Explainable Artificial Intelligence For Intelligent Decision-Making In Complex Real-World Applications

The implementation of artificial intelligence systems in highly sensitive fields such as medicine, finance, self-driving cars, and agriculture has revealed an important trade-off between the accuracy of prediction and decision transparency. In this paper, a systematic approach is suggested for embedding XAI in intelligent decision-making systems in practical applications. A combined architecture consisting of the design of intrinsically interpretable models and post hoc explanation procedures, supported by a human-in-the-loop verification stage, is considered. The methodology includes a multi-step pipeline involving data pre-processing, modeling, explanations, and human feedback. Based on the quantitative evaluation of the system performance in various application areas, it is shown that the introduced approach results in significant gains in decision reliability without compromising computational efficiency. The comparison with baseline models demonstrates the effectiveness of the suggested approach in terms of decision uncertainty and expert trust.