As artificial intelligence technologies continue to evolve at a staggering pace, their applications in fields as diverse as education, communication, and historical research become ever more widespread. Yet, as AI-powered tools like ChatGPT promise unprecedented abilities to summarize and interpret vast troves of data, experts in the humanities are sounding cautionary notes. Dr. Jan Burzlaff, a postdoctoral associate specializing in Jewish Studies at Cornell University, recently published a poignant critique in the esteemed journal Rethinking History, emphasizing why human historians remain irreplaceable in an age dominated by machine learning and generative AI.
Dr. Burzlaff’s investigation centered on the AI’s handling of Holocaust survivor testimonies, a topic that serves as a profound litmus test for the capabilities and shortcomings of artificial intelligence in historical scholarship. When tasked with condensing the harrowing accounts of survivors from diverse locations such as La Paz, Kraków, and Connecticut, AI notably failed to grasp the emotional depth and moral intricacy imbued in these narratives. For instance, the testimony of Luisa D., who was seven years old during the Holocaust, recounted her mother’s desperate act of cutting her own finger to provide drops of blood to a dying child. This intimate and heart-wrenching detail was omitted entirely in the AI’s summary, underscoring a critical blind spot in current machine learning systems.
The omission of such vital emotional nuances is not an isolated oversight but illuminates fundamental limitations intrinsic to AI’s architecture. These systems rely predominantly on pattern recognition and statistical correlations drawn from vast textual datasets. While such capabilities allow them to identify themes or generate coherent overviews, they falter when confronted with the fractured, contradictory, and deeply personal fragments that constitute human history—especially histories marked by trauma and suffering. Dr. Burzlaff articulates this as the difference between simply summarizing facts and genuinely interpreting the ethical weight and emotional textures embedded in historical testimony.
Why does this matter? Historical writing transcends the mere cataloging of events; it demands a profound sensitivity to the human condition, encompassing suffering, resilience, and moral ambiguity. AI, by its very design, excels at coherence and simplification but struggles with contradictions and emotional undercurrents. This presents a danger: if machines begin to replace human historians wholesale, there is a risk that the subtlety and ethical dimensions of history may be flattened, sanitized, or distorted. In the context of Holocaust studies—the gravest testimony of human suffering in recent memory—this risk is particularly acute. Dr. Burzlaff warns that failure to faithfully engage with the emotional gravity of such histories could reverberate into the handling of even more nuanced or less widely known historical narratives.
Moreover, the study’s findings raise compelling questions about the evolving roles of historians. A recent Microsoft study ranked historians highly among professions AI might replace, given the proliferation of tools capable of rapid data processing and content generation. Yet, Dr. Burzlaff argues that historians’ unique value lies in their capacity to interpret silences, fractures, and ambiguities—elements current AI cannot authentically replicate. Historians do not merely recount; they listen, contextualize, and wrestle with contradiction, preserving the full spectrum of human experience with ethical rigor.
The article published on September 15, 2025, in Rethinking History not only critiques AI’s limitations but also offers proactive guidance. Dr. Burzlaff proposes five core principles tailored to historians and educators grappling with the integration of AI into their workflows, particularly when addressing trauma, genocide, and other sensitive topics. Chief among his recommendations is a departure from the AI’s dependency on frequency, proximity, and pattern analysis. Instead, human scholars should embrace the fragmentary and heterogeneous nature of testimonies, refusing to allow their narratives to be distilled merely into a uniform collection of texts.
This philosophy reframes the historian’s mission in the digital era: the goal is not to outcompete AI at its own game but to affirm a distinctly human mode of historical inquiry—one that foregrounds ethical considerations and stylistic particularities that machines cannot emulate. The stakes extend far beyond remembering the Holocaust accurately. They touch on the very future of historical memory and its role in shaping societal understanding of past injustices, moral failings, and resilience.
AI-driven summarization, while a powerful educational and analytical tool, inherently lacks the ability to ‘listen’—to absorb the often fragmented and contradictory nature of oral histories, eyewitness accounts, and personal testimonies. These sources do not fit neatly into the associative logic followed by large language models. Rather, they contain pauses, silences, and ruptures that are essential to preserving historical truth and moral complexity. In this context, silence is not an absence but a form of communication laden with meaning.
Semantic coherence, often prized in computational text generation, can paradoxically erode the very complexity that defines historical authenticity. AI’s drive to smooth and summarize can efface the ‘fractures’—the inconsistencies and tensions—that signal deeper layers of individual and collective experience. Such ‘fractures’ are the ethical fulcrums upon which much historical scholarship pivots, especially when dealing with trauma where memory is fragmented and pain often resists neat categorization.
Dr. Burzlaff’s analysis challenges historians to reflect on their unique competencies at a moment when AI claims increasing influence in educational environments and public discourse. While acknowledging AI’s potential for uncovering new analytical angles, he stresses that these advantages cannot substitute for the historian’s interpretive acumen. The danger lies not in the existence of AI per se, but in whether human researchers retain their commitment to ‘doing history’ in a profoundly reflective and ethically accountable manner.
His piece thus serves as both a sober assessment and a call to action. As generative AI becomes more embedded in knowledge production, historians must reckon with how technology shapes collective memory. Will historical narratives become homogenized templates optimized for clarity and coherence? Or will historians insist on preserving the fractured, painful, and sometimes ambiguous testimonies that constitute real historical experience?
In this critical juncture, Dr. Burzlaff’s work is a testament to the enduring necessity of human sensitivity in the face of technological prowess. His recommendations underline that historical writing is neither reducible to data nor replaceable by algorithms—it is an indelible human endeavor, perpetually engaged in negotiating the tension between fact, memory, and meaning.
Subject of Research:
Not applicable
Article Title:
Fragments, not prompts: five principles for writing history in the age of AI
News Publication Date:
15-Sep-2025
