AI for the Historian’s Craft

In 2022, OpenAI's ChatGPT introduced the general public to large language models and their astonishing capabilities to empower humans. Since then, this and several other chatbots, such as Gemini, Claude, and Deepseek, have demonstrated the ability to keep up in many fields, including history. Some studies estimate that at least two-thirds of students, particularly in North American universities, use it regularly. Beyond the debate on plagiarism, beyond ChatGPT, we propose to organise a session on the uses and misuses in history of Artificial Intelligence (AI) as we understand it today (machine learning and deep learning), according to the main stages of historical research: sources (collection and processing), interpretation and writing. Very soon this topic has been explored in the circles of digital history, digital humanities, and beyond: Wulf Kansteiner has asked himself what AI could be for a historian (History & Theory), Andrea Nanetti and Li Boyang explored the process of text-to-image generation comparing Renaissance illuminated manuscripts and generative AI, Sarah Gensburger and Frédéric Clavert co-edited an issue on collective memory and AI (Memory Studies Review), Frédéric Clavert and Caroline Muller are looking at AI as a ‘discrete’ (undocumented) digital practice of historians (20&21), Victoria G. Walden and Mykola Makhortykh have considered the uses of AI in Holocaust studies, to give just a few examples. We feel it is important to bring the debate to the CISH congress, a landmark event in the life of international historians. To have a global impact and serve the international member associations equally, we propose to focus on methodologies rather than regional historical themes. In practice, the panel wants to invite representatives from different regions of the world to lead co-authored papers with scholars of their choice who represent best practices in their respective regions in exploring how AI can empower the historian's craft. For a comprehensive and transparent discussion panel, contributions will be encouraged to work on the following four themes. The first theme focuses on the already common uses of AI in history (the use of Transkribus, for example, which is based on machine learning) or in digital history (topic modelling, word vectors). The second theme focuses on uses that are not well documented and are likely to develop: generative AI (ChatGPT, LLAMA, Google Gemini, or Mistral to give just a few examples) but also the growing use of AI in archive centres. The third theme attempts to show what history can contribute to AI research, including a critical evaluation of its results, its methods and training datasets, and the biases they introduce. The fourth one is the impact of AI on history education, both in terms of practical solutions and potential impact (positive and negative) on historical culture of today’s students and future adults.