[MetaSTEM] Next week: Aditya Dasgupta on AI for Social Science Discovery (NEW LOCATION)

[Tyler Marghetis]

See everybody tomorrow for a talk from Adi Dasgupta!

Recall that we’ll squeeze into a new space: *SSM 269*, on the *exact
opposite corner of SSM from our usual room*.

Yours in MetaSTEM,
Tyler

On Thu, Mar 12, 2026 at 3:21 PM Tyler Marghetis <tyler.marghetis at gmail.com>
wrote:

> Dear MetaSTEMmers,
>
> Next week we're joined by Adi Dasgupta (CCed) from PoliSci.
>
> We'll meet at our usual time and place: Thursday March 19, 12:30-1:30pm.
>
> Please note that we'll assemble in a *new location*: my lab, which is
> located in *SSM 269A*, on the *exact opposite corner of SSM from our
> usual room*.
>
> Thanks again to Christian for the boxed lunches! And a big thanks to
> Aditya for sharing his work.
>
> Best,
> Tyler
>
> *Title:* AI for Social Science Discovery
> *Authors:* Aditya Dasgupta, Tesalia Rizzo, Yuhang Zhu
> *Time & Place: *Thursday, March 19, 12:30-1:30, in *SSM 269A*
> *Abstract:* Traditionally, social scientists understand human behavior by
> choosing and measuring explanatory variables, and conducting statistical
> tests of hypothesized relationships with a target behavior of interest. Yet
> individuals make decisions in complex, high-dimensional environments, the
> relevant features of which may be difficult to identify/theorize ex ante.
> We propose a framework to use artificial intelligence algorithms for text
> and imagery to assist in the discovery of determinants of human behavior.
> Our framework is based on training a deep learning algorithm end-to-end to
> predict a target behavior of interest directly from unstructured data such
> as imagery and text that provides a naturalistic representation of the
> setting in which the behavior takes place. Instead of manually selecting
> explanatory variables ex ante, we propose usage of explainable artificial
> intelligence tools to discover interpretable features that drive
> predictions ex post. We provide an information-theoretic grounding for the
> framework and illustrate how it can facilitate discovery of novel
> determinants of behavior in diverse settings, including understanding what
> content gets censored by Chinese social media monitors, how the built
> environment affects the turnout decisions of Mexican voters, and how visual
> biases enter the criminal sentencing decisions of American judges.
>
> --
> Tyler Marghetis, Ph.D.
> Assistant Professor, Cognitive & Information Sciences
> University of California, Merced
> www.tylermarghetis.com
> tyler.marghetis at gmail.com | +1.619.252.7798
> I use the pronouns: he, him, his
> I have moderate face-blindness
> <https://en.wikipedia.org/wiki/Prosopagnosia>; if you see me out of
> context, I may not recognize you
>
-- 

Sent from a tiny device...

-- 
Tyler Marghetis, Ph.D.
Assistant Professor, Cognitive & Information Sciences
University of California, Merced
www.tylermarghetis.com
tyler.marghetis at gmail.com | +1.619.252.7798
I use the pronouns: he, him, his
I have moderate face-blindness; if you see me out of context, I may not
recognize you, so please say hello.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ucmerced.edu/pipermail/metastem/attachments/20260318/063365ee/attachment.htm>