根據教育部《十二年國民教育課程綱要》中可以發現歷史教學目標發展之梗概乃由過往對闡釋歷史事實的重視，轉而強調教導學生認識歷史知識的特質，並培養探究、分析、詮釋、反思及創新的態度與能力，而在讀者解讀多文本的史料閱讀歷程中，Wineburg (1990)提出的歷史多文本閱讀三種策略─佐證、探究史源、脈絡化，其有助於解讀和批判歷史的多文本資料。隨著現今科技的發展，目前針對生成式AI－ChatGPT解讀歷史事件的研究中，尚未探究不同讀者使用ChatGPT解讀歷史多文本之閱讀歷程及使用策略，特別是針對ChatGPT的分析上，針對受試者的思考模式與認知歷程的相關研究更是相當有限。
本研究之研究對象為來自台灣八位史學相關研究者和八位台灣大專院校之大學生，實驗過程全程採用放聲思考法（think aloud method）深入紀錄受試者從閱讀多文本到使用ChatGPT的思考模式與認知歷程，針對受試者之回答進行收集與分析，以衡量台灣史學相關研究者和非史學相關研究者在解讀歷史事件及使用ChatGPT時的認知過程差異。本實驗操作時間為2024年7、8月，採用研究當下最新之版本ChatGPT4o，受試者將會再透過匹配抽樣進行基礎和進階的ChatGPT使用教學，基礎組僅學習ChatGPT簡要的功能說明，而進階組除基礎功能外，還會進行詳細ChatGPT指令之影片教學，針對角色設定、結構化的提問、延伸詢問及其他注意事項進行教學與示範，隨後分析受試者透過放聲思考後的紀錄，比較不同的訓練後，基礎組與進階組利用ChatGPT提供觀點和解釋的過程差異。資料收集完成後，資料編碼轉譯遵循但不限於Wineburg（1990）提出之閱讀策略和Ferguson等人(2012)針對知識證成的分類所提出的編碼準則。
研究結果顯示，解讀書面資料的歷程分析，史學組在運用Wineburg（1990）所提出的「佐證、探究史源、脈絡化」三種策略次數上皆顯著高於非史學組，展現出更豐富的交叉比對能力與批判思維，史學組傾向從多元來源進行知識證成，並能檢視文本作者背景與時代脈絡；非史學組則較易依賴權威資料與個人直觀判斷，缺乏深入質疑與多文本比對的行為，在建構連貫的歷史脈絡時則顯得較為被動。史學組更常使用「描述、分析、限定說明」等策略，能進一步探究圖片可能的政治與文化意涵，並辨別潛藏的偏見；非史學組則多停留在表層特徵描述與情緒反應，較少進行深層次的批判與解釋。於資料可信度排序任務中，史學組能正確識別關鍵差異顯著高於非史學組，且史學組多優先考量一手資料（如官方檔案、當時報紙），並認為課本因簡化目的而可信度偏低；非史學組則較依賴個人知識關聯度來判斷文本可靠性。
在運用ChatGPT進行歷史事件分析時，受試者下達指令經編碼後分析為五大類別「第一層序提問、第二層序提問、追問、評價回饋、外部驗證」，史學組傾向提出第二層序提問、追問與評價回饋等，顯示其較強的批判性與探究性；非史學組的提問多集中於第一層序提問，進行單一事實查詢，缺乏對生成內容進行再探討的動機與能力，另外值得注意的是，進階教學對非史學組的大學生產生明顯影響，其指令字數顯著提升，在第一句指令的目的性上也更傾向開放式探索；然而，史學組在教學前後的指令複雜度差異有限，顯示其基礎提問能力已相對成熟。此結果顯示，教師在教學中應注重基礎知識與批判性思維的同步培養，並透過設計專門的訓練活動，提升學生在使用生成式AI進行多文本閱讀時的分析與整合能力。

According to the Ministry of Education's Curriculum Guidelines for 12-Year Basic Ed-ucation, the goals of history education have shifted from emphasizing the interpretation of his-torical facts to fostering students' understanding of the nature of historical knowledge. These goals aim to cultivate inquiry, analysis, interpretation, reflection, and innovation. In the process of interpreting historical multi-text materials, Wineburg (1990) proposed 3 historical reading strategies—corroboration, sourcing, and contextualization—which are instrumental in the anal-ysis and critique of multi-text historical data. With the advancement of technology, research on the application of generative AI, such as ChatGPT, in interpreting historical events remains lim-ited, particularly regarding its impact on the reading processes of different audiences and the strategies employed when using ChatGPT.
