A new study published in Nature Human Behaviour has shed light on the impact of emotional language on online news consumption. The findings reveal that negative words in news headlines significantly increase the likelihood of users clicking on the news stories, with each additional negative word boosting the click-through rate by 2.3% on average.
The digital age has transformed how we consume news. Today, most people obtain their news online, engaging with content for brief periods despite vast available sources. This shift has driven media outlets to fiercely compete for reader attention, often resorting to emotionally charged “clickbait” headlines.
Prior research has consistently shown that negative stimuli attract more attention than positive ones, a pattern observable from infancy through adulthood. The new study aimed to quantitatively assess how this dynamic plays out in online news consumption, leveraging data from Upworthy.com, a site known for its mastery in crafting viral content.
“During Covid, I noticed myself engaging in ‘doomscrolling,’ or the tendency to let myself be sucked down into a negative news spiral,” said study author Claire E. Robertson, a PhD candidate in psychology at New York University. “I knew it made me feel bad, but I kept going back to the news anyway. That got me thinking about the fundamental process of negativity bias, which goes all the way back to our evolved cognitive mechanisms for threat detection.”
The core of the research involved analyzing a vast dataset from the Upworthy Research Archive. This unique dataset enabled them to investigate the causal effects of language on news engagement within a realistic online environment.
Upworthy was founded in 2012 and quickly became a pivotal player in the U.S. media landscape, especially between 2013 and 2015, when it was recognized as the fastest-growing media company globally. The publisher gained significant traction by sharing viral content on platforms like Facebook, where it outperformed all U.S. mainstream media in terms of shares.
To maximize its reach, Upworthy engaged in thousands of A/B tests, experimenting with different headlines and images to see which combinations most effectively drew readers to their articles. In 2017, Upworthy merged with Good Worldwide and contributed their extensive historical data to create the Upworthy Research Archive.
The researchers had access to data from 22,743 A/B tests conducted between January 2013 and April 2015. In these tests, different headlines were randomly assigned to news articles, and the subsequent user engagement with each headline variation was recorded. The primary metrics collected were the number of impressions (the number of times a headline was displayed) and the number of clicks each headline received. The click-through rate (CTR) for each headline was calculated by dividing the number of clicks by the number of impressions.
The dataset included over 105,000 different headline variations from Upworthy, which collectively garnered approximately 5.7 million clicks and achieved over 370 million impressions.
Robertson and her colleagues employed a text mining approach to analyze the headlines. This involved extracting words from the headlines and categorizing them as either positive or negative based on predefined dictionaries. The analysis quantified the frequency of these emotional words relative to the total word count of the headline, allowing the researchers to assess the impact of each additional emotional word on the likelihood of a headline being clicked.
They discovered that the presence of negative words in news headlines significantly increased the likelihood of those headlines being clicked by readers. Specifically, for an average-length headline, each additional negative word boosted the click-through rate (CTR) by 2.3%. This result underscores a broader trend observed in media consumption where negative information tends to draw more attention than positive information, a phenomenon often referred to as “negativity bias.”
On the other hand, the researchers found that positive words in headlines had the opposite effect. Increased presence of positive language in headlines correlated with a decrease in click-through rates. For headlines of average length, each positive word reduced the likelihood of a click by 1%. This finding suggests that while positive news might be perceived as less urgent or compelling.
“I think the best thing people can do is be mindful of how news is framed,” Robertson told PsyPost. “What I think is really unique about our study is that we were able to look at the effects negative language on news consumption when multiple headlines were for the exact same story. And, we found that simply adding one negative word in a headline increased the likelihood that that headline would be clicked on by about 2.3% on average. While this may not seem like a lot, it can easily aggregate online where millions of people are clicking on hundreds of articles a day.”
The researchers employed multilevel binomial regression models that allowed them to account for other factors that could influence engagement, such as the length and complexity of headlines and the novelty of the publication’s tactics over time. Despite these controls, the influence of emotional language remained significant.
This pattern of engagement was observed despite the varying subjects of the news stories, suggesting a general preference for negative headlines across different types of news content. Furthermore, the study noted that while the effectiveness of emotionally charged headlines gradually diminished as the novelty of Upworthy’s editorial strategies wore off, the overall trend of negativity fostering more clicks remained consistent.
The researchers also found that specific emotions significantly influenced user interactions. Sadness had a statistically significant positive effect on CTRs, increasing the likelihood of a click by 0.7% for each standard deviation increase in sadness within a headline. But emotions such as joy and fear showed a negative impact on CTR, with increases in these emotions reducing the likelihood of clicks by 0.9% and 0.7%, respectively.
Interestingly, anger, which was hypothesized to influence CTR based on previous studies suggesting its prevalence in viral content, showed no significant effect on news consumption.
“We were surprised that we found no effect of anger on news consumption, and instead found that sadness was driving clicks,” Robertson said. “A lot of prior work has found that anger motivates sharing behavior online. However, sharing is a very public behavior, and our study focused on a private behavior — clicking on headlines. So, we think (and are currently investigating) whether sharing content online and consuming content online are driven by different motivations.”
Despite its insights, the study has limitations. The focus on Upworthy’s specific style of sensationalist journalism may not generalize across all news platforms.
“Upworthy was one of the originators of ‘clickbait’ which is a specific journalistic style,” Robertson noted. “Arguably, many media companies and online influencers now use some sort of clickbait tactic to garner attention online, but the data we looked at ranged from 2012-2015. The effect of negativity was present throughout the time series of the data, but we did find that stories were clicked on less as time went on, suggesting that Upworthy’s clickbait tactics were most effective at gaining attention when they were more novel to consumers.”
The study provides several avenues for future research. One area of interest is exploring how these findings apply across different platforms and types of media to validate the generalizability of the results. Additionally, further studies could investigate the psychological impact of negative versus positive news on readers over longer periods. Researchers could also examine the effects of discrete emotional words and how different emotional tones influence news consumption across various cultural contexts and demographics.
“I’m very interested in differences between our public and private behavior online,” Robertson said. “There has been a lot of research on the type of content people share online, but we also know that only a small subset of online users (who tend to be the most extreme ideologues, at least in the political realm) post on social media with any frequency.”
“There are a large subset of ‘lurkers’ so to speak that consume media but do not share their own opinions. Thus, moderate opinions online may become invisible, and extreme opinions dominate and are misperceived as the norm. This may create an overly-extreme online environment — we recently wrote about this hypothesis in this preprint.”
Robertson also noted that the paper was the result of a unique situation — two research teams simultaneously and unknowingly conducted similar studies using the Upworthy Research Archive and submitted their findings to Nature Human Behaviour around the same time. “If you want to read more about the unorthodox way the paper came to be, check out my post about it here.”
The study, “Negativity drives online news consumption,” was authored by Claire E. Robertson, Nicolas Pröllochs, Kaoru Schwarzenegger, Philip Pärnamets, Jay J. Van Bavel, and Stefan Feuerriegel.
