Embedded Files
Research Vision
Research Vision
The NLP for Human Sciences Lab is dedicated to redefining the interface between advanced computational linguistics and the human-centered sciences. We operate as an interdisciplinary hub, driven by the dual mission of providing a scientific accelerator for diverse fields - from medicine to the humanities - while simultaneously pioneering new methodologies that leverage the internal logic of AI to model human complexity.
Our lab thrives on an open, collaborative spirit, welcoming partnerships that challenge the boundaries of how we analyze cognition, culture, and clinical interaction. We believe that by treating Large Language Models as both powerful tools and testable scientific environments, we can unlock insights into the human experience that were previously beyond reach.
Research Lines
Research Lines
1. Computational Clinical Psychology
1. Computational Clinical Psychology
This line focuses on the transformation of clinical interaction into quantitative scientific data. We apply state-of-the-art NLP to decode the "DNA" of the therapeutic process and understand the longitudinal trajectories of mental well-being.
The Horizon: Mapping the mechanisms of psychological change by analyzing therapist-patient dynamics, identifying digital biomarkers of mental states in social media, and modeling the impact of specific clinical interventions.
The Goal: To provide clinicians and researchers with a data-driven "map" of human interaction, moving from subjective observation to empirical discovery.
II. NLP as a Scientific Accelerator
Beyond specialized clinical work, we serve as a high-performance engine for the broader scientific community. This pillar represents our commitment to scaling the classical scientific method across diverse and complex data landscapes.
The Horizon: Bridging the gap between raw textual data and scientific insight in fields ranging from Medical and Life Sciences to the Digital Humanities (e.g., historical and spiritual manuscripts). We provide the computational machinery for high-fidelity automated annotation and large-scale synthesis.
The Goal: To accelerate the pace of discovery across disciplines, allowing researchers to navigate and analyze massive corpora with a depth and speed previously unreachable.
III. Mechanistic Interpretability & the In-Silico Laboratory
In this frontier, we move from using AI to analyzing its core. We treat Large Language Models as "white-box" environments where we can explore the internal geometry of meaning and simulate human-focused experiments.
The Horizon: Identifying the conceptual dimensions of latent representations (e.g., via low-rank subspace) that correspond to high-level human constructs and critical scientific variables. By perturbing these internal directions, we use models as specialized laboratories to ask "What if?" questions about intra-personal and inter-personal dynamics.
The Goal: To upgrade scientific methodology itself — moving from passive observation of text to active, causal simulations within the learned structure of language models.
Page updated
Google Sites
Report abuse