APS
2026 APS Annual Convention
Dimensions of Bias In Language Models: Linguistic, Cognitive, and Social
Biases permeate society, and recent evidence suggests language models also reflect such patterns. This symposium considers implicit, linguistic and cognitive biases, how such patterns are captured by current language models, what language models can tell us about human behaviour, and the challenges these models pose for researchers and society.
Chairs & Discussants
- Dermot LynottChair
Maynooth University - Kevin KiyCoChair
Maynooth University
Presentations
- When It Comes to Implicit Bias, Are All Language Models Created Equal?Dermot Lynott, Alexander Porshnev, Kevin Kiy, Diarmuid O'Donoghue, Manokamna Singh, Cai Wingfield
- Llms Displaying Less Cognitive Bias Are Not Necessarily Better Decision MakersVittoria Dentella, Marco Marelli, Luca Rinaldi
- Charting the Limits of the Inductive Bias In Large Language Models through Testing Impossible Languages Evelina Leivada, Raquel Montero, Paolo Morosi, Natalia Moskvina, Tamara Serrano, Marcel Aguilar, et al.
- Assessing the Pragmatic Knowledge of Large Language Models Under the Light of the Decontextualization BiasPaolo Morosi, Elena Pagliarini, Nikoleta Pantelidou, Evelina Leivada
