Introduction

In today’s era, artificial intelligence (AI) has permeated every aspect of human life, profoundly changing how we understand and reshape the world. In academic research, AI offers efficient text processing and outstanding content mining capabilities, but it also presents inherent limitations and ethical risks, making it a hot topic across disciplines. This article invites three young scholars engaged in different national studies to discuss how AI is applied in world history research, its potential to expand research boundaries, and the challenges faced by the younger generation of historians in coexisting with AI.

How AI Drives World History Research

Moderator: In recent years, AI technology has rapidly developed, prompting scholars from various disciplines to explore its application potential in their fields, including world history. Can each of you share how AI plays a role in your specific research area?

Wang Sijie: In my research on German history, the application of AI in both Chinese and foreign historiography mainly focuses on optical character recognition and transcription of historical manuscripts and archives, as well as content mining using topic modeling and text reuse detection. AI has significantly advanced existing digital history work, such as identifying implicit relationships and intermediary nodes in social network analysis of archives, and compensating for missing geographic information. Although digital historians have long used programming languages for frequency statistics and co-occurrence analysis to identify potential themes, these methods are often limited to statistical associations at the word level, making it difficult to capture deeper historical representations like semantic evolution and rhetorical differences. Recent advances in deep learning pre-trained language models can transform text into vector structures that reflect contextual meanings, identifying the same historical theme under different expressions and generating explanatory summaries or labels.

Yao Nianda: In the field of American history, AI applications extend beyond large language models to encompass a comprehensive set of computational analysis methods centered on natural language processing and machine learning. This approach quantifies diverse historical materials, such as newspapers and government documents, using topic modeling, text embedding, and semantic analysis to reveal long-term changes in language, concepts, and political discourse, providing new clues and evidence for historical interpretation. For instance, the Stanford team led by Nikil Garg analyzed 20th-century corpora to quantify shifts in gender and ethnic stereotypes in language, linking them to transformations in social structures. Another study by Melissa Lee tracked the transition of the term “United States” from a plural to a singular usage in 19th-century newspapers and congressional debates, reflecting changing understandings of national sovereignty.

Yi Jinming: Recently, the integration of AI in medieval European history has focused on automating the transcription, completion, and structural analysis of medieval materials, enhancing the readability and retrievability of ancient texts. For example, Transkribus is one of the most commonly used tools for handwritten text recognition in European academia. Additionally, knowledge graph and semantic web technologies are used to structure relationships among people, places, and institutions found in charters, ledgers, and letters into queryable data networks. A research team from Spain proposed establishing a knowledge graph for medieval charters, combining expert annotations and community contributions to support systematic analysis of medieval social, legal, and economic relationships. Large language models are also used for text completion of Latin inscriptions, such as Aeneas, which is trained on about 200,000 Latin inscriptions to help scholars interpret damaged or missing historical texts.

The Limitations of AI in World History Research

Moderator: While AI significantly enhances research efficiency, it also has notable limitations. What are the current challenges AI faces in historical research?

Yao Nianda: There are several bottlenecks in applying AI to historical research. These challenges reflect a structural mismatch between current AI technology and historical research rather than mere technical immaturity. First, AI struggles to resonate emotionally with human society. As Croce noted, all history is contemporary history. A vital historical research topic often responds to contemporary social issues and evokes emotional resonance in readers. Thus, determining which historical questions are meaningful today relies heavily on researchers’ sensitivity to public issues and human experiences. AI can summarize existing discussions but cannot genuinely understand the emotional connections between historical issues and human life.

Second, AI faces unavoidable semantic drift when analyzing historical texts. Most language models are trained on contemporary corpora, leading to potential misinterpretations of past language practices. Even attempts to train models on historical corpora are limited by the incompleteness and imbalance of existing historical texts. Moreover, AI’s value judgments are not neutral; they inevitably reflect mainstream norms and contemporary values from the training data. When these models are used in historical research, they may inadvertently measure the past against contemporary standards, weakening historical context.

