AI for the humanities and the humanities for AI

Beáta Megyesi, Professor of Computational Linguistics at Stockholm University

Abstract

Many texts remain only partially understood, encoded or written in undeciphered scripts and languages. Addressing such materials challenges traditional linguistic approaches and calls for closer collaboration between the humanities and artificial intelligence. This talk examines how AI can support humanistic research—and how humanistic perspectives can shape AI—through computational methods for analyzing hidden codes and unknown languages. Drawing on the DESCRYPT project, I show how hybrid approaches that combine neural models with linguistic and philological expertise can work effectively with sparse and uncertain data, while also highlighting the limits of AI and the continuing role of human interpretation.

Beáta Megyesi is Professor of Computational Linguistics at Stockholm University. She earned her Ph.D. from the Royal Institute of Technology (KTH) and served as associate and full professor at Uppsala University. She has led numerous externally funded research initiatives and is Principal Investigator of DECRYPT (Swedish Research Council, 2018–2024) and DESCRYPT (Riksbankens Jubileumsfond, 2025–2032), projects that advance computational methods and research infrastructures for historical cryptology and the automatic analysis of rare scripts. 

Megyesi is the author of more than 120 scientific publications and has delivered many keynote lectures at international conferences. Her previous leadership roles include President of the Northern European Association for Language Technology, Chair of the Linguistics Review Panel at the Swedish Research Council, and Head of the Department of Linguistics and Philology at Uppsala University.

Olle Häggström, Professor of Mathematical Statistics at Chalmers University of Technology. 

About the seminar

Olle will discuss the situation regarding the threats posed by AI and what can be done about them. 

His lecture will briefly cover this:  AI is developing at breakneck speed, and the CEOs of the leading AI companies expect within a single-digit number of years to have AI so transformative that at a minimum it will thoroughly disrupt the labor market and other parts of society. 
”It might even wipe out humanity altogether. Yet, these same CEOs claim to be locked in a race against each other that prevent them from pulling the breaks. This is not OK. I will explain the situation in more detail, and briefly address what can be done about it”, says Olle Häggström.

Olle Häggström is professor of mathematical statistics at Chalmers University of Technology and a member of the Royal Swedish Academy of Sciences. 
His research is in probability theory and statistical mechanics, but in recent years his interests have shifted towards futurology, global risk and AI safety. He is the author of five books.
 

Evie Coussé is a linguist at University of Gothenburg

Abstract

In recent decades, linguistics has increasingly relied on corpora to study language. As more texts have become digitally available, corpora have grown dramatically in size, and datasets containing billions of words are now common. Such big data allow linguists to investigate infrequent phenomena that only emerge in very large datasets and to study multiple phenomena simultaneously, opening new perspectives on how the language system works as a whole. However, this abundance of data also creates methodological challenges: annotating and analyzing datasets of this scale quickly moves beyond the reach of human annotators. Artificial intelligence has therefore been explored as a solution. In this presentation, I illustrate the development of corpus linguistics towards big data and AI with examples from my research on Dutch and Swedish. I also take a step back to consider whether the move toward ever larger datasets is always necessary—or even desirable—in linguistic research.

Evie Coussé is a linguist specializing in the Germanic languages. She obtained her PhD in Linguistics from Ghent University (Belgium) in 2008. Since 2010, she has been employed at the University of Gothenburg, where she became Associate Professor of Linguistics in 2014. Her research focuses on grammatical change in the Germanic languages—especially Dutch and Swedish—including word order change and the development of auxiliary verbs (grammaticalization). She studies these processes across a wide range of corpora, from medieval Bible translations to contemporary social media.

Anton Törnberg professor at the Department of Sociology and Work Science, University of Gothenburg

About the seminar

A recent policy report underscores a critical shift: climate misinformation is no longer confined to the digital fringes but increasingly originates from within the political mainstream. Political elites – especially in fossil-fuel aligned, populist, or authoritarian regimes – play a central role in amplifying and legitimizing misleading climate narratives. 

This presentation explores how AI in the form of Large Language Models (LLMs) can be used to study how such elites contribute to the global circulation of climate misinformation. Drawing on a global dataset of approximately 26 million tweets posted over five years by politicians from all major parties in 26 countries, the presentation explores which types of political actors spread climate misinformation; how this varies across regime types, regions, and ideological orientations; and what kinds of narratives elites deploy to delay climate action. Combining LLMs with qualitative human coding, the presentation uses a comparative framework mapping patterns of elite misinformation across democratic, hybrid, and authoritarian contexts, and traces temporal surges around key political and environmental events such as COP meetings, spikes of climate movement mobilization, and extreme weather crises.

Anton Törnberg is an associate professor in the Department of Sociology and Work Science at the University of Gothenburg. His research focuses primarily on the online far right, combining computational methods with qualitative approaches. He is the author of Intimate Communities of Hate: Why Social Media Fuels Far-Right Extremism (Routledge, 2024).