Project

The Electome: Measuring Responsiveness in the 2016 Election

Groups

The Electome: Where AI Meets Political Journalism

The Electome project is a machine-driven mapping and analysis of public sphere content and conversation associated with the 2016 presidential election campaign. Through its unprecedented view of the national election conversation, LSM aims to shift some of our collective focus from who’s winning/losing (traditional “horse race” polls and projections) to the issues the campaign is being fought over (the “Horse Race of Ideas”). The Electome is fueled by two primary data streams: the entire Twitter archive and daily output (the so-called 500m Tweet per day “fire hose”) as well as a sample of daily content from 30 digital news sites (5k-6k stories per day). A series of machine learning algorithms identify those Tweets and stories specifically about the election, then classify them by topic, candidate, organization and a number of other filters. The classified data is then run through various semantic and network analytics that continuously measure and visualize:

The Electome: Where AI Meets Political Journalism

The Electome project is a machine-driven mapping and analysis of public sphere content and conversation associated with the 2016 presidential election campaign. Through its unprecedented view of the national election conversation, LSM aims to shift some of our collective focus from who’s winning/losing (traditional “horse race” polls and projections) to the issues the campaign is being fought over (the “Horse Race of Ideas”). The Electome is fueled by two primary data streams: the entire Twitter archive and daily output (the so-called 500m Tweet per day “fire hose”) as well as a sample of daily content from 30 digital news sites (5k-6k stories per day). A series of machine learning algorithms identify those Tweets and stories specifically about the election, then classify them by topic, candidate, organization and a number of other filters. The classified data is then run through various semantic and network analytics that continuously measure and visualize:

  • the share of conversation or coverage that any given issue or candidate commands on Twitter and in the news media, respectively—and how the two platforms are aligned
  • which issues are most closely associated with each candidate on Twitter (via co-occurrent candidate/issue references in single Tweets)
  • how much of the public sphere conversation and coverage is about substantive issues as compared to politics (polls, projections, process) and the candidates’ character and personality
  • specific sub-topics and representative Tweets within broader conversations about specific issues or candidates
  • the level of “incivility” (profanity, insults, violence, ethnic/sexual slurs) within the public Twitter conversation about any given issue or candidate
  • who is influencing the public sphere election conversation (via a composite Twitter/media influence metrics)

LSM’s deployment of Electome analytics has been supported by the Knight Foundation, with the goal of fueling news coverage that is more responsive to what matters most to the public. To that end, LSM has: 

  • collaborated with the Commission on Presidential Debates to offer Electome analysis to the general election debates’ moderators and credentialed journalists
  • also collaborated with the Roper Center for Public Opinion Research at Cornell University for integration of the Center’s polling in Electome analytics/dashboard
  • and built a self-service dashboard featuring several Electome analytic tools for journalists and analysts to produce their own issue-driven analyses and visualizations.

Looking beyond the 2016 election, LSM sees Electome technology as enabling new forms —and, importantly, creators—of investigative and explanatory journalism by democratizing access to powerful data mapping, analysis and visualization tools.

Copyright

LSM

Project at a glance