Semra Sevi @semrasevi
Assistant Professor of Political Science @UofT semrasevi.com Joined September 2016-
Tweets581
-
Followers7K
-
Following922
-
Likes8K
Introducing ReproAI, an author-facing agentic AI plugin for replication packages: reproai.org Since we released the Scaling Reproducibility paper, many people have asked how they can make use of the pipeline. Recall that, even after in-house reproducibility checks by journals, we still find many human errors in replication packages. Gary @kinggary gave us a truly inspiring direction: "With AI tools, the relationship among authors, journals, and replicators can become much more collaborative." He suggested that we build a tool to help authors prepare replication packages before submission. So we did, and it is fully open source. Building on our experience replicating >500 papers, we summarized the most common problems we encountered. Our goal is simple: help authors increase the chance that their replication packages can be run more easily by data editors, reviewers, and future researchers. This task is actually much easier than third-party replication from scratch because authors know their papers best😺: they have the source files, understand the intended workflow, and can often fix bugs quickly once problems are identified. Here is the first version. Let me show it using one of my old papers. 🧵
Why do arguments often change people’s beliefs without changing their attitudes? In a new APSR with @patrickpliu and Scott Clifford, we point to belief relevance: arguments are more persuasive when they grapple with the idiosyncratic reasons people hold their political views.
NEW in @ScienceAdvances, after 3 years of work with a great team: We review and meta-analyze 100 immigrant conjoint experiments in 36 countries. Immigration preferences are surprisingly similar across people and countries, but changing over time and structured by politics. 🧵
A monumental loss. Ran was not only a trailblazing and brilliant scholar; he was also a remarkably generous colleague and, what's more, a dear friend. politics.utoronto.ca/news/memoriam-…
Field experiment presents causal evidence that Social media promotion improves job market outcomes | PNAS pnas.org/doi/10.1073/pn…
So excited to finally share my job market paper! The post-Floyd “Great Awokening” was driven by affluent white liberals and emphasized recognition over redistribution. Consistent with elite capture of identity politics. Feedback very welcome! Link below.
@RoyceKoop This piece might also be of interest to you. cambridge.org/core/journals/…
Excited to present my JMP at MPSA! The post-Floyd “Great Awokening” was driven by affluent white liberals and emphasized recognition over redistribution. Evidence from surveys, public discourse, and implicit bias data. Consistent with elite capture of identity politics. ⬇️
It turns out people vote based on policy. A sneak peek at our paper for MPSA next week, in a not-to-be-missed panel: convention2.allacademic.com/one/mpsa/mpsa2…
Turns out most Canadians vote the party, not the person. Local candidates were decisive for just 5-8% of voters outside Quebec and 2-5% in Quebec. The data don't lie! /w @marcomavina & @ablais_ See paper here: cambridge.org/core/journals/…
“bUt In CaNaDa YoU vOtE fOr ThE pErSoN nOt ThE pArTy” 🤡
My work with @marcomavina & @ablais_ shows that most Canadians vote the party, not the person! Local candidates were a decisive factor for 5-8% of voters outside Quebec, and just 2-5% in Quebec. See: cambridge.org/core/journals/…
Regardless of what you think, you’re voting for a person in Canada, not a party. There’s a reason why the name is the one in bolder font and not the party!
@Nature has published three groundbreaking papers on reproducibility, analytical robustness, and replicability across the social sciences. Sincere thanks are due to the many folks who contributed to these projects. It’s painstaking work, and a great service to social science.
Our PNAS paper (@semrasevi + Don Green) is a small step toward building this information layer, but as RAG gives way to massive context windows and agentic retrieval, the central challenge becomes curating and prioritizing high-quality information.
Amidst understandable concerns of AI dystopia, no one is offering a positive vision for how we can use AI to remake our institutions and reinvent how we govern. That’s what I try to offer today. My argument is that we need an explicit research agenda to build “political
Great to welcome @UofT Professor Sevi’s 3rd-year political science class to #QueensPark this week! From the gallery during Question Period to a conversation in the council chamber with Premier @fordnation, they got a front row look at how provincial politics works. It was clear from their insightful questions that these students are receiving a world-class education that will set them up for success. I look forward to seeing all they will accomplish in their future careers!
New paper w/ @YamilRVelez! A lot of great research on political microtargeting discounts personalization: tailored ads (using AI or not) rarely beat a single-best message. We define two types of microtargeting, clarify when tailoring matters, & showcase a novel audio-based design
1/ Will voters participate in the primary of a party they oppose to prevent the nomination of a candidate they fear? In a new paper in AJPS with @HayleyCohen, we study crossover voting using surveys and a large field experiment (N=83,902) in the 2024 New Hampshire Republican presidential primary. onlinelibrary.wiley.com/doi/10.1111/aj…
@markovitis @j_kalla @HayleyCohen Very cool!!
Conditionally accepted at the APSR (w/ @ScottClifford & @patrickpliu): Why does political information so often change beliefs but NOT attitudes? We highlight the role of belief relevance, or the extent to which beliefs bear on attitudes.
Trends for United States
You might like
