Content

Among the Humanities and Social Sciences, History trains students in an unusually diverse set of skills. Central among these are how to critically engage with primary sources from the past, how to situate them in the time and place that produced them, and how this helps us understand or explain changes in human societies over time.
None of this can be fully realized unless students learn to work with primary sources in their original form, which for written sources is often in a language that students do not speak or read with native fluency. This is especially the case with Latin, a language that all students in History at Heidelberg must demonstrate proficiency in, both in its classical and its medieval variants. Medieval Latin is tested through an hour-long translation exam within the intensive introductory seminar (Proseminar) in medieval history that students must pass as a prerequisite for taking the more advanced research seminars (Hauptseminare).
This talk presents the use of privacy-conscious Large Language Models to reduce student anxiety in preparing for an exam that many perceive as particularly daunting. I explain how my Proseminar - focused on the 11th-century chronicler Adam of Bremen - integrated self-directed LLM support as part of the bwGPT pilot project and discuss student experiences, as well as the pedagogical opportunities and challenges that this approach presents.