基于生成式人工智能的认知外包: 交互行为模式与认知结构特征分析

doi:10.3724/SP.J.1041.2025.0967

摘要/Abstract

摘要：

人类通过外包部分认知任务给外部生成式人工智能技术来提升任务完成的效率和质量, 但认知外包的效果因人而异。为了揭示有效认知外包的关键特征和内在要求, 研究设计了一个面向研究生的认知外包活动, 参与者在生成式人工智能系统的协助下撰写开放性主题文章, 并依据文章得分被划分为高绩效组和低绩效组。通过对知识前测的差异性分析发现高绩效组的先前领域知识水平显著高于低绩效组。通过对交互过程数据进行滞后序列分析和认知网络分析, 发现两组群体在交互行为模式和认知结构特征上存在差异: 高绩效组的行为转换更加多元, 形成“快速自主的任务理解与规划——高效精准的人机互动——选择性提取与深度加工”的行为模式; 高绩效组的认知结构较为均衡和完整, 表现为交互中各认知元素间相对多样且紧密的关联, 而低绩效组的认知结构相对失衡和松散, 表现为对低层次认知元素的偏向和各元素间相对单一且微弱的关联。综合来看, 有效认知外包是个体在认知活动中积极参与、深入加工的复杂过程, 需要内外部认知网络的平衡与有效连接的建立。

关键词: 认知外包, 人机协同, 认知网络分析, 滞后序列分析, 内外部认知的连接

Abstract:

The emergence of generative AI has profoundly impacted the field of education by enabling individuals to enhance both the efficiency and quality of cognitive tasks by delegating part of the tasks to generative AI. This process is referred to as cognitive outsourcing. However, individuals’ effectiveness in using AI varies. Empirical research on the educational applications of generative AI remains limited, primarily focusing on evaluating technical capabilities and the effects of learning support. At present, the cognitive and behavioral prerequisites for effective cognitive outsourcing remain unclear. Furthermore, the differences in prior knowledge, behavioral patterns, and cognitive structures among individuals with varying performances have yet to be thoroughly explored. In this study, we designed a cognitive outsourcing activity for graduate students involving a sample of 46 participants (10 males, 36 females; age: M = 26.39, SD = 6.91). The activity consisted of two sessions. In the first session, participants were allotted 30 minutes to independently construct a concept map on the topic "Artificial Intelligence and Teachers" using pen and paper, which served as a measure of their prior knowledge. In the second session, participants engaged with a generative AI system to compose an essay on the same topic within a 100-minute time frame using a computer. The entire process was video-recorded. Based on expert evaluations, participants were categorized into high-performance and low-performance groups according to their essay scores. Interactive behaviors and contents were coded, and behavioral sequence transitions between the two groups were mapped using Lag Sequence Analysis. Additionally, Epistemic Network Analysis was employed to construct cognitive structure mappings, followed by a comparative analysis of the differences between the two groups.

The results indicate that the high-performance group exhibited significantly higher prior domain knowledge compared to the low-performance group. Significant differences were observed between the two groups, including the frequency of different interactive behaviors, the frequency of different cognitive elements, the behavioral sequences, and the cognitive network structures. From the behavioral perspective, the high-performance group demonstrated significantly more diversified behavioral transitions, forming a distinctive pattern characterized by "rapid and autonomous task comprehension and planning, efficient and precise human-computer interaction, selective information extraction and deep processing." From the cognitive perspective, the high-performance group exhibited a well-balanced and comprehensive cognitive structure characterized by diverse and tightly interconnected cognitive elements. In contrast, the low-performance group displayed an unbalanced and loosely connected cognitive structure, primarily engaging with lower cognitive-level interaction. Overall, the findings indicate that effective cognitive outsourcing is a multifaceted process that necessitates active participation and profound cognitive processing. It demands proficient integration between internal cognitive frameworks and external technological tools. These findings highlight the distinct behavioral patterns and cognitive structures of individuals with varying levels of success in cognitive outsourcing activities and elucidate the cognitive and behavioral requirements for effective cognitive outsourcing. By focusing on individuals’ prior knowledge and interactive processes, this study examines the influence of cognitive and behavioral characteristics on the efficacy of generative AI-assisted writing, thus contributing to empirical research on generative AI-supported education. Additionally, it extends the theoretical understanding of cognitive outsourcing and provides insight for future research and educational practices. Furthermore, the interactive behavior and content coding framework established in this study, along with the application of Lag Sequence Analysis and Epistemic Network Analysis, provide valuable methodological references. Future research should further investigate the long-term and deep-seated effects of cognitive outsourcing on individuals with different characteristics, as well as the intrinsic neural mechanisms underlying effective cognitive outsourcing.