Lastly, a critical bottleneck is the “black box” nature of AI. In many cases, humanists struggle to explain how AI arrives at certain conclusions. For disciplines that prioritize explainability and discussability, a lack of clarity in the analysis process makes it difficult for researchers to take academic responsibility for their conclusions.

Yi Jinming: In text analysis, AI is primarily applied to well-preserved and digitized materials, such as contracts and correspondence, while its application in other areas remains limited. This limitation stems from two main reasons. First, AI model training heavily relies on large-scale, readable data. For instance, a study by Fabio Gatti’s team at the University of Bern utilized over 6,000 letters to analyze the banking correspondence network of Florentine merchants. However, many medieval materials do not reach such scale and quality. Second, medieval texts often suffer from complex handwriting, numerous abbreviations, and poor preservation, increasing the costs of text recognition and transcription. Although platforms like Transkribus have improved large-scale reading possibilities, training and proofreading still require significant human input, leading researchers to prefer using already organized material databases.

Wang Sijie: As mentioned, the unevenness of corpora affects the scope of AI usage. A similar issue arises from the training data of general large language models, which predominantly comes from the English-speaking world, leading to a Western-centric perspective in historical narratives. AI still struggles with semantic recognition and comprehension of long and complex sentences in lesser-used languages. Furthermore, the digitalization and open access advantages of English and American archives facilitate automated batch retrieval and deep processing for historians. This “digital divide” is particularly prominent in transnational history research, where scholars tend to use easily accessible and highly structured English and American materials, impacting the overall understanding of historical events.

Coexisting with AI in Historical Research

Moderator: Given the limitations of AI, what methods can be employed to address these challenges?

Yao Nianda: The fundamental solution to these limitations is to anticipate technological advancements that can eliminate these issues. However, a more realistic approach for humanists is to mitigate these limitations through methodological design and research norms, ensuring that AI remains controllable and verifiable. First, it is crucial to maintain the leading role of human researchers in the problem-setting phase. Decisions about which historical questions are worth asking and why they are significant must stem from the researchers’ understanding of contemporary society and historiographical traditions, rather than being generated by models. Second, when using AI to analyze historical texts, researchers must clearly distinguish between contemporary language models and historical language, striving to restore the historical context of the materials. Lastly, in light of AI’s “black box” nature, historians should enhance the transparency and accountability of the research process. Even if algorithms are not fully explainable, researchers should clarify the types of models used, the scope of the data, and the analysis steps, ensuring that the research path remains traceable and that conclusions can be subjected to academic scrutiny.

Wang Sijie: We could attempt to construct specialized models for specific fields, such as those serving early American history or German historiography. These specialized models can utilize retrieval-augmented generation (RAG) techniques to conduct material searches through local structured knowledge bases, anchoring context while ensuring quality and controllability. Specialized models possess independent memory and parameters, allowing for deep training on specific languages and historical backgrounds. Importantly, local knowledge bases can include diverse historical narratives, enabling researchers to incorporate insights from local historians into prompts to counteract potential geopolitical biases in the models.

Yi Jinming: AI should be viewed as a hypothesis-generating tool rather than a conclusion-verifying tool. To prevent AI from becoming merely an efficiency tool for existing historiographical propositions, it is essential to redefine its methodological role. Rather than using models to validate established economic trends or institutional judgments, we should position them as mechanisms for generating hypotheses, actively identifying historical issues that have not been adequately explained by theoretical frameworks. For instance, algorithms can reveal latent networks of low-frequency individuals across regions or identify semantic combinations of unconventional contractual clauses. These outputs do not directly constitute historical conclusions but provide historians with new clues and research directions, which can then be interpreted and validated in the context of archival materials and institutional backgrounds.

Moderator: In the context of AI profoundly influencing academic research paradigms, how should young world history researchers seek a balance between adhering to historiographical traditions and embracing technological changes?

