TEXT will examine LLM-hard problems in terms of the linguistic performance of LLMs, and implications for linguistics, cognitive science, literacy, and media analytics 

This work package explores how large language models (LLMs) contribute to our understanding of human language processing, linguistic abilities, and the broader implications for literacy, cognitive science, and media analysis. It addresses key challenges, such as analyzing how LLMs handle resource-scarce languages, tracking linguistic changes over time, distinguishing between human and AI-generated texts, and assessing the impact of AI-generated content on public discourse. 

A central focus is on the nature of human language acquisition and linguistic competence. The ability of modern LLMs to generate coherent text has surprised experts, reigniting discussions on whether human language learning is primarily driven by statistical patterns. This project examines how LLMs process language under conditions that mirror human learning, including psycholinguistic experiments that test their ability to generalize linguistic structures. 

Another key aspect is investigating how LLMs perform across a diverse range of languages, particularly those with limited digital resources. By studying LLM performance on under-resourced languages, the project explores whether certain linguistic structures pose particular challenges for AI models. This research also assesses whether knowledge gained from high-resource languages can be effectively transferred to low-resource ones. 

The project further examines the ability to differentiate between human and AI-generated texts, developing scalable and interpretable models for detecting machine-written content. Understanding these distinctions will shed light on how AI influences writing culture and public communication. 

Finally, the work package develops new methods for analyzing the role of AI-generated text in literacy and online engagement. By combining linguistic research with computational tools, this research will provide insights into how people interact with AI-generated content and its broader societal impact.