Human Reading

Human reading research aims to understand the cognitive processes underlying text comprehension, focusing on how readers process information and how this can be modeled computationally. Current research leverages large language models (LLMs), such as transformers, and eye-tracking data to investigate fine-grained aspects of comprehension, including the prediction of reading times and comprehension accuracy from eye movements, and the role of both explicit and implicit knowledge in multi-hop question answering. These studies are improving our understanding of human reading behavior and informing the development of AI-powered tools for applications like speed reading assistance and fake news detection. Furthermore, the development of synthetic scanpath generation models is addressing the limitations of scarce eye-tracking data.