This paper explores the development of human-centered proactive conversational agents (PCAs), emphasizing the importance of aligning technological capabilities with human needs, ethical considerations, and social implications. Unlike traditional reactive systems, proactive systems take initiative to anticipate and plan actions to achieve goals before users explicitly request them. However, without thoughtful design, these systems risk being perceived as intrusive. To address this, the paper introduces a new taxonomy for human-centered PCAs based on three key dimensions: INTELLIGENCE, ADAPTIVITY, and CIVILITY. These dimensions guide the design and evaluation of PCAs across five stages of system construction: Task Formulation, Data Preparation, Model Learning, Evaluation, and System Deployment. The paper discusses the importance of balancing technological advancement with human-centered design principles in the development of PCAs. It highlights the need for PCAs to not only be intelligent and adaptive but also to be civil, respecting user boundaries and ethical standards. The paper also explores the challenges and opportunities in each stage of PCA development, emphasizing the importance of considering user needs, preferences, and ethical considerations in all aspects of PCA design. In Task Formulation, the paper discusses the importance of integrating user emotions, preferences, and values into the design of PCAs. It also highlights the need for PCAs to be adaptable and respectful in their interactions with users. In Data Preparation, the paper discusses the challenges of collecting and annotating data for PCAs, emphasizing the need for real-world data and the importance of ethical considerations in data collection. In Model Learning, the paper discusses the importance of aligning language models with human values and expectations, emphasizing the need for human alignment approaches such as In-context Learning, Supervised Fine-tuning, and Reinforcement Learning. In Evaluation, the paper discusses the need for multidimensional evaluation frameworks that assess the performance of PCAs across the three key dimensions of INTELLIGENCE, ADAPTIVITY, and CIVILITY. In System Deployment, the paper discusses the importance of designing user interfaces that are minimally intrusive and that allow users to have control over system-initiated interactions. It also emphasizes the importance of trust and reliance in human-centered PCAs, highlighting the need for explanations, reliability, and controllability in the design of these systems. The paper concludes by emphasizing the importance of human-centered design in the development of proactive conversational agents, and the need for further research in this area.This paper explores the development of human-centered proactive conversational agents (PCAs), emphasizing the importance of aligning technological capabilities with human needs, ethical considerations, and social implications. Unlike traditional reactive systems, proactive systems take initiative to anticipate and plan actions to achieve goals before users explicitly request them. However, without thoughtful design, these systems risk being perceived as intrusive. To address this, the paper introduces a new taxonomy for human-centered PCAs based on three key dimensions: INTELLIGENCE, ADAPTIVITY, and CIVILITY. These dimensions guide the design and evaluation of PCAs across five stages of system construction: Task Formulation, Data Preparation, Model Learning, Evaluation, and System Deployment. The paper discusses the importance of balancing technological advancement with human-centered design principles in the development of PCAs. It highlights the need for PCAs to not only be intelligent and adaptive but also to be civil, respecting user boundaries and ethical standards. The paper also explores the challenges and opportunities in each stage of PCA development, emphasizing the importance of considering user needs, preferences, and ethical considerations in all aspects of PCA design. In Task Formulation, the paper discusses the importance of integrating user emotions, preferences, and values into the design of PCAs. It also highlights the need for PCAs to be adaptable and respectful in their interactions with users. In Data Preparation, the paper discusses the challenges of collecting and annotating data for PCAs, emphasizing the need for real-world data and the importance of ethical considerations in data collection. In Model Learning, the paper discusses the importance of aligning language models with human values and expectations, emphasizing the need for human alignment approaches such as In-context Learning, Supervised Fine-tuning, and Reinforcement Learning. In Evaluation, the paper discusses the need for multidimensional evaluation frameworks that assess the performance of PCAs across the three key dimensions of INTELLIGENCE, ADAPTIVITY, and CIVILITY. In System Deployment, the paper discusses the importance of designing user interfaces that are minimally intrusive and that allow users to have control over system-initiated interactions. It also emphasizes the importance of trust and reliance in human-centered PCAs, highlighting the need for explanations, reliability, and controllability in the design of these systems. The paper concludes by emphasizing the importance of human-centered design in the development of proactive conversational agents, and the need for further research in this area.