Explainable AI (XAI): History, Basic Ideas and Methods
Prachi Zodage,Hussain Harianawala,Hafsa Shaikh,Asad Kharodia
DOI: https://doi.org/10.48175/ijarsct-16988
2024-04-07
Abstract:Explainable Artificial Intelligence (XAI) is a field that aims to make artificial intelligence (AI) processes more transparent, explainable, and understandable. As AI processes become more complex, the need to reveal the "black box" nature of these models and provide explanations for their results and decisions is increasing. XAI aims to bridge the gap between the opaque inner workings of AI and human understanding by creating AI that is accurate, useful, and can explain reasoning and decision-making processes in ways that humans can understand. XAI's importance stems from several factors. First, it addresses trust and accountability issues in AI systems, particularly in high-risk sectors like healthcare, finance, and technology. By providing explanations for AI decisions, stakeholders can better understand the logic behind them, detect inconsistencies, and ensure moral and administrative compliance. Second, XAI encourages collaboration and decision-making between people and intelligence, allowing experts and decision-makers to use their knowledge and experience to make better decisions. Thirdly, XAI plays a crucial role in modeling, debugging, and continuous improvement by identifying flaws, biases, or inconsistencies and working to improve performance standards and reliability. Various methods and techniques are used in XAI, each with their own advantages and limitations. Model-free explanations such as LIME, Anchor and SHAP are particularly important because they can be applied to any AI model, regardless of its design or complexity