In this study, eight historical researchers and eight university students from Taiwan were recruited as participants. The experiment adopted a think-aloud method throughout, cap-turing in depth each participant’s thought processes and cognitive development as they pro-gressed from reading multiple historical texts to using ChatGPT. By collecting and analyzing the participants’ responses, this research aimed to evaluate differences in cognitive processes between historical researchers and non-historical participants in Taiwan, specifically regarding their interpretation of historical events and use of ChatGPT. The experiment was conducted in July and August of 2024, employing ChatGPT-4o, the latest version available at the time of the study. Participants were further matched into basic or advanced ChatGPT training groups. In the basic group, participants were introduced only to the fundamental functions of ChatGPT, whereas the advanced group received additional training via instructional videos on specialized ChatGPT commands, including role setting, structured questioning, extended inquiries, and other relevant considerations. The subsequent analysis examined participants’ recorded think-aloud data to compare differences in how the basic and advanced groups used ChatGPT to pro-vide perspectives and explanations. The data collected through the think aloud method were transcribed and coded according to Wineburg (1990) reading strategies and the knowledge jus-tification categories proposed by Ferguson et al. (2012). These guidelines provided the frame-work for analyzing and comparing participants’ processes of reading historical multiple-texts and using ChatGPT.
Results indicate that, in the process analysis of reading written materials, the historical group employed significantly higher frequencies of Wineburg’s (1990) three strategies—corroboration, sourcing, and contextualization—than the non-historical group, demonstrating more extensive cross-referencing capabilities and critical thinking. The history group tends to conduct knowledge justification by multiple sources, closely examining the background and historical context of text authors; by contrast, the non-historical group relies more on authorita-tive sources and personal intuitive judgments, showing less inclination toward in-depth ques-tioning or multi-text comparison, and appearing more passive in constructing coherent historical contexts. Moreover, the historical group more frequently adopted higher-level strategies such as “description, analysis, and bounded explanation,” thereby exploring potential political and cul-tural connotations in images and identifying underlying biases; the non-history group, however, remained more focused on surface features and emotional reactions, with limited deeper critique or interpretation. In a credibility-ranking task, the history group’s rate of accurately identifying key differences was significantly higher than that of the non-history group. The history group generally prioritized primary sources (e.g., official documents, contemporary newspapers) and considered textbooks less credible due to their simplified nature. By contrast, the non-history group tended to rely on personal knowledge relevance to determine textual reliability.
When using ChatGPT for historical event analysis, participants’ instructions were cod-ed into five main categories: “first-order questions, second-order questions, follow-up queries, evaluative feedback, and external verification.” The historical group tended to pose second-order questions, follow-up queries, and evaluative feedback, indicating stronger critical and in-vestigative tendencies. The non-historical group, on the other hand, primarily focused on first-order questions—seeking single factual answers—lacking the motivation or ability to probe further into the content generated. Notably, advanced training significantly affected the universi-ty students in the non-historical group, substantially increasing their average instruction length and shifting their initial instructions toward more open-ended exploration. However, the histor-ical group exhibited minimal changes in instruction complexity before and after the instructional intervention, suggesting that their foundational questioning ability was already relatively mature. These findings suggest that educators should emphasize the simultaneous cultivation of founda-tional knowledge and critical thinking, and design specialized training activities to strengthen students’ analytical and integrative abilities when using generative AI for multi-text reading.

中文文獻
宋佩芬 （2008）。講述中的歷史思維教學: 一些可能與問題。師大學報：教育類，53(1)，175-197。
李涵鈺、宋家復（2016）。像史家一般閱讀——在中學歷史課堂裡教讀寫素養。教科書研究， 9(3)， 165-184。
林慈淑（2010）。歷史要教什麽?: 英美歷史教育的爭議。臺灣學生書局。https://books.google.com.tw/books?id=8y6NZwEACAAJ
張元（2015）。利用文本閱讀題進行教學的幾點想法。清華歷史教學，24，51-64。
教育部 （2018）。十二年國民基本教育課程綱要。台北市。
陳玟樺、劉美慧（2020）。探究為本的國中歷史教學實作: 以芬蘭一間國際文憑學校課室為例。課程與教學，23(4)，59-88。
譚光鼎（2008） 。被扭曲的他者：教科書中原住民偏見的檢討 [The Twisted Others: A Review of Biases against Aboriginals in Taiwan's Textbooks] 。課程與教學，11(4)，27-49。 https://doi.org/10.6384/ciq.200811.0027 。

外文文獻
Bråten, I., & Strømsø, H. I. (2006). Epistemological beliefs, interest, and gender as predictors of Internet-based learning activities. Computers in Human Behavior, 22(6), 1027-1042.
Ericsson, K. A., & Simon, H. A. (1993). Protocol analysis verbal reports as data (Rev. ed.). MIT Press.
