[Theory] [TTIC Talks] 7/29 Thesis Defense Announcement

[Brandie Jones via Theory]

*Who: *Pushkar Shukla, TTIC

*When: *Tuesday, July 29 2025 from 11:00 am -1:00 pm CT

*Location: TTIC , 6045 South Kenwood Avenue, 5th Floor Room 530*

*Virtually: via Zoom
<https://us02web.zoom.us/j/9271945247?pwd=UVJ4TmVqVUJPc0gxbkRRY3pXV2lZUT09&omn=81415756789>*

*Title:* Understanding and evaluating computer vision models through the
lens of counterfactuals

*Abstract: *As machine learning systems increasingly influence decisions in
critical areas such as healthcare, finance, and creative industries, there
is a growing need to understand how these models make decisions and whether
those decisions are fair and trustworthy. This thesis investigates computer
vision models through the framework of counterfactual reasoning —
systematically varying specific attributes while holding others fixed to
evaluate model behavior, biases, and assumptions.

Across several contributions, this work advances methods to (1) quantify
the influence of human-defined concepts on model decisions (CAVLI), (2)
mitigate bias through adversarially generated counterfactuals (ASAC), (3)
dynamically evaluate biases in text-to-image generative models (TIBET), and
(4) diagnose and mitigate intersectional biases through causal
interventions (BiasConnect and InterMit). Together, these contributions
demonstrate how counterfactual reasoning provides a unified, interpretable,
and practically useful toolkit for auditing and improving computer vision
models.

The defense will present a detailed exploration of these methods,
experimental results across classification and generative tasks, and
broader reflections on fairness, explainability, and the future of
trustworthy AI in vision systems.

*Committee:*
Prof. Prof Matthew Turk (Chair), TTIC
Prof. Prof Matthew Walter , TTIC
Prof. Professor Leonid Sigal, UBC


-- 
*Brandie Jones *
*Executive **Administrative Assistant*
Toyota Technological Institute
6045 S. Kenwood Avenue
Chicago, IL  60637
www.ttic.edu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/theory/attachments/20250721/800fbf75/attachment.html>