Cognition - Multiple sclerosis (MS)
Multiple sclerosis (MS) is a chronic autoimmune disease of the central nervous system characterised by inflammatory demyelinating lesions in both grey and white matter, leading to secondary neurodegeneration. 30–70% of people with MS experience cognitive impairment, which can significantly impact daily functioning and well-being. The most commonly affected domains are processing speed and memory issues, followed by attention, executive functioning, verbal fluency, and visuospatial abilities. Cognitive symptoms vary widely in severity and progression, but their early occurrence is associated with worse disease outcomes. Early detection and continuous monitoring are therefore essential to ensure timely intervention and prevent long‑term cognitive disability.
Predicting cognitives trajectories remains, however, a clinical challenge. Standard cognitive screening relies on in‑clinic neuropsychological testing, which is costly, infrequent, and captures cognition only in artificial settings. Available digital assessments requiring active participation also face adherence challenges, underscoring the need for tools that better fit patients’ daily lives.
Our research focuses on developing sensitive, real‑world methods to screen and monitor cognitive decline. Specifically, we are investigating digital markers such as naturalistic smartphone interaction patterns (e.g., app usage, typing dynamics) and acoustic features of speech (e.g., pause duration, loudness). These digital traces may serve as proxies for cognitive health, enabling high‑frequency monitoring in everyday environments. Combined with clinical markers such as MRI metrics, they may offer a more comprehensive picture of how cognitive symptoms manifest in MS, supporting more tailored interventions.
Cognition and use of Large Language Models (LLMs)
Large language models (LLMs) are rapidly becoming embedded in everyday cognitive activity, from writing and studying to problem solving and decision making. While their societal impact is widely discussed, their effects on the human brain remain largely unexplored. New evidence suggests a striking paradox: LLMs can improve short-term task performance while simultaneously reducing cognitive effort and neural engagement during memory processing.
This project investigates how interaction with LLMs translates into changes in brain activity, combining behavioral measures with EEG recordings to capture neural responses during AI-assisted cognition.
At the core of this question lies the hippocampus, a key structure supporting memory encoding and long-term consolidation. Effective memory formation depends on active engagement of hippocampal-cortical networks, particularly during effortful encoding and retrieval. When cognitive processes are partially outsourced to AI, this engagement may fundamentally change, with potential consequences for brain plasticity and long-term memory formation.
Using behavioral testing combined with EEG and intracranial recordings, we study both healthy individuals and people with epilepsy, a population characterized by increased vulnerability of memory networks.
By integrating cognitive and neurophysiological data, this research seeks to determine whether LLMs function as supportive cognitive tools or whether they may contribute to reduced neural engagement and long-term memory formation. The findings will provide fundamental insights into how artificial intelligence interacts with human cognition and inform responsible use of AI in clinical and educational contexts.