Key words: cognitive outsourcing, human-computer collaboration, epistemic network analysis, lag sequence analysis, internal and external cognitive connections

中图分类号:

B842

汪凡淙, 汤筱玙, 余胜泉. (2025). 基于生成式人工智能的认知外包: 交互行为模式与认知结构特征分析. 心理学报, 57(6), 967-986.

WANG Fancong, TANG Xiaoyu, YU Shengquan. (2025). Cognitive outsourcing based on generative artificial intelligence: An Analysis of interactive behavioral patterns and cognitive structural features. Acta Psychologica Sinica, 57(6), 967-986.

图/表 13

参考文献 53

[54]	Sweller J., Ayres P., & Kalyuga S. (2011). Cognitive load theory. New York: Springer.
[55]	Terry C. A., Mishra P., & Roseth C. J. (2016). Preference for multitasking, technological dependency, student metacognition, & pervasive technology use: An experimental intervention. Computers in Human Behavior, 65, 241-251.
[56]	Tsai C. Y., Lin Y. T., & Brown I. K. (2024). Impacts of ChatGPT-assisted writing for EFL English majors: Feasibility and challenges. Education and Information Technologies, 29, 22427-22445.
[57]	Voinea C., Vică C., Mihailov E., & Savulescu J. (2020). The internet as cognitive enhancement. Science and Engineering Ethics, 26(4), 2345-2362. doi: 10.1007/s11948-020-00210-8 pmid: 32253711
[58]	Wang X., Liu Q., Pang H., Tan S. C., Lei J., Wallace M. P., & Li L. (2023). What matters in AI-supported learning: A study of human-AI interactions in language learning using cluster analysis and epistemic network analysis. Computers & Education, 194, 104703. https://doi.org/10.1016/j. compedu.2022.104703
[59]	White J., Fu Q., Hays S., Sandborn M., Olea C., Gilbert H., … Schmidt D. C. (2023). A prompt pattern catalog to enhance prompt engineering with ChatGPT. arXiv:2302. 11382 [Preprint]. https://doi.org/10.48550/arXiv.2302.11382
[60]	Witherby A. E., & Carpenter S. K. (2022). The rich-get- richer effect: Prior knowledge predicts new learning of domain-relevant information. Journal of Experimental Psychology: Learning, Memory, and Cognition, 48(4), 483-498.
[61]	Xia Q., Chiu T. K., Chai C. S., & Xie K. (2023). The mediating effects of needs satisfaction on the relationships between prior knowledge and self-regulated learning through artificial intelligence chatbot. British Journal of Educational Technology, 54(4), 967-986.
[62]	Yilmaz R., & Yilmaz F. G. K. (2023). The effect of generative artificial intelligence (AI)-based tool use on students’ computational thinking skills, programming self-efficacy and motivation. Computers and Education: Artificial Intelligence, 4, 100147. https://doi.org/10.1016/j.caeai.2023.100147
[63]	Yu S., & Wang F. (2023). Cognitive outsourcing pitfalls of artificial intelligence for educational applications and their crossing. e-Education Research, (12), 5-13.
	[余胜泉, 汪凡淙. (2023). 人工智能教育应用的认知外包陷阱及其跨越. 电化教育研究, (12), 5-13.]
[64]	Yu S., & Wang Q. (2019). Analysis of collaborative path development of "AI+teachers". e-Education Research, (4), 14-22+29.
[1]	Abdelhalim S. M. (2024). Using ChatGPT to promote research competency: English as a foreign language undergraduates’ perceptions and practices across varied metacognitive awareness levels. Journal of Computer Assisted Learning, 40(3), 1261-1275.
[2]	Alexander P. A., & Judy J. E. (1988). The interaction of domain-specific and strategic knowledge in academic performance. Review of Educational Research, 58(4), 375-404.
[3]	Alkaissi H., & McFarlane S. I. (2023). Artificial hallucinations in ChatGPT: Implications in scientific writing. Cureus, 15(2), Article e35179. https://doi.org/10.7759/cureus.35179
[4]	Anderson L. W., & Krathwohl D. R. (2001). A taxonomy for learning, teaching, and assessing: A revision of Bloom's taxonomy of educational objectives: Complete edition. Addison Wesley Longman, Inc.
