Abstract: [Purpose/Significance] Generative artificial intelligence products are becoming a new channel for people to obtain information. Extensive empirical research is required to understand how people utilize and perceive generative artificial intelligence products, and how those products perform in helping people complete tasks in their everyday life contexts. This study focuses on skilled users of ChatGPT 3.5, and analyzes their interaction behavior with generative AI in daily life situations. [Design/Methodology] This study conducted a longitudinal study combined with diary and interview methods, recruiting college students who were ChatGPT proficient users. We recorded the specific process of their interaction with ChatGPT in daily life. The research integrated qualitative and quantitative analysis to categorize the types of tasks users completed with generative AI, and compared the differences in their interaction behaviors and levels of satisfaction when completing different types of tasks. [Findings/Conclusion] It is found that in daily life, the types of tasks completed by college users with ChatGPT can be categorized into five types according to cognitive complexity: information-acquiring tasks, comprehending tasks, operative tasks, analytic and evaluative tasks, and creative and expressive tasks. A relatively high level of users' satisfaction is shown in information-acquiring tasks, comprehending tasks, and operative tasks, while their satisfaction with analytic and evaluative tasks is slightly lower. There is significant differences in users' satisfaction with creative and expressive tasks. [Originality/Value] The research summarizes the task types from the cognitive level, providing an operable standard for the evaluation of AIGC in the future. It also helps explore users' needs for information in the era of intelligent communication, and enrich the research content and achievements in the field of information behaviors under the perspective of AI empowerment.

Keywords: Generative artificial intelligence, ChatGPT, Daily life information behavior, Types of tasks, Human-computer interaction