Human-Centered Explainable Artificial Intelligence: Concepts, Techniques, and Applications

Abstract

Artificial intelligence (AI) drives the development of various fields in society,and the wide application of complex black box models, such as neural networks and random forests, has opaque decision-making, causing users to distrust. Explainable Artificial Intelligence (XAI) mitigates this by providing certain explanations, yet it balances fidelity and usability, lacks unified evaluation, and fails at multimodal problems. Human-Centered Explainable Artificial Intelligence (HCXAI) emerges as a solution by incorporating human-computer interaction, psychology, sociology, and other concepts into the explanation of the occasion and pays more attention to the user's needs, situational features, and social context. This study systematically sorts out the relevant concepts and core features of HCXAI, including user-orientedness, situationalization, and operability. By combining the three typical domains of healthcare, education, and industry, it explores the application value of HCXAI in scenarios such as clinical decision support systems. In addition, this paper analyzes the challenges faced by HCXAI in terms of trust management and assessment methodology, i.e., challenges in interdisciplinary integration, and proposes future development directions. The study shows that HCXAI enhances user trust, facilitates human-computer collaboration, and promotes AI in critical areas, boosting sustainable development of AI.

Ähnliche Arbeiten