[5]	Anderson R. C. (1984). Some reflections on the acquisition of knowledge. Educational Researcher, 13(9), 5-10.
[6]	Baidoo-Anu D., & Ansah L. O. (2023). Education in the era of generative artificial intelligence (AI): Understanding the potential benefits of ChatGPT in promoting teaching and learning. Journal of AI, 7(1), 52-62.
[7]	Bakeman R., & Gottman J. M. (1997). Observing interaction: An introduction to sequential analysis. Cambridge University Press.
[8]	Barr N., Pennycook G., Stolz J. A., & Fugelsang J. A. (2015). The brain in your pocket: Evidence that smartphones are used to supplant thinking. Computers in Human Behavior, 48, 473-480.
[9]	Bašić Ž., Banovac A., Kružić I., & Jerković I. (2023). ChatGPT-3.5 as writing assistance in students’ essays. Humanities and Social Sciences Communications, 10, 750. https://doi.org/10.1057/s41599-023-02269-7
[10]	Beier M. E., & Ackerman P. L. (2005). Age, ability, and the role of prior knowledge on the acquisition of new domain knowledge: Promising results in a real-world learning environment. Psychology and Aging, 20(2), 341-355. doi: 10.1037/0882-7974.20.2.341 pmid: 16029097
[11]	Bittermann A., McNamara D., Simonsmeier B. A., & Schneider M. (2023). The landscape of research on prior knowledge and learning: A bibliometric analysis. Educational Psychology Review, 35, 58. https://doi.org/10.1007/s10648-023-09775-9
[12]	Bostrom N., & Sandberg A. (2009). Cognitive enhancement: Methods, ethics, regulatory challenges. Science and Engineering Ethics, 15(3), 311-341. doi: 10.1007/s11948-009-9142-5 pmid: 19543814
[64]	[余胜泉, 王琦. (2019). “AI+教师”的协作路径发展分析. 电化教育研究, (4), 14-22+29.]
[65]	Zamfirescu-Pereira J. D., Wong R. Y., Hartmann B., & Yang Q. (2023, April). Why Johnny can’t prompt: How non-AI experts try (and fail) to design LLM prompts. Paper presented at the meeting of the 2023 CHI Conference on Human Factors in Computing Systems, Hamburg, Germany.
[66]	Zhang J., Bao K., Zhang Y., Wang W., Feng F., & He X. (2023, September). Is ChatGPT fair for recommendation? evaluating fairness in large language model recommendation. Paper presented at the meeting of the 17th ACM Conference on Recommender Systems, Singapore.
[67]	Zimmerman B. J., & Pons M. M. (1986). Development of a structured interview for assessing student use of self-regulated learning strategies. American Educational Research Journal, 23(4), 614-628.
[13]	Bouton E., Yosef D., & Asterhan C. S. (2025). Differences between low and high achievers in whole-classroom dialogue participation quality. Learning and Instruction, 96, 102088. https://doi.org/10.1016/j.learninstruc.2025.102088
[14]	Chang C. Y., Lin H. C., Yin C., & Yang K. H. (2025). Generative AI-assisted reflective writing for improving students’ higher order thinking: Evidence from quantitative and epistemic network analysis. Educational Technology & Society, 28(1), 270-285.
[15]	Chang K. E., Sung Y. T., Chang R. B., & Lin S. C. (2005). A new assessment for computer-based concept mapping. Journal of Educational Technology & Society, 8(3), 138-148.
[16]	Cheng Y. W., Wang Y., Cheng I. L., & Chen N. S. (2019). An in-depth analysis of the interaction transitions in a collaborative Augmented Reality-based mathematic game. Interactive Learning Environments, 27(5-6), 782-796.
[17]	Chu H. C., Hwang G. J., & Tsai C. C. (2010). A knowledge engineering approach to developing mindtools for context-aware ubiquitous learning. Computers & Education, 54(1), 289-297.
[18]	Clark A. (2001). Natural-born cyborgs? In Beynon, M., Nehaniv, C. L., & Dautenhahn, K. (Eds.), Lecture notes in computer science: Vol. 2117: Cognitive technology: Instruments of mind (pp. 17-24). Springer, Berlin, Heidelberg.
