L’impatto delle tecnologie generative sulla ricerca storica: alterazioni, riscritture e previsioni

Abstract

Artificial Intelligence (AI) application to historical research is revolutionizing how scholars analyze and interpret sources. AI offers numerous advantages, such as the ability to process large datasets and perform advanced semantic searches. However, it also raises significant challenges regarding the interpretation of sources selected by AI, including biases and the machine’s “laziness.” Recent advancements in AI, machine learning, and language models (LLMs) have prompted reconsideration of the use of generative technologies in academic research. AI’s growing “humanoid” capabilities have attracted even social scientists to explore these tools. Particularly useful in public history, generative technologies expand the audience of historical sciences by implementing new learning models, like what happened in the past with Wikipedia. Initially met with skepticism by academic historians, Wikipedia is now regarded as a valuable tool for public historians. One of AI’s primary implications in historical research is Optical Character Recognition (OCR), which automatically transcribes printed and handwritten texts, facilitating understanding documents written in ancient or less-known languages or simply in unclear handwriting. A specific case will be analyzed where, due to the machine’s aforementioned laziness, it chose to focus on the first part of the provided text, attempting to predict the content of subsequent slides. By analyzing available data, AI can “foresee” future developments or simulate scenarios, often creating historical distortions and rewriting reality.