People don’t always read past the headlines – but headlines don’t tell the whole story. Here we introduce the first feature in AOI’s Lucid Lens project: a proof-of-concept tool that automatically rewrites news headlines to reflect the content of articles in a more accurate, less sensational way.
On Nov 26, 2021, the New York Times published an article with the headline: “After Murders ‘Doubled Overnight,’ the N.Y.P.D. is Solving Fewer Cases.” A reader could be forgiven for concluding that murders in New York City had doubled overnight. In fact this was not the case, and was indeed never claimed in the content of the article. A correction later noted that: “detectives’ caseloads doubled overnight, not murders“ [emphasis added]. This came alongside complaints on social media from discerning readers. The editorial move to imply murders had doubled in America’s largest city turned a complicated story into a simple, sensational, terrifying one – that is, if you only read the headline.
People often share articles after reading only the headline.1 What drives sharing is largely emotional content, especially negative emotion. As the journalistic saying goes, “When it bleeds, it leads.” This incentivizes users to share content that they think will be engaged with. Journalists may internalize these incentives (even without knowing it) and may craft headlines that are more sensational, misleading, or essentially “clickbait.” Headlines are powerful enough to shape readers’ understanding of articles even when they read the articles.