Algorithmic Decision Systems: Design, Transparency, And Accountability In AI Models

Abstract

Algorithmic Decision Systems (ADS) have been widely applied in healthcare diagnostics for facilitating effective decision-making. Nevertheless, insufficient transparency, accountability, and interpretability in Artificial Intelligence (AI) systems still pose considerable challenges, especially in life-critical healthcare applications. To promote accurate and AI-driven decision support for healthcare, this research presents human-centric healthcare diagnostic approach by applying the Intelligent Ant Lion Optimizer-Dense Convolutional Network (IntALO-DenseCNet) model, which enables reliable classification while guaranteeing trustworthy AI-based decision-making. Data utilized in proposed model are in form of structured healthcare datasets, including patients' demographic information, diseases, medications, and diagnostic information. Data processing usingmin-max normalization and standardization to preprocess collected data. Additionally, dimensionality reduction is performed using Linear Discriminant Analysis (LDA). The proposed IntALO-DenseCNet decision system for healthcare diagnostics integrates the optimal solutions generated by IntALO for feature selection and parameter tuning. Finally, this research proves that proposed approach ensures high levels of transparency and accountability through interpretable output, confidence score, and decision trace capabilities. Experiments (Python) conducted in the research indicate that IntALO-DenseCNet is capable of achieving state-of-the-art performance at 97.5% accuracy and 97.96% precision. The proposed IntALO-DenseCNet architecture provides clear, accountable, and very precise health diagnosis systems, showing its efficiency as an effective human-centric AI decision-making system.