Our Opinion Score

The algorithm behind Horizon's core feature, the Opinion Score.

Expression

Is an article expressed as mostly fact statements, analysis statements, opinion statements, or very highly opinionated statements. Those that are expressed as mostly opinion statements or very highly opinionated statements should be considered inherently less reliable than those with mostly fact and analysis statements.

  • Use of opinion-laden words (e.g., "amazing," "terrible," "undoubtedly")
  • Language that speculates or assumes: "probably," "likely," "seems"

Language

A first type of linguistic indicator is the words used to describe political positions. Notice if the article uses words normally associated with one side or another, such as “pro-choice” vs “pro-life” or “illegal aliens” vs “undocumented immigrants.” A second type of linguistic indicator is the words used to characterize one’s political allies and opponents. For example, calling a politician smart or savvy may indicate mild positive bias toward that politician’s party, while calling a politician heroic or brilliant may indicate a stronger positive bias. Conversely, calling a politician crafty or sneaky may indicate a mild negative bias, while calling that same politician stupid or senile would indicate an extreme negative bias.

Headline and Graphic

Does the headline and graphic align with the sentiment of the article content itself, and how opinionated is the headline? Is the research sponsored?