What is it about?
Can AI be truly neutral, or does it inherit our historical prejudices? Our research reveals that even when AI models are "aligned" to avoid offensive language, they still perpetuate ancient stereotypes through subtle, underlying personality archetypes. Using a novel, multi-stage methodology to analyze character generation, we found that models consistently associate Jewish identity with tropes of the "isolated genius" or the "cold-blooded manipulator," embedding historical biases into the digital narratives they create.
Featured Image
Photo by Growtika on Unsplash
Why is it important?
As AI becomes a primary tool for critical decision making in areas such as hiring, legal services, and educational support, these invisible biases carry substantial real world consequences. Because this form of prejudice bypasses standard safety filters by appearing as personality traits rather than hate speech, it is uniquely difficult to detect and address. Our work highlights a critical systemic flaw. Without addressing these deep seated cultural patterns, we risk automating and scaling historical prejudices under the guise of objective technology.
Perspectives
By analyzing LLMs as vast repositories of our long term digital history, we leveraged their massive multi year training data to expose how AI reassembles human cultural patterns. Using a multi stage process of decomposition and reconstruction, our methodology mapped underlying statistical associations to specific recurring archetypes, such as characters reminiscent of Sherlock Holmes, Dr. House, and Walter White, all of whom share a pattern of high intelligence combined with social alienation and moral ambiguity. These findings suggest that AI is not merely processing data. Rather, it is normalizing and modernizing ancient cultural myths, transforming them into new digital realities that require urgent and reflexive design.
Gal Gutman
Ben-Gurion University of the Negev
Read the Original
This page is a summary of: From myth to model: Representation of “the Jew” in generative AI., American Psychologist, May 2026, American Psychological Association (APA),
DOI: 10.1037/amp0001668.
You can read the full text:
Contributors
The following have contributed to this page
