With the rapid advancement of artificial intelligence (AI), machines are increasingly evolving into autonomous agents capable of independent decision−making. Consequently, the traditional paradigm of human–computer interaction is transitioning toward a new model of human–AI collaboration. However, technology−centered AI development approaches have gradually revealed critical limitations, including fragility, bias, and low interpretability, underscoring the urgency of adopting a human−centered AI (HCAI) design philosophy. As a systems−engineering paradigm, the successful implementation of HCAI relies on the design and optimization of high−quality human–AI interaction (HAII). This paper systematically reviews our team's nearly decade−long exploration and practice in the field of HCAI. At the conceptual level, we propose HAII as an emerging interdisciplinary domain and construct a human−centered framework for human–AI collaboration. At the foundational theoretical level, we introduce a joint cognitive system framework for human–AI interaction, an intelligent agent teaming situation awareness model, and a shared social understanding model, thereby forming a relatively comprehensive theoretical system. At the methodological level, we develop a hierarchical HCAI framework and propose five categories of implementation methods. At the applied level, we conduct a series of empirical studies in domains including autonomous driving, intelligent aircraft cockpits, and trust in human–AI collaboration, thereby validating the effectiveness of the proposed framework. Looking forward, research on HCAI and HAII should continue to advance along three dimensions: theoretical refinement, methodological innovation, and application expansion, in order to promote the development of a human−centered intelligent society characterized by harmonious human–AI coexistence.