[19]	Clark A., & Chalmers D. (1998). The extended mind. Analysis, 58(1), 7-19.
[20]	Cohen J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20(1), 37-46.
[21]	Cole M., & Engeström Y. (1993). A cultural-historical approach to distributed cognition. In G. Salomon (Ed.), Distributed cognitions: Psychological and educational considerations (pp. 1-46). Cambridge University Press.
[22]	Fan Y., Tang L., Le H., Shen K., Tan S., Zhao Y.,... Gašević D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56(2), 489-530.
[23]	Farina M., & Levin S. (2021). The Extended Mind Thesis and Its Applications. In: Robinson, M. D., & Thomas, L. E. (Eds.), Handbook of embodied psychology (pp. 127-147). Springer, Cham.
[24]	Farrokhnia M., Banihashem S. K., Noroozi O., & Wals A. (2024). A SWOT analysis of ChatGPT: Implications for educational practice and research. Innovations in Education and Teaching International, 61(3), 460-474.
[25]	Fisher M., Goddu M. K., & Keil F. C. (2015). Searching for explanations: How the Internet inflates estimates of internal knowledge. Journal of Experimental Psychology: General, 144(3), 674-687.
[26]	Gilson A., Safranek C. W., Huang T., Socrates V., Chi L., Taylor R. A., & Chartash D. (2023). How does ChatGPT perform on the United States medical licensing examination? The implications of large language models for medical education and knowledge assessment. JMIR Medical Education, 9(1), Article e45312. https://doi.org/10.2196/45312
[27]	Greve A., Cooper E., Tibon R., & Henson R. N. (2019). Knowledge is power: Prior knowledge aids memory for both congruent and incongruent events, but in different ways. Journal of Experimental Psychology: General, 148(2), 325-341.
[28]	Gunawardena C. N., Lowe C. A., & Anderson T. (1997). Analysis of a global online debate and the development of an interaction analysis model for examining social construction of knowledge in computer conferencing. Journal of Educational Computing Research, 17(4), 397-431.
[29]	Güner H., & Er E. (2025). AI in the classroom: Exploring students’ interaction with ChatGPT in programming learning. Education and Information Technologies. Advance online publication. https://doi.org/10.1007/s10639-025-13337-7
[30]	Herbold S., Hautli-Janisz A., Heuer U., Kikteva Z., & Trautsch A. (2023). A large-scale comparison of human-written versus ChatGPT-generated essays. Scientific Reports, 13, 18617. https://doi.org/10.1038/s41598-023-45644-9
[31]	Hollan J., Hutchins E., & Kirsh D. (2000). Distributed cognition: Toward a new foundation for human-computer interaction research. ACM Transactions on Computer- Human Interaction, 7(2), 174-196.
[32]	Hwang G. J., Hung P. H., Chen N. S., & Liu G. Z. (2014). Mindtool-assisted in-field learning (MAIL): An advanced ubiquitous learning project in Taiwan. Journal of Educational Technology & Society, 17(2), 4-16.
[33]	Kang J., & Liu M. (2022). Investigating navigational behavior patterns of students across at-risk categories within an open-ended serious game. Technology, Knowledge and Learning, 27, 183-205.
[34]	Kasneci E., Sessler K., Küchemann S., Bannert M., Dementieva D., Fischer F., … Kasneci G. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences, 103, 102274. https://doi.org/10.1016/j.lindif.2023.102274
[35]	Kocoń J., Cichecki I., Kaszyca O., Kochanek M., Szydło D., Baran J.,... Kazienko P. (2023). ChatGPT: Jack of all trades, master of none. Information Fusion, 99, 101861. https://doi.org/10.1016/j.inffus.2023.101861
[36]	Lin X. F., Wang Z., Zhou W., Luo G., Hwang G. J., Zhou Y.,... Liang Z. M. (2023). Technological support to foster students’ artificial intelligence ethics: An augmented reality-based contextualized dilemma discussion approach. Computers & Education, 201, 104813. https://doi.org/10.1016/j.compedu.2023.104813
