AI and the Study of History

Though GenAI may be helpful in particular instances (such as in the transcription and translation of primary sources), for undergraduate students in particular, the detriments of its use far outweigh the benefits. Frequent use of GenAI results in “cognitive offloading,” a tendency to rely on GenAI tools to solve problems, which in turn reduces students’ capacity for critical thinking. The key pedagogical problem with GenAI is that its widespread adoption replaces the very skills and aptitudes students develop in our discipline, including the practice of brainstorming ideas, developing good research questions, learning how to locate appropriate primary and secondary source materials, evaluating evidence, generating strong arguments, organizing essays, and revising poor writing. Students who use GenAI to complete assignments may be able to fulfill course requirements and graduate with a history degree, but they will have done so while exempting themselves from the acquisition of both the knowledge of historical content and the research, analysis and writing skills central to the practice of history.

The massive data centres used in GenAI require enormous quantities of water, air conditioning, and energy. The shocking environmental cost of this technology runs counter to Concordia University’s commitment to sustainability. In addition, GenAI companies rely heavily on exploitative labour practices, using poorly paid human “trainers” in the “global south” while promoting their product as a replacement for fairly compensated human labour everywhere. Moreover, GenAI has run roughshod over the intellectual and artistic rights of the producers of the material in the text and image libraries upon which it depends. Finally, the intrusion of GenAI bots and gadgets into basic software makes its users unwitting labourers and content-providers for the big AI companies.

The inaccuracies and built-in biases of GenAI mean its productions are deeply flawed. Its errors include “hallucinations” of invented facts, quotations, and sources, and these are often undetectable to users without expertise in a given subject. The source bases upon which GenAI depends mean that it often operates as a cultural echo chamber, reiterating the racial, gendered, and political biases of the sources upon which it draws and from the training parameters established at its creation. Thirdly, GenAI analysis flattens the interpretative landscape of the subjects it treats, often converging in the generation of bland, median conclusions rather than producing critical evaluations based on either the preponderance of reliable evidence or on the obvious inaccuracies of weak evidence.