class: center, middle, inverse, title-slide .title[ # Unpacking polarization: Antagonism and Alignment in Signed Networks of Online Interaction ] .author[ ### David Garcia, University of Konstanz <br><br> with Emma Fraxanet, Max Pellert,<br> Simon Schweighofer & Vicenç Gómez ] .date[ ### 13.09.2023, MMM ETHZ Workshop <br> Slides vailable at: dgarcia.eu/UnpackingPolarization ] --- layout: true <div class="my-footer"><span>David Garcia - Unpacking Polarization</span></div> --- # Mitigating Online Polarization  <font size="5"> - Changes in feed algorithms have weak effects if applied only to part of the population - Alternative: recommend content with cross-partisan appeal - It can be contentious but not necessarily across the polarization fault line - What and when? Finding this content requires new models and methods [Influence of Facebook algorithms on political polarization tested. David Garcia. Nature (2023)](https://rdcu.be/djT2c) [Breaking the Social Media Prism: How to Make Our Platforms Less Polarizing. Chris Bail (2021)](https://press.princeton.edu/books/hardcover/9780691203423/breaking-the-social-media-prism) </font> --- # Unpacking (Affective) Polarization .center[] --- # Data on Signed Online Discussions .center[] Birdwatch (now Community Notes) and DerStandard comments (Austrian news) They contain political discussions with explicit signed, timestamped interactions --- ## From Interactions to Relations to Polarization  --- # Edge Frustration and Signed Alignment .center[] <font size="5"> [Balance and frustration in signed networks. S. Aref and M. Wilson. Complex Networks (2019)](https://doi.org/10.1093/comnet/cny015) </font> --- # Alignment versus Antagonism .center[] --- # Unpacking Polarization Pipeline .center[] --- .center[] --- ## Temporal Evolution of Birdwatch .center[] --- ## Peaks and events in Birdwatch .center[] --- ## Antagonism and Alignment in Der Standard .center[] --- # Effects Over Time in Der Standard .center[] --- # Elections and Alignment in Der Standard .center[] --- # Summary - Aim: Finding topics and times where Alignment is low but Antagonism is high - Developed a method to unpack Antagonism and Alignment based on frustrated edges (and a maximizing partition of the network) - Cohesiveness and Divisiveness components of Alignment - Needs new data: Birdwatch and DerStandard - A first application is aligned with previous observations in Birdwatch, and highlights contientious topics - Decreases in alignment after elections as a validity test [Unpacking polarization: Antagonism and Alignment in Signed Networks of Online Interaction. Emma Fraxanet, Max Pellert, Simon Schweighofer, Vicenç Gómez, David Garcia. Arxiv preprint (2023)](https://arxiv.org/abs/2307.06571) .center[**More at: [www.dgarcia.eu](https://dgarcia.eu) and [@dgarcia_eu](https://twitter.com/dgarcia_eu)** **Thanks for listening!**]