[37]	Lin Z. (2024). How to write effective prompts for large language models. Nature Human Behaviour, 8(4), 611-615. doi: 10.1038/s41562-024-01847-2 pmid: 38438650
[38]	Liu M., Zhang L. J., & Biebricher C. (2024). Investigating students’ cognitive processes in generative AI-assisted digital multimodal composing and traditional writing. Computers & Education, 211, 104977. https://doi.org/10.1016/j.compedu.2023.104977
[39]	Liu P., Yuan W., Fu J., Jiang Z., Hayashi H., & Neubig G. (2023). Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing. ACM Computing Surveys, 55(9), 1-35.
[40]	Mao Z., Li X., Li Y., Shao J., & Bao H. (2025). Capturing patterns of argumentation elements, encountered challenges, and social regulation in collaborative argumentation: An epistemic network analysis study. Interactive Learning Environments. Advance online publication. https://doi.org/10.1080/10494820.2025.2460583
[41]	Molenaar I., Horvers A., & Baker R. S. J. D. (2021). What can moment-by-moment learning curves tell about students' self-regulated learning? Learning and Instruction, 72, 101206. https://doi.org/10.1016/j.learninstruc.2019.05.003
[42]	Muraven M., & Baumeister R. F. (2000). Self-regulation and depletion of limited resources: Does self-control resemble a muscle? Psychological Bulletin, 126(2), 247-259. doi: 10.1037/0033-2909.126.2.247 pmid: 10748642
[43]	Noy S., & Zhang W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654), 187-192. doi: 10.1126/science.adh2586 pmid: 37440646
[44]	Papamitsiou Z., & Economides A. A. (2019). Exploring autonomous learning capacity from a self-regulated learning perspective using learning analytics. British Journal of Educational Technology, 50(6), 3138-3155. doi: 10.1111/bjet.12747
[45]	Pifarre M., & Cobos R. (2010). Promoting metacognitive skills through peer scaffolding in a CSCL environment. International Journal of Computer-Supported Collaborative Learning, 5, 237-253.
[46]	Shaffer D. W., Collier W., & Ruis A. R. (2016). A tutorial on epistemic network analysis: Analyzing the structure of connections in cognitive, social, and interaction data. Journal of Learning Analytics, 3(3), 9-45.
[47]	Shapiro A. M. (2004). How including prior knowledge as a subject variable may change outcomes of learning research. American Educational Research Journal, 41(1), 159-189.
[48]	Shing Y. L., & Brod G. (2016). Effects of prior knowledge on memory: Implications for education. Mind, Brain, and Education, 10(3), 153-161.
[49]	Shoufan A. (2023). Can students without prior knowledge use ChatGPT to answer test questions? An empirical study. ACM Transactions on Computing Education, 23(4), 1-29.
[50]	Simonsmeier B. A., Flaig M., Deiglmayr A., Schalk L., & Schneider M. (2022). Domain-specific prior knowledge and learning: A meta-analysis. Educational Psychologist, 57(1), 31-54.
[51]	Song Y., Huang L., Zheng L., Fan M., & Liu Z. (2025). Interactions with generative AI chatbots: Unveiling dialogic dynamics, students’ perceptions, and practical competencies in creative problem-solving. International Journal of Educational Technology in Higher Education, 22, 12. https://doi.org/10.1186/s41239-025-00508-2
[52]	Sparrow B., Liu J., & Wegner D. M. (2011). Google effects on memory: Cognitive consequences of having information at our fingertips. Science, 333(6043), 776-778. doi: 10.1126/science.1207745 pmid: 21764755
[53]	Stojanov A., Liu Q., & Koh J. H. L. (2024). University students’ self-reported reliance on ChatGPT for learning: A latent profile analysis. Computers and Education: Artificial Intelligence, 6, 100243. https://doi.org/10.1016/j.caeai.2024.100243

