Political Communication Research
Polarization, astroturf detection and political abuse on Twitter: the Truthy project and OSoMe.
Overview
Well before platform manipulation became a mainstream concern, Bruno Gonçalves was part of the Indiana University group that built the tools to measure it. This body of work — carried out largely between 2011 and 2017 — asked how political speech actually moves through social media: who talks to whom, how sharply audiences separate into camps, and how to tell an organic groundswell apart from a handful of accounts manufacturing one.
Research Areas
Astroturf & Political Abuse
Detecting coordinated campaigns in which a few actors operate many accounts to manufacture the appearance of grassroots support, using topological, content-based and crowdsourced signals.
Polarization
Measuring how politically distinct communities do — and do not — talk to one another, by contrasting the retweet and mention networks of US political Twitter.
Inferring Alignment
Classifying the political alignment of individual users at scale, so that aggregate opinion signals are not washed out by mixing opposing camps together.
Research Infrastructure
Building open analytics platforms that give social scientists access to large-scale social media data without requiring them to write the collection pipeline themselves.
Truthy and the Detection of Astroturf
Microblogs had become grounds on which public relations, marketing and political battles were fought. The Truthy project introduced an extensible framework for tracking political memes on Twitter in real time, and paired it with a machine learning approach combining topological, content-based and crowdsourced features to identify astroturf campaigns — politically motivated actors running multiple centrally controlled accounts to simulate widespread support for a candidate or opinion.
The same group later generalized this infrastructure into OSoMe, an open analytics platform that gave researchers without a data-engineering background direct access to large-scale social media data.
Measuring Polarization
Analyzing more than 250,000 tweets from the six weeks before the 2010 US congressional midterm elections revealed that the retweet and mention networks behave very differently: the retweet network splits into two dense, homogeneous, barely connected partisan clusters, while the mention network remains a single connected community in which users encounter opposing views. A companion study of over 18,000 politically active users then traced asymmetries in how the two sides engage.
Key Publications
Everyday the Same Picture: Popularity and Content Diversity
Bessi, Zollo, Del Vicario, Scala, Caldarelli, Petroni, Gonçalves, Quattrociocchi. CompleNet (2017)
OSoMe: The IUNI Observatory on Social Media
Davis, Ciampaglia, Aiello, … Gonçalves, … Menczer. PeerJ Computer Science (2016)
Partisan Asymmetries in Online Political Activity
Conover, Gonçalves, Flammini, Menczer. EPJ Data Science (2012)
Truthy: Mapping the Spread of Astroturf in Microblog Streams
Ratkiewicz, Conover, Meiss, Gonçalves, Patil, Flammini, Menczer. WWW (2011)
Political Polarization on Twitter
Conover, Ratkiewicz, Francisco, Gonçalves, Flammini, Menczer. ICWSM (2011)
Detecting and Tracking Political Abuse in Social Media
Ratkiewicz, Conover, Meiss, Gonçalves, Flammini, Menczer. ICWSM (2011)
Predicting the Political Alignment of Twitter Users
Conover, Gonçalves, Ratkiewicz, Flammini, Menczer. IEEE SocialCom (2011)
A complete list is available on the Publications page.