Building Trustworthy AI Systems: Explainable Decision Support, Privacy-Preserving Learning, and Agentic Federated Workflows
Abstract
The rapid development of generative artificial intelligence, particularly large language models (LLMs), has created new opportunities for deploying intelligent systems in high-stakes domains such as healthcare and education. At the same time, practical adoption remains constrained by several persistent challenges, including the need for trustworthy and explainable decision support, the protection of sensitive data, and the difficulty of training robust models under heterogeneous and decentralized data conditions. This dissertation addresses these challenges through three complementary studies that develop AI frameworks integrating explainable machine learning, knowledge-augmented LLMs, agentic AI, federated learning, and privacy-preserving data processing for medical and educational applications. The first study develops an explainable AI framework for clinical decision support. Although existing explainable machine learning methods can reveal useful information about model behavior, their outputs are often technical and difficult for healthcare practitioners to interpret directly. To bridge this gap, the proposed framework combines model-agnostic explanations with knowledge-augmented LLMs. Binary classifiers are trained for clinical depression diagnosis using biological indicators, and explanation-related information is organized into a structured knowledge graph that preserves explanatory evidence and captures relationships among models, features, predictions, and…
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.