维度	描述	评分标准
概念节点	与主题相关的概念	正确: 得分S = 1 + 0.5 × (L - 1), 其中L表示概念节点在概念图中的层级, 且L≥1; 无关: 0分; 错误: -1分;
交叉连接	以连接线表示非父子节点的两个概念之间的意义关系	正确: 2分; 无关: 0分; 错误: -2分;
文字标注	对节点上的概念、不同节点上概念的关系的详细阐述, 以及对整幅概念图的说明。	正确: 1分; 无关: 0分; 错误: -1分;

维度	描述	评分标准
概念节点	与主题相关的概念	正确: 得分S = 1 + 0.5 × (L - 1), 其中L表示概念节点在概念图中的层级, 且L≥1; 无关: 0分; 错误: -1分;
交叉连接	以连接线表示非父子节点的两个概念之间的意义关系	正确: 2分; 无关: 0分; 错误: -2分;
文字标注	对节点上的概念、不同节点上概念的关系的详细阐述, 以及对整幅概念图的说明。	正确: 1分; 无关: 0分; 错误: -1分;

编码	交互行为类别	描述
TR	任务审视	审视写作任务的要求和目标。
AQ	提问	向系统提出问题以获取信息、解释或观点等。
FQ	追问	在收到系统的回答后, 用户基于回答的内容提出进一步的问题。
EF	评价反馈	在收到系统的回答后, 对回答内容进行评价, 包括对信息准确性、有用性的评估, 或对回答方式的反馈等。
CCP	复制粘贴内容	从系统的回答中选择部分或全部内容复制粘贴至文章中。
CPC	内容规划与构思	拟定主题大纲或核心要点, 以及规划内容架构。
CC	内容删减	对已有内容进行删减, 去除不必要或不相关的部分, 保留核心信息。
CR	内容改写	对已有内容的某些部分或句子进行重写。
ACC	自主内容创作	独立产生或在现有信息基础上进行文章内容的写作。

编码	交互行为类别	描述
TR	任务审视	审视写作任务的要求和目标。
AQ	提问	向系统提出问题以获取信息、解释或观点等。
FQ	追问	在收到系统的回答后, 用户基于回答的内容提出进一步的问题。
EF	评价反馈	在收到系统的回答后, 对回答内容进行评价, 包括对信息准确性、有用性的评估, 或对回答方式的反馈等。
CCP	复制粘贴内容	从系统的回答中选择部分或全部内容复制粘贴至文章中。
CPC	内容规划与构思	拟定主题大纲或核心要点, 以及规划内容架构。
CC	内容删减	对已有内容进行删减, 去除不必要或不相关的部分, 保留核心信息。
CR	内容改写	对已有内容的某些部分或句子进行重写。
ACC	自主内容创作	独立产生或在现有信息基础上进行文章内容的写作。

认知元素类别	描述	举例
记忆	回忆、验证基本概念与事实信息。	什么是生成式人工智能? 请介绍“教育”的定义。
理解	解释、概括信息。	人工智能对于生产实践有哪些帮助?
应用	在新情境中应用信息。	生成式人工智能时代下如何促进教师专业发展和学习?
分析	明确各概念间的相互关系。	分析一下当前的人工智能趋势及其对教师教育的影响。
评价	评价和判断信息的价值。	你怎么看待人工智能在隐私保护方面的挑战?
创造	产生新的或原创的作品。	制定一个使用人工智能提升课堂教学效果的方案。