Year: 2024
Type: article
Abstract: The rise of AI, especially in critical domains, has raised transparency and accountability concerns due to opaque black-box algorithms. This article explores Explainable AI (XAI) and its application, ‘focusing on Remote Sensing and Signal Processing. AI is increasingly used in sectors like autonomous driving, healthcare, and finance, necessitating transparent decision-making.’ Opaque AI models impact trust, bias, and accountability, driving the need for XAI. XAI provides insights into AI decision rationale, supported by GDPR's right to explanations. In Reinforcement Learning (RL), XAI faces unique challenges due to RL's sequential nature and the lack of human-labeled data. XAI methods include model interpretation, post-hoc explanations, interactive explanations, and hybrid approaches. ‘XAI categories include transparent models, opaque models, model-agnostic and model-specific approaches, explanation by simplification, explanation by feature relevance, visual explanation, and local explanation. Applications span healthcare, criminal justice, natural language processing, autonomous systems, agriculture, finance, computer vision, forecasting, remote sensing, social media, and transportation, enhancing trust and fairness. Challenges in natural language generation involve evaluating, handling ambiguous language, constructing narratives, and communicating data quality. This article highlights XAI's role in making AI transparent, addressing black-box algorithm challenges, and fostering trust and accountability in AI decision-making.
Author: Kinjal Gandhi
Institution: Charotar University of Science and Technology
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7217940964
