Advancing human–AI teams: evolving from instrumental tools to trusted partners

Abstract Human–Computer Interaction (HCI) has undergone fundamental transformations as AI capabilities have advanced, necessitating new theoretical frameworks for understanding human–AI collaboration (HAIC). This review traces HCI’s evolution through four paradigm shifts: the Equipment Era (pre-1970s), Interactive System Era (1980s–1990s), Autonomous Agent Era (1990s–2010s), and the emerging Coexistential AI Era (2020s–present), reflecting changing metaphors from tools to dialog partners, autonomous agents, and co-creative partners. The analysis reveals how anthropomorphism and affective computing have become central to contemporary AI systems, enabling emotional intelligence and pseudo-intimate relationships that fundamentally alter human–technology dynamics. Traditional performance metrics such as efficiency and accuracy are inadequate for evaluating these partnerships, necessitating new measures including intimacy, mutual adaptability, and social bonding. Multi-agent systems and embodied AI are examined as manifestations of social intelligence enabling collaborative architectures across healthcare, transportation, and emergency response. Significant challenges include cognitive load management, transparency deficits, coordination complexity, and ethical concerns regarding emotional manipulation and artificial alienation. Four research priorities are identified: developing novel performance metrics, investigating ethical implications, designing inclusive multi-agent systems, and fostering interdisciplinary collaboration. This analysis provides a roadmap for advancing human–AI partnerships that prioritize human values, social responsibility, and psychological well-being.