Researchers have developed an AI to detect clickbait along with headlines written by machines.
There are many problems facing reporting today including fake news, alt-truths, general lack of factual accuracy, state manipulation, deepfake content, imprisonment of journalists, and those pesky misleading headlines we call clickbait.
Many of us claim we don’t fall for clickbait, but most of us will have clicked on one at some point. The headline is normally something along the lines of “You wouldn’t believe what the Teletubbies baby looks like now!” (Yes, ok, I fell for that one embarrassingly recently.)
Researchers at Penn State and Arizona State University tasked a group of humans to write their own clickbait. Machines were then programmed to write artificial clickbait headlines. This data was then used to train an algorithm for detecting clickbait written by people or machines.
The researchers claim their clickbait-detecting algorithm is around 14.5 percent more accurate than other systems. Beyond its use for detecting clickbait, the researchers claim their data can help to improve the performance of other AIs.
Dongwon Lee, principal investigator of the project and an associate professor in the College of Information Sciences and Technology, said:
“This result is quite interesting as we successfully demonstrated that machine-generated clickbait training data can be fed back into the training pipeline to train a wide variety of machine learning models to have improved performance.
This is a step toward addressing the fundamental bottleneck of supervised machine learning that requires a large amount of high-quality training data.”
One problem the researchers faced is the wide variety of clickbait available and the need for it to be labelled as such in order to train the model.
Shyam Sundar, professor of communications at Penn State University and another author of the report, explains:
“There are clickbaits that are lists, or listicles; there are clickbaits that are phrased as questions; there are ones that start with who-what-where-when; and all kinds of other variations of clickbait that we have identified in our research over the years.
So, finding sufficient samples of all these types of clickbait is a challenge. Even though we all moan about the number of clickbaits around, when you get around to obtaining them and labelling them, there aren’t many of those datasets.”
125 journalism students and 85 workers were recruited from Amazon Turk to supply the human-written clickbait for the research. The recruits were tasked with reading short 500-word articles and then writing a clickbait headline for each.
AIs such as the one developed by the Penn State and Arizona State University researchers could one day be essential for combating the threats to journalism which have a knock-on effect to the freedoms and liberties we hold dear.
You can find a copy of the full research paper here (PDF)
Interested in hearing industry leaders discuss subjects like this? Attend the co-located 5G Expo, IoT Tech Expo, Blockchain Expo, AI & Big Data Expo, and Cyber Security & Cloud Expo World Series with upcoming events in Silicon Valley, London, and Amsterdam.