Emerging trends in historical scholarship in age of AI
The quality of scholarly output will depend largely on the ability of researchers to collaborate effectively with AI. Photo: TUCHONG
Over the past two years, the advent of large language models (LLMs) such as ChatGPT and DeepSeek has vividly demonstrated the rapid progress of artificial intelligence (AI) technology. AI is poised to redefine the boundaries of human knowledge, transform cognitive processes, reshape the political and economic structures of society, and profoundly influence academic disciplines—including history.
AI empowering humanities and social sciences
With the fast-paced development of AI, several new trends have emerged in historical research.
A shift from text-driven to data-driven approach: Traditional historical studies have relied on scholars individually reading and analyzing texts, a process inherently constrained by the finite amount of material any individual can absorb in a lifetime. AI, however, can process vast datasets, allowing scholars to examine patterns and underlying historical logic across extended time periods and larger spatial scales.
Overcoming language barriers: In the past, historians typically invested enormous effort in reading and processing foreign-language texts. Today, the advent of LLMs has greatly reduced these linguistic obstacles, and future improvements in AI accuracy may extend their utility to more challenging tasks such as deciphering ancient scripts and reconstructing fragmentary manuscripts.
Facilitating reading and text analysis: Current mainstream AI models already excel at organizing literature, summarizing key arguments, and assisting with tasks such as literature reviews, research design, data collection, and theoretical development. By streamlining these processes, AI empowers historical scholarship as well as research in other humanities and social sciences.
Human-AI collaboration: In the future, AI is expected to take on an increasingly prominent role in collaborative research, while human scholars act as mentors—guiding AI systems, formulating questions, and directing research through prompts and oversight. Effective collaboration with AI will require researchers to develop proficiency in engaging with these models, alongside strong oversight and imagination. The quality of scholarly output will depend largely on the ability of researchers to collaborate effectively with AI.
Transition from “digital humanities” to “intelligent digital humanities”: Digital humanities integrates theories, methods, and tools from disciplines such as literature, history, geography, computer science, as well as library and information science. As AI continues to advance, digital humanities will evolve by leveraging the immense computational power of LLMs, along with their model-building, multimodal simulation, and automated analysis capabilities, transforming into “intelligent digital humanities.”
Historical sources shifting from texts to data: As production activities and daily life at both the individual and societal levels become increasingly digitized, historical sources are shifting from tangible textual archives to intangible digital datasets. Historians who once navigated between physical archives will now navigate through various databases.
Enhancing risk awareness in research
While AI provides powerful tools for humanities and social science scholarship, it also introduces certain risks that warrant serious attention.
“Data deserts”: Globally, substantial regional and temporal disparities exist in the digitization of historical sources. However, the absence of digital records does not equate to an absence of history. Greater emphasis should be placed on overlooked groups and regions, as well as the diversity of historical sources. In addition to textual data, images, archaeological artifacts, climate records, environmental data, and other forms of information should be integrated to avoid biases and misinterpretations arising from overreliance on textual data.
Hidden biases: AI models are prone to inadvertently reproducing and amplifying biases embedded within their training data, potentially distorting representations of race, gender, linguistic, or cultural groups. Researchers in the humanities and social sciences must remain vigilant against ideological and value-laden biases lurking in emerging technologies.
Research ethics and norms: Many universities worldwide have already implemented regulations on AI-generated content. Several major academic publishers also prohibit listing LLMs like ChatGPT as authors. It is foreseeable that norms governing the use of AI in academic research and publishing will continue to be updated and refined.
Anxieties over the displacement of humans: Historians remain active agents in generating source materials, curating datasets, selecting and framing research questions, providing historical interpretations, and assigning historical value to events—roles that AI, despite its capabilities, has yet to fully replicate. Their humanistic perspective, with its temporal sensitivity, remains irreplaceable.
The rise of AI may usher in a post-human future, catalyzing transformative changes worldwide. This calls for more insightful historical scholarship that narrates and records the grand story of human destiny to preserve humanistic values and alleviate existential anxieties. Academic disciplines should collaborate to navigate these transformations and challenges, collectively shaping the present and the future of human history.
Wu Chou is a lecturer from the School of History at Beijing Normal University.
Edited by WANG YOURAN