As artificial intelligence (AI) systems advance rapidly across virtual and physical domains, adopting a human-centered AI (HCAI) perspective in system design is essential for enhancing both system capabilities and human–AI teaming effectiveness. As a subdomain of human–AI interaction, human–AI teaming focuses on the task interdependency between the AI systems and humans to achieve a joint goal. Learning from human social intelligence, artificial social intelligence (ASI) enables AI agents to infer human states, predict their actions, and intervene in responce to complex human cues in dynamic environments, making it a foundational competency for effective teaming. This chapter explores the conceptual foundations, design principles, and empirical evidence for ASI, with an emphasis on high-stakes domains such as search and rescue and command-and-control operations. The case studies of the DARPA’s Artificial Social Intelligence (ASIST) and Adaptive Distributed Allocation of Probabilistic Tasks (ADAPTII) programs showcase experimental testbeds and metrics used to evaluate socially competent AI, and discuss challenges such as cultural variability, ethical concerns, and evaluation standards. The chapter concludes with a vision for generalizable ASI that supports long-term, co-adaptive human–AI teaming, advocating for standardized frameworks, simulation environments, and sustained interdisciplinary research.