Lurid titles called “clickbait” on the web no longer bring readers. US scientists come to this conclusion. […]
Particularly raunchy titles, also called “clickbait”, are not as well received by readers as thought, as researchers from Pennsylvania State University (Penn State) and Michigan State University show in several studies. In some cases, this even reduces access, it is said. Another finding: Artificial intelligence (AI) is finding it extremely difficult to distinguish clickbait articles from normal posts.
“Much more complicated than we thought”
“Because fake news is a huge problem on social media, research is working hard to find ways to use AI to systematically identify and block clickbait,” said Shyman Sundar, director of the Media Effects Research Laboratory at Penn State. However, the latest studies would suggest that it is not so easy to track down false news in this way. “Some believe that you only have to solve the clickbait problem to solve the fake news problem as well. But our studies show that clickbait is much more complicated than we thought.»
“One of our initial questions was whether certain clickbait properties can generate more clicks than others,” explains Maria Molina, assistant professor in the Department of Advertising and Public Relations at Michigan State University. These include headings with question marks at the end or with exaggerated negative or positive superlatives. The result surprised the experts themselves. “When we analyzed this in more detail, we found that there were no significant differences. People seem to find even regular headlines more appealing than clickbait, ” Molina says.
Three investigations carried out
The team of both universities conducted three studies. In the first, 150 people were presented with one of eight randomly selected headings on various topics and checked whether they subsequently read or shared the corresponding articles. In the second, 249 participants were recruited to see eight headlines on the same topic – only one of them was Clickbait. “In both studies, the clickbait posts could not get more hits than the normal texts,” says Molina.
In the course of a third study, it should be clarified whether AI can help to automatically track down clickbait in order to combat the spread of fake news. For this purpose, different systems and models of machine learning were tested. “These systems only agreed in 47 percent of cases on what is and isn’t classifiable as clickbait. So people who see clickbait headlines as an element of tracking fake news should reconsider their assessment, ” Sundar says.