Colloquia

Recent years have seen a surge of efforts to adapt machine learning techniques for healthcare.These tools have provoked heated debates about privacy, safety, bias, and inequality, but lawsand official guidance lag behind technological advances. This talk investigates how healthresearchers develop rules and norms around the use of data-intensive technologies in the absenceof formal regulation, and how these new ideas are poised to change healthcare for clinicians andpatients alike. Drawing on three years of ethnographic research and interviews, I investigate thistransition within digital psychiatry, a field that uses machine learning and other data-intensivetechniques to study mental illness and provide mental healthcare. I analyze how clinician-researchers settle norms in digital psychiatry as they develop data values, moral sentimentsaround digital data’s objectivity, authoritativeness, impartiality, and scalability. I argue datavalues have substantive implications for mental health work and care. While psychiatry hashistorically emphasized clinical judgment, digital psychiatry hybridizes expertise in psychiatry as

Gender/sex is in AI. Sociologists and tech critics alike have written extensively about the consequences of it being there. Many technologists and not a few scientists celebrate its presence. But how exactly do gender, sex, and their ambiguous conflations enter into ‘AI’? This talk develops an account of analytically distinct sociotechnical processes by which social facts such as gender/sex become embedded in computer systems, with a focus on contemporary machine learning. As I will show, individual agency, social (infra)structure, and the logic of learning algorithms themselves all have roles to play.

Details to be dsiplayed soon.

Abstract:

"Panel and Open Discussion on the 2024 Election in Sociological Perspective."    Moderator: Kim Voss   Panelists: Cristina Mora, Cihan Tugal, Dylan Riley, and Michael Rodriguez-Muniz.   

   

Presented by the UC Berkeley Computational Research for Equity in the Legal System Training Program (CRELS)

Individuals do not respond uniformly to treatments, such as events or interventions. Sociologists routinely partition samples into subgroups to explore how the effects of treatments vary by selected covariates, such as race and gender, based on theoretical priors.  Emerging machine learning methods based on decision trees allow researchers to explore sources of variation that they may not have previously considered. In this talk, Brand describes a range of approaches to study effect heterogeneity, including tree-based machine learning.

Abstact: