BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Towards Causally Reliable Concept-based Models    - Giovanni De Fe
 lice (Università della Svizzera Italiana)\, Arianna Casanova Flores (Univ
 ersity of Liechtenstein)\, and Francesco De Santis (Politecnico di Torino)
DTSTART:20260212T110000Z
DTEND:20260212T120000Z
UID:TALK244321@talks.cam.ac.uk
CONTACT:Mateja Jamnik
DESCRIPTION:Concept-based models are an emerging paradigm in deep learning
  that constrains the inference process to operate through human-understand
 able variables\, facilitating interpretability and human interaction. Howe
 ver\, these architectures\, on par with popular opaque neural models\, fai
 l to account for the true causal mechanisms underlying the target phenomen
 a represented in the data. This hampers their ability to support causal re
 asoning tasks\, limits out-of-distribution generalization\, and hinders th
 e implementation of fairness constraints. To overcome these issues\, we pr
 esent Causally reliable Concept Bottleneck Models (C2BMs)\, a class of con
 cept-based architectures that enforce reasoning through a bottleneck of co
 ncepts structured according to a model of the real-world causal mechanisms
 . Finally\, in light of the limitations of such approach\, we anticipate o
 ur intended future directions while also discussing their relevance within
  the broader landscape of concept-based interpretability.\n\nArianna Casan
 ova\, Francesco De Santis\, and Giovanni De Felice are academic researcher
 s with a shared focus on concept-based interpretability. Arianna Casanova 
 Flores is a postdoctoral researcher at the University of Liechtenstein. Pr
 eviously\, she received a PhD in Informatics at the Dalle Molle Institute 
 for Artificial Intelligence (CH). Francesco De Santis is a PhD candidate a
 t Politecnico di Torino (IT). He received a Master’s degree in Data Scie
 nce from Politecnico di Torino. Giovanni De Felice is a postdoctoral resea
 rcher at Università della Svizzera italiana (CH). Previously\, he complet
 ed his Ph.D. (2025) in the Data Mining & Machine Learning group at the Uni
 versity of Liverpool (UK). He is a co-creator of the open-source machine l
 earning library PyTorch Concepts.
LOCATION:Computer Laboratory\, William Gates Building\, Lecture Theatre 1
END:VEVENT
END:VCALENDAR