Toward Human-Centered Human-AI Interaction: Advances in Theoretical Frameworks and Practice

TL;DR

This paper reviews a decade of research on human-centered AI interaction, proposing frameworks and methods to enhance collaboration, explainability, and trust in AI systems across various domains.

Contribution

It introduces a comprehensive interdisciplinary framework for human-AI interaction, including theoretical models, hierarchical methodologies, and empirical validation in real-world applications.

Findings

Validated frameworks in autonomous driving and aircraft cockpits

Demonstrated improved human-AI collaboration and trust

Established a taxonomy of HCAI implementation methods

Abstract

With the rapid development of artificial intelligence (AI), machines are increasingly evolving into intelligent agents, and the human-machine relationship is shifting from traditional "human-computer interaction" toward a new paradigm of "human-AI collaboration." However, technology-centered approaches to AI development have gradually revealed limitations such as fragility, bias, and low explainability, highlighting the urgent need for human-centered AI (HCAI) design philosophy. As a systems engineering approach, the successful implementation of HCAI depends critically on the design and optimization of high-quality human-AI interaction (HAII). This paper systematically reviews our research team's nearly decade-long exploration and practice in HCAI. At the level of research vision, we were among the first in China to systematically propose HAII as an interdisciplinary field and to develop a human-centered conceptual framework for human--AI collaboration. At the theoretical level, we introduced frameworks for human-AI joint cognitive systems, team-level situation awareness among intelligent agents, and shared social understanding, forming a relatively comprehensive theoretical system. At the methodological level, we established a hierarchical HCAI framework and a taxonomy of HCAI implementation methods. At the application level, we conducted a series of studies in domains such as autonomous driving, intelligent aircraft cockpit, and trust in human-AI collaboration, empirically validating the effectiveness of the proposed frameworks. Looking ahead, research on HCAI and HAII must continue to advance along three dimensions: theoretical deepening, methodological innovation, and application expansion, promoting the development of an intelligent society that is human-centered and characterized by harmonious human-AI coexistence.