On July 29th, the CSSLab held its inaugural Student Research Conference to share the summer progress made in the Lab’s various projects.
The Summer 2022 Student Research Conference created a space where student members of the CSSLab presented on their summer project work, something learned (about themselves, about the research world, about social science or computer science, about a methodological approach) or something built (a technology, a piece of code, a mapping process, a technique, a lit review, an idea for further exploration). Covering six unique projects, the Conference helped to connect and explain active research across the Lab.
Read on for descriptions of each project, along with links to the corresponding Conference presentations:
Part of the Lab’s work on High-Throughput Experiments on Group Dynamics, the Team Performance project aims to analyze and learn how teams collaborate, along with which factors impact their success. In the process, it is working to transform social science from the “one at a time” paradigm into a replicable, scalable, and generalizable process.
To help develop the data infrastructure for research in the Penn Media Accountability Project (PennMAP), we investigate stacking topic models as a form of unsupervised machine learning. This project tackles two key flaws: automated evaluations do not agree with human evaluations, and automated evaluations overestimate performance of neural topic models.
Another project under the High-Throughput umbrella, Nudge Cartography is building an interactive “nudge map” that policymakers, marketers, managers, and academics seeking to motivate behavior change can use to search for candidate nudges. The goal is to make searching for appropriate nudges as easy as searching for a flight on Kayak.
The Lab’s in-progress Living Journal initiative is creating functional, public-facing resources that update in real time, subverting the old paradigm of static research and replacing it with “living papers.” The first prototype of these living papers was developed to present research on YouTube radicalization as part of PennMAP.
The COVID-19 pandemic has made apparent the need for accurate models for epidemics, and the proliferation of GPS mobility data has made studying human mobility patterns easier than ever. As part of the Lab’s COVID-Philadelphia research, this project explores the potential of collaborative filtering (CF) — a machine learning algorithm that can learn relationships between two domains.
Part of the High-Throughput series, the Deliberation project researches which interventions most reduce social polarization in deliberative groups and populations. Its short-run goal is to design and build the core infrastructure for high-throughput experiments that will power an open science platform, allowing researchers to “map” the space of possible deliberation scenarios.