        keywords = {Explainable AI, Machine Learning, Model Interpretability, SHAP, LIME, Black-box Models.},
        month = {September},
        }