All Events
Bi-Co Mathematics Colloquium with Dr. Phil Chodrow
Nov
2
2026
4:15pm - 5:15pm
On Campus Event - Park Science, Room 245
"Learning Social Influence"
Abstract:
Information environments help shape the beliefs of individuals and populations. The question of how they do this--how social influence operates--has been a long-running concern of social psychologists and, more recently, mathematical modelers. In both experiments and models, researchers often assume that individuals are always susceptible to influence from every potentially salient element of their information environment simultaneously. The implied narrative is one in which a user opens Twitter and is immediately influenced by the opinions of every user they follow. Of course, the user is actually exposed only to a subset of recent posts, and may not be susceptible to influence by all of them.
In this talk, I'll briefly survey some standard mathematical models of social influence with an eye towards their underlying mechanical assumptions. I'll then introduce a new model of social influence in which we explicitly represent an agent's selective attention to their information environment. Since we don't usually get to directly observe attention in most experiments and data sets, we model this as a latent mechanism and develop an expectation-maximization algorithm for statistical inference. The incorporation of selective attention turns out to be a very impactful choice: incorporating selective attention can qualitatively alter the outcome of model selection on both synthetic real data. It also turns out to be a helpful choice in the context of statistical prediction. Models with latent mechanisms consistently outperform standard methods for fitting social influence functions on a wide range of empirical data sets ranging from controlled psychological studies, congressional political alignments, Reddit arguments, and prediction markets. We hope that this work contributes to a reexamination of some common assumptions in mathematical models and to the development of new directions in quantitative research on social influence.
Audience: BMC Community
Type(s): Seminar/Colloquium
Submitted by:
Contact:
Selvi Kara
Bryn Mawr College welcomes the full participation of all individuals in all aspects of campus life. Should you wish to request a disability-related accommodation for this event, please contact the event sponsor/coordinator. Requests should be made as early as possible.