Yi Jinming: As AI gradually enters historical research practices, the importance of historiographical training has not diminished; rather, it has become more pronounced. First, the formation of problem awareness relies on long-term historiographical training rather than mere technical proficiency. Truly innovative research often arises from questioning and reconstructing existing explanations, a skill cultivated through familiarity with historiographical traditions, theoretical lineages, and methodological debates. Without an understanding of the history of historiography, it becomes challenging to discern whether a pattern generated by AI represents a “new discovery” or a “repetition of old problems.” Second, historiographical training fosters a keen awareness of the absence of voices, marginalized groups, and unrecorded narratives in historical research. Only scholars with extensive historiographical training can recognize which groups are systematically absent in contracts or administrative documents and design supplementary paths accordingly. Lastly, the ability to critique sources is indispensable. Regardless of how many text patterns a model identifies, researchers must evaluate whether these patterns stem from archival generation mechanisms or preservation biases. Therefore, while actively utilizing AI technologies, historians should prioritize traditional historiographical training.

Wang Sijie: Young scholars should allow AI to handle preliminary tasks such as document screening, text recognition, and literature translation, focusing their efforts on more creative interpretative aspects. As archival materials continue to be made publicly available and digitized, young scholars can gradually build a personal knowledge base composed of structured materials and diverse scholarly outputs, transitioning from readers of archives to managers of data. Supported by RAG technology, personal knowledge bases can search, recognize semantic associations, and integrate research perspectives across multilingual corpora, significantly enhancing work efficiency. Additionally, young scholars should actively explore potential applications of AI in history, such as engaging in dialogues with historical figures based on letters, diaries, and writings, or simulating key wartime decisions or diplomatic negotiations through historical reenactments. These applications can not only assist in historical teaching but also inspire researchers’ academic creativity.

Yao Nianda: I believe that the relationship between world historians and AI should not be viewed as adversarial or substitutive but rather as a conscious coexistence with boundaries. It is essential to emphasize that the importance of human agency in research does not negate technology. Historians are not difficult to replace by machines, not merely because technology is still maturing, but because their core value derives from the researchers’ awareness of questions and the significance they assign to history. Therefore, humanists do not need to prove their irreplaceability by rejecting AI. At the same time, we must be cautious of another extreme tendency: the high efficiency brought by AI may inadvertently diminish researchers’ subjectivity. If researchers rely solely on models to generate conclusions, summaries, or analytical paths, research may devolve into merely organizing and restating model outputs. The key to coexisting with AI lies in clearly distinguishing between enhancing labor efficiency and substituting human thought.

Expert Commentary

Wang Tao, Professor at Nanjing University: The transformation of research methods in history tends to be slow, but it does not reject methodological updates. The emergence of new historiography and various historiographical schools indicates an active engagement with interdisciplinary thinking. If Sima Qian could traverse to the present and see young historians discussing AI in historical research, he would likely experience a familiar strangeness.

The strangeness lies in the high-tech jargon that can be overwhelming. Since the advent of quantitative history, methodologies like digital humanities, big data, spatial analysis, and text mining have emerged, and now, under the impact of AI, terms like large language models and intelligent historiography are being coined. The technological shift in historiography should be validated. Historians are not seeking technology for its own sake but hope that tedious research work can be enhanced by technology. Whether capturing semantics from vast texts or transcribing manuscripts, these are areas where large language models can excel. Young scholars, being in the early stages of their careers, are naturally more sensitive to this discussion and may feel excited about it, as they need to publish papers efficiently and quickly establish their academic reputation.

If Sima Qian were to enter the AI era, he might not understand the technical concepts mentioned by the three young scholars, but he would certainly notice that, beneath the technological glamour, they are still discussing the comprehensibility, discussability, significance, and evaluation of historiography. This is a familiar topic for him, and he could join the lively discussion of the three young scholars, perhaps adding a note of his own.

It is encouraging that young scholars, while closely following the latest methodologies, remain guided by the core of historiography to define or evaluate the effectiveness and limitations of AI. They emphasize that, in the context of AI entering historical research, foundational historiographical training should not be neglected, which is a crucial reminder. Only in this way can historical research counter the illusions brought by AI and the exacerbated “digital divide,” breaking through the “black box” of technology.

