关键词: GenAI素养, 高阶思维能力, 人智协作绩效, 认知加工策略

Abstract: With the development of Generative Artificial Intelligence (GenAI), optimizing Human-GenAI collaborative performance has become the main topic for both researchers and practitioners. Although higher levels of GenAI literacy are considered beneficial for improving human-GenAI collaborative performance, emerging evidence presents mixed findings. It is necessary to examine its measurement, boundary conditions and influencing processes. Drawing from human-machine complementarity theory, this study proposes that effective collaboration stems from the interaction between human GenAI literacy and higher-order thinking skills. In Study 1, 216 university students with GenAI experience were recruited. The researchers developed a 25-item multiple-choice question test grounded in an established four-dimensional framework of GenAI literacy, encompassing Knowledge & Understanding, Use & Application, Evaluation & Creation, and Ethics. This instrument was rigorously validated using both Classical Test Theory and Item Response Theory. Subsequently, participants completed a “Career Planning Program” writing task by using ChatGPT-5. Their outputs were evaluated by three independent raters using the Creative Product Semantic Scale, and the regression results demonstrated that GenAI literacy had a significant positive influence on Human-GenAI collaborative performance. In Study 2, 261 employees from an e-commerce company were recruited. Participants completed the same GenAI literacy test from Study 1 and assessed the three dimensions of their higher-order thinking skills: Creativity, Domain Expertise, and Problem-Solving Capability. Then participants used ChatGPT-5 to write a marketing copy. Their performance was also rated by using the Creative Product Semantic Scale. The results showed the positive effect of GenAI literacy on performance. More importantly, regression analyses revealed significant positive interaction effects between GenAI literacy and all three dimensions of higher-order thinking skills on performance. Additionally, this study adopted Epistemic Network Analysis to code user-GenAI dialogues, providing evidence for the differentiated cognitive processing strategies of users with various higher-order thinking skills. The findings provide several implications. First, this study offers a validated tool for assessing GenAI literacy. Second, higher-order thinking skills—encompassing creativity, domain expertise, and problem-solving capability—serve as a key moderator for the effectiveness of GenAI literacy. Third, this study reveals differentiated cognitive processing strategies among individuals with varying higher-order thinking skills during human-GenAI collaboration. Organizations should cultivate employees’ GenAI literacy and these human skills simultaneously. In Study 1, 216 university students with GenAI experience were recruited. The researchers developed a 25-item multiple-choice question test grounded in an established four-dimensional framework of GenAI literacy, encompassing Knowledge & Understanding, Use & Application, Evaluation & Creation, and Ethics. This instrument was rigorously validated using both Classical Test Theory and Item Response Theory. Subsequently, participants completed a “Career Planning Program” writing task by using ChatGPT-5. Their outputs were evaluated by three independent raters using the Creative Product Semantic Scale, and the regression results demonstrated that GenAI literacy had a significant positive influence on Human-GenAI collaborative performance. In Study 2, 261 employees from an e-commerce company were recruited. Participants completed the same GenAI literacy test from Study 1 and assessed the three dimensions of their higher-order thinking skills: Creativity, Domain Expertise, and Problem-Solving Capability. Then participants used ChatGPT-5 to write marketing copy. Their performance was also rated by using the Creative Product Semantic Scale. The results showed the positive effect of GenAI literacy on performance. More importantly, regression analyses revealed significant positive interaction effects between GenAI literacy and all three dimensions of higher-order thinking skills on performance. Additionally, this study employs Epistemic Network Analysis to code user-GenAI dialogues, providing evidence for the mediating role of differentiated cognitive processing strategies of users with various higher-order thinking skills. The findings provide several implications. First, this study offers a validated tool for assessing GenAI literacy. Second, higher-order thinking skills—encompassing creativity, domain expertise, and problem-solving capability—serve as a key moderator for the effectiveness of GenAI literacy. Third, this study advocates for a “process-oriented” research paradigm and reveals differentiated cognitive processing strategies among individuals with varying higher-order thinking skills during human-GenAI collaboration. Organizations should cultivate employees’ technical GenAI literacy and these human skills simultaneously. In Study 1, 216 university students with GenAI experience were recruited. And the researchers developed a 25-item multiple-choice question test grounded in an established four-dimensional framework of AI literacy, encompassing Knowledge & Understanding, Use & Application, Evaluation & Creation, and Ethics. This instrument was rigorously validated using both Classical Test Theory and Item Response Theory. Subsequently, participants completed a “Career Planning Program” writing task by using ChatGPT-5. Their outputs were evaluated by three independent raters using the Creative Product Semantic Scale, and the regression results demonstrated that AI literacy had a significant positive influence on Human-AI collaborative performance. In Study 2, 261 employees from an e-commerce company were recruited. Participants completed the same AI literacy test from Study 1 and assessed the three dimensions of their higher-order thinking skills: Creativity, Domain Expertise, and Problem-Solving Capability. Then participants used ChatGPT-5 to write marketing copy. Their performance was also rated by using the Creative Product Semantic Scale. The results showed the positive effect of AI literacy on performance. More importantly, regression analyses revealed significant positive interaction effects between AI literacy and all three dimensions of higher-order thinking skills on performance. Additionally, this study employs Epistemic Network Analysis to code user-AI dialogues, providing evidence for the mediating role of differentiated cognitive processing strategies of users with various higher-order thinking skills. The findings provide several implications. Firstly, AI literacy is an antecedent for high-performance Human-AI collaboration. Secondly, higher-order thinking skills—encompassing creativity, domain expertise, and problem-solving capability—serve as a key moderator for the effectiveness of AI literacy. Third, this study advocates for a “process-oriented” research paradigm and reveals differentiated cognitive processing strategies among individuals with varying higher-order thinking skills during human-AI collaboration. Fourth, this study offers a validated tool for assessing AI literacy and underscores the imperative of fostering a "human-centered" approach to collaboration in the GenAI era. Organizations should cultivate employees’ technical AI skills and these human capabilities simultaneously.

Key words: GenAI literacy, higher-order thinking skills, human-GenAI collaborative performance, cognitive processing strategies