Ferguson, L. E., Bråten, I., & Strømsø, H. I. (2012). Epistemic cognition when students read multiple documents containing conflicting scientific evidence: A think-aloud study. Learning and Instruction, 22(2), 103-120. https://doi.org/10.1016/j.learninstruc.2011.08.002
Flower, L., & Hayes, J. R. (1980). The cognition of discovery: Defining a rhetorical problem. College composition and communication, 31(1), 21-32.
Gao, J., Ding, X., Qin, B., & Liu, T. (2023). Is chatgpt a good causal reasoner? a comprehensive evaluation. arXiv preprint arXiv:2305.07375.
Gao, J., Zhao, H., Yu, C., & Xu, R. (2023). Exploring the feasibility of chatgpt for event extraction. arXiv preprint arXiv:2303.03836.
González-Gallardo, C. E., Boros, E., Girdhar, N., Hamdi, A., Moreno, J. G., & Doucet, A. (2023, 26-30 June 2023). Yes but.. Can ChatGPT Identify Entities in Historical Documents? 2023 ACM/IEEE Joint Conference on Digital Libraries (JCDL),
Greene, J. A., Azevedo, R., & Torney-Purta, J. (2008). Modeling Epistemic and Ontological Cognition: Philosophical Perspectives and Methodological Directions. Educational psychologist, 43(3), 142-160. https://doi.org/10.1080/00461520802178458
Hartman, D., & Allison, J. (1996). Promoting inquiry-oriented discussions using multiple texts. Lively discussions, 106-133.
Hofer, B. K. (2004). Epistemological understanding as a metacognitive process: Thinking aloud during online searching. Educational psychologist, 39(1), 43-55.
Hofer, B. K., & Bendixen, L. D. (2012). Personal epistemology: Theory, research, and future directions. In K. R. Harris, S. Graham, T. Urdan, C. B. McCormick, G. M. Sinatra, & J. Sweller (Eds.), APA educational psychology handbook, Vol 1: Theories, constructs, and critical issues. (pp. 227-256, 621 Pages). American Psychological Association, American Psychological Association. https://doi.org/https://doi.org/10.1037/13273-009
Hofer, B. K., & Pintrich, P. R. (1997). The development of epistemological theories: Beliefs about knowledge and knowing and their relation to learning. Review of educational research, 67(1), 88-140.
Hynd, C. R. (1999). Teaching students to think critically using multiple texts in history. Journal of Adolescent & Adult Literacy, 42, 428-436.
Lee, P., & Ashby, R. (2000). Progression in historical understanding among students ages 7-14. In (pp. 199-222).
Lee, P. J. (2005). Putting principles into practice: Understanding history. National Academy Press, pp. 31-78.
Mason, L., & Boldrin, A. (2008). Epistemic Metacognition in the Context of Information Searching on the Web. In M. S. Khine (Ed.), Knowing, Knowledge and Beliefs: Epistemological Studies across Diverse Cultures (pp. 377-404). Springer Netherlands. https://doi.org/10.1007/978-1-4020-6596-5_18
Newell, A., & Simon, H. A. (1972). Human problem solving (Vol. 104). Prentice-hall Englewood Cliffs, NJ.
Nguyen, X.-H., Nguyen, H.-A., & Cao, L. (2023). ChatGPT's Understanding of History A Comparison to Vietnamese Students and its Potential in History Education.
Nokes, J. D., Dole, J. A., & Hacker, D. J. (2007). Teaching high school students to use heuristics while reading historical texts. Journal of Educational Psychology, 99(3), 492-504.
Perry Jr, W. G. (1999). Forms of Intellectual and Ethical Development in the College Years: A Scheme. Jossey-Bass Higher and Adult Education Series. ERIC.
Tiunova, A., & Muñoz, F. (2023). Chatgpt: Using Ai in Social Studies Academic Research. Available at SSRN 4451612.
Whitmire, E. (2003). Epistemological beliefs and the information-seeking behavior of undergraduates. Library & Information Science Research, 25(2), 127-142.
Williamson, B., Macgilchrist, F., & Potter, J. (2023). Re-examining AI, automation and datafication in education. In (Vol. 48, pp. 1-5): Taylor & Francis.
Wineburg, S. S. (1990). Historical problem solving : a study of the cognitive processes used in the evaluation of documentary evidence Thesis (Ph. D.)--Stanford University, 1990.].
Wu, T., He, S., Liu, J., Sun, S., Liu, K., Han, Q. L., & Tang, Y. (2023). A Brief Overview of ChatGPT: The History, Status Quo and Potential Future Development. IEEE/CAA Journal of Automatica Sinica, 10(5), 1122-1136. https://doi.org/10.1109/JAS.2023.123618
Yell, M. M. (2023). Social Studies, ChatGPT, and Lateral Reading. Social Education 87, 138-141.