Nevertheless, traditional historiographical methodologies and developmental inertia are becoming increasingly untenable. Undoubtedly, for comprehensive research methodologies, history may no longer exist. AI undoubtedly leads in completing comprehensive and summary academic reviews. The future development path and how to maintain technological control, such as the application of retrieval-augmented generation technology in world history research, require more historians to explore and advance through practice. They also emphasize the subjectivity of historians, asserting that the value of historical research comes from human creativity. This understanding is crucial. While some scholars have discussed that history written by humans may not necessarily be human history, we should insist that human history must be written by humans. Writing history aims to achieve a sympathetic understanding of historical figures and empathize with them. If AI participates in the entire process of historical research, why should human readers read a history written by a non-human species? Merely because it is more fluent or interesting?

Zhao Xiurong, Professor at Renmin University of China: The core value of AI lies in its ability to process and analyze large-scale data, designed to handle the primary materials cherished by historians. This includes, but is not limited to, natural language processing, topic modeling, social network analysis, and geographic information systems.

The three young scholars affirm that historians can enhance research efficiency by leveraging AI. Indeed, a significant amount of historical materials has been digitized and transformed into fully searchable corpora, including newspapers, journals, diaries, and even manuscript archives. The construction of various databases has surpassed human cognitive capabilities, making it impossible to read and analyze these materials using traditional close reading methods. For instance, the “Tomason pamphlet” is a collection of documents compiled by 17th-century London bookseller George Tomason, containing 22,255 pamphlets, flyers, manuscripts, books, and newspapers published between 1640 and 1661. This collection is considered one of the treasures of the British Library and an invaluable resource for studying the history of the English Civil War. Clearly, reading and organizing these materials exceeds the capacity of any historian, as French historian Christian Henriot noted, unless historians master the necessary skills to navigate this complex and unknown realm, this “information-rich world” will remain out of reach.

The young scholars also recognize the limitations of AI. One is the bias in algorithms brought by AI, which resembles biases in archives. AI can reflect and even amplify existing biases in archives, such as those related to race, gender, and colonialism, highlighting the crucial role of historians. Second, the “black box” problem of AI poses a fundamental challenge to verifiable historical research, as many AI systems are opaque, meaning their internal decision-making processes are not transparent even to their designers. Some AI systems have begun to address this issue by establishing mechanisms for human participation in verification and correction.

Thus, historians are not passive consumers of AI; their unique disciplinary training enables them to identify the problems brought by AI. For example, the biases arising from training AI on modern languages are familiar to historians, as archives often conceal biases, making it challenging to find materials written by women, children, or lower-class individuals before the Victorian era. Regarding the “black box” issue, the training in historical writing methods can effectively overcome this problem, as professional historical writing has been based on the principle of showcasing the sources used through footnotes since the 19th century. The call for AI to be annotated is an extension of the footnote principle into the 21st century.

AI can discover patterns but cannot explain why these patterns are significant, nor can it craft engaging and meaningful historical narratives. AI can generate models but cannot provide contextual interpretations or conduct source critiques, nor can it assess the biases hidden within sources. This means that using AI comes with significant responsibilities. Assisting research with AI requires adopting a new, more rigorous critical framework. The traditional skills of historians are not outdated; rather, they have become more crucial than ever in the age of AI. The profound and long-standing critical tradition of history provides a solid intellectual foundation for addressing the most challenging issues posed by AI.

AI is a transformative technology that is changing the tools used by historians and broadening their research horizons. The ultimate value of AI in historical research lies in enhancing historians’ skills, enabling them to explore broader historical contexts and write richer, more data-driven, and detailed histories than ever before. However, it is essential to remember that AI cannot think like historians, ask questions, or judge which topics hold research value. Therefore, in the age of AI, the humanistic qualities of historians become increasingly invaluable.