Curiosity in news consumption

Abstract

We analyze how curiosity drives news consumption. We test predictions of the information-gap theory of curiosity using over 100,000 WeChat news articles, applying NLP methods to construct measures of salience, importance, and surprisingness associated with news headlines, experimentally validating these measures, and using them to predict clicks. Our findings confirm that people tend to consume news when: the headline sparks a salient question; the content appears more important (e.g., emphasized by the headline's position on the webpage or an exclamation mark); the headline refers to more surprising topics (measured as the KL-divergence from a baseline topic distribution); and the headline has lower valence. Information-gap theory helps predict aggregate news consumption. Yet our data also reveal a small negative correlation between the number of clicks and the ratio of likes to clicks, suggesting that while inducing curiosity can drive short-term news consumption, it doesn't necessarily enhance long-term reader engagement.