Dr. Jan Burzlaff, a postdoctoral associate in the Jewish Studies Program at Cornell University, sounds a dire alarm. His criticism touches on AI’s inability to understand historical story-arcs, particularly complex stories that include trauma. A leading expert on Nazi Germany, Dr. Burzlaff most recently challenged ChatGPT to synthesize the lived recovery journeys of Holocaust survivors. The results were indeed alarming, as the AI was not able to pick up on essential emotional and intimate nuances from the testimonies.
Dr. Burzlaff shared his analysis with the public in the journal Rethinking History. He appraised how accurately ChatGPT summarized recorded testimonies of Holocaust survivors from La Paz, Kraków, and Connecticut, collected in 1995. He argues that this failure highlights a broader issue: if AI cannot accurately portray the most extreme cases of suffering in history, it will likely distort more nuanced historical accounts as well.
AI’s Limitations in Capturing Trauma
Dr. Burzlaff’s findings indicate that the AI failed to capture significant emotional distress in applicant personal narratives. These stories are critical to the full story of the Holocaust. It was an astute observation on the largest drawback with ChatGPT. Or a poignant detail from the written testimony of seven-year-old survivor Luisa D. According to Dr. Burzlaff, the AI neglected to mention that Luisa’s mother cut her own finger to provide her dying child with drops of blood to stay alive—”the faintest trace of moisture.”
“Essentially it ignored the extent these individuals suffered on an emotional level,” – Dr. Burzlaff
This omission calls into question the entire premise of using AI as a trustworthy tool for historical research. Dr. Burzlaff makes a case for Holocaust testimony to serve as a “litmus test for AI.” He especially foregrounds the tension between the desire for true-to-life representation and AI’s urge to flatten and condense stories. He reminds us that historians need to be sensitive to the ethical burden and silence embedded in these testimonies.
The Importance of Human Experience
She feels that the depth and quality of historical narratives just can’t be calculated down to algorithms and recurring patterns. Ordinarily categorized, which suggests that historians should stop confining the idiosyncrasies of human experience to formalized boxes and molds.
“The accounts of people from the past differ according to their individual experiences and some are different to categorize. Historians need to embrace this lack of uniformity and moments of human experience that algorithms cannot anticipate,” – Dr. Burzlaff
He warns that as tools like ChatGPT become more prevalent in education, research, and public discourse, historians face a critical challenge: determining what these systems can accurately convey.
“As tools like ChatGPT increasingly saturate education, research, and public discourse, historians must reckon with what these systems can and cannot do,” – Dr. Burzlaff
Guidelines for Historical Writing in the Age of AI
Dr. Burzlaff was well aware of the difficulties of crafting history in our current climate. In response, he created five strong principles for educators to consider when teaching about trauma, genocide, and other historical injustices. These new standards reflect the importance of maintaining the emotional complexity and moral urgency that often inform historical stories.
“Essentially, as historians we should not try to ‘outperform the machine’ but to sound nothing like it,” – Dr. Burzlaff
His article, titled “Fragments, not prompts: five principles for writing history in the age of AI,” outlines these principles and calls for historians to draw from written testimonies instead of becoming mere collections of texts.
We’re grateful for Dr. Burzlaff for raising this important question. His point is, if a machine could do it great – then we know we really misread what should count as good history.
“If historical writing can be done by a machine, then it was never historical enough,” – Dr. Burzlaff