Sono Shah

Sono Shah

Political scientist · leading data science & machine learning at Theknow.io

I study how AI and digital platforms shape political opinion and behavior, and I build the systems that measure it. Ph.D. in Political Science, UC Riverside.

What I study

I'm a political scientist working in computational social science. My research combines behavioral data, like web browsing and social media activity linked to survey responses, with representative surveys and experiments to study how technology affects political attitudes and behavior.

My most recent research asked whether large language models can stand in for human survey-takers. I conceived and proposed the Pew Research Center study and co-authored it. Because we already had real answers from American Trends Panel members, we could build AI personas from those panelists and check them against what the same people actually said, on nearly 300 questions. The personas often missed, especially for Republicans and Black adults. Accuracy for the public as a whole can hide large errors for specific groups.

I've used similar data to study how Google's AI Overviews change what people click on in search results, how Democrats and Republicans behave differently on Twitter, and how negative posts from members of Congress get more engagement than their other posts. Earlier in my career I ran survey experiments on how elite rhetoric and information about political representation affect public attitudes and participation.

What I'm working on

I lead data science and machine learning at Theknow.io, a social-listening and analytics platform. I work on both the statistical methods and the production systems that deliver them to clients.

Many of the questions are ones I worked on at Pew: when one data source can substitute for another, and how to use language models while keeping their output accurate and accountable to the people who rely on it. Most of this work is internal, but feel free to ask me about it.

Before this

I was Associate Director of Data Labs at Pew Research Center, where I was the technical and methodological lead for an eight-person team. I managed four people directly, and the team published more than 20 research projects. I set the research agenda with the Managing Director, ran the team's research infrastructure, and set our standards for using LLMs to classify text, audio, and video. Bloomberg, Forbes, and Vox have covered the work.

Before Pew I spent seven years at AAPI Data, doing public opinion research on Asian American and Latino communities.

Research

Peer-reviewed work, working papers, and the public reports I authored or directed at Pew.

2026
Can AI Stand In for Human Survey-Takers? Not Really
Co-author; conceived and proposed the study, then supervised the lead researcher
Pew Research Center
2026
Investigating Click Behaviors On Google Search Result Pages That Produce an AI Overview
with Athena Chapekis, Anna Lieb, and Aaron Smith
arXiv preprint
2021
Using Administrative Records and Survey Data to Construct Samples of Tweeters and Tweets
with Stefan D. McCabe, William Hobbs, Adam Hughes, Emma Remy, and David Lazer
Public Opinion Quarterly

All research & reports →

Get in touch

Happy to hear from people working on AI and politics, measurement, or applied ML.