IIITH’s research reveals that readers don’t trust content with clickbait headlines; spend less time on them and that online Hindi content suffers from the same malaise.
Here’s What Will Happen If This Heatwave Doesn’t Abate!
With unrelenting temperatures scorching most parts of India and bringing forth visions of hell, a headline like the above would definitely ignite (pun unintended) curiosity. But what if the actual article itself left you none the wiser? Well, that’s what clickbait does. An established tactic employed by digital media houses to attract reader attention, clickbait’s USP lies in piquing enough interest in content via a tantalising title.
Vivek Kaushal, MS by Research student at IIITH’s Cognitive Science research centre under the guidance of his supervisor Prof. Kavita Vemuri, has been studying the phenomenon of clickbait and has a few research papers to his credit. “Journalism has changed more in the last 20 years than it has in the 100 before that. Digitisation of media has led to the democratisation of journalism where now anyone can create a website or a blog but the flipside is the presence of clickbait. It is not only a trend that’s going to continue, but one that hasn’t been properly understood,” he says.
Clicks That Don’t Convince
In their research study titled “Clickbait: Trust and Credibility of Digital News” that was published in IEEE Transactions on Technology and Society, Vivek and Kavita set out to determine the impact of clickbait on content credibility. “The question was, do I trust an article with a clickbait headline less than an article with a non-clickbait headline?,” states Vivek. For this, they enlisted 200 participants (100 each from North America and India) who were asked to read through any 2 of 6 news articles where each article was given a clickbait and a non-clickbait headline. They were then asked to fill out a credibility questionnaire. The researchers found that clickbait headlines significantly reduced the credibility of news items. What does this mean? Vivek explains by citing a 2017 study that revealed clickbait content in top tier media outlets is as high as 34%. “The typical underlying assumption is that such news sources don’t have as much clickbait content. But the study revealed that the phenomenon of clickbait is not limited to the fringes of journalism; it’s quite prevalent in the mainstream as well. That’s why our findings – that readers are less likely to trust what they’re reading if it has a clickbait headline versus if it has a non-clickbait headline – have serious implications for the media houses.”
Seriously Skimming
In another research study with equal ramifications for the digital media industry, Vivek and Kavita studied the impact of clickbait headlines on visual attention paid to the articles. The study titled, “Clickbait’s Impact On Visual Attention – An Eye Tracker Study” has been accepted at CogSci2022 – the 44th annual conference to held by the Cognitive Science Society in Toronto in July. According to Vivek, the core idea of the study was derived from Facebook. “In 2014, FB launched an algorithm to filter out clickbait content from their homepage. Essentially they measured the time users spent on each article to understand if the article was clickbait or not. FB’s assumption was that if it’s a clickbait headline, you are probably not spending a lot of time on the article since you are not going to find the relevant information anyway. There was however no analysis done on whether visual attention on articles was impacted by the headlines.” Hence to test the assumption held by FB’s algorithm, the IIITH study used an eye-tracker setup that analysed gaze-fixation from 60 participants recruited from IIITH campus. With attention measured via gaze fixation, they found that an article with a clickbait headline, is likely not to have as much visual attention paid to it, than the same article with a non-clickbait headline.
Another interesting finding was that while participants began reading through articles with clickbait headlines, there was an expected drop-off in gaze after the initial few lines. “Essentially no one really goes ahead and reads the entire article or spends too much time on one with a clickbait headline,” says Vivek, adding, “Let’s say, there’s a headline: This is the best city to live in. There is probably the name of the city mentioned in the first few lines of the article itself. Hence the gaze drop-off happens quickly enough.”
Vernacular Void
When the IIITH researchers first began analysing the prevalence of clickbait, they were surprised at how widespread and common the phenomenon was. Intrigued, they set out to explore its prevalence in online Hindi media. After scraping data for about 4 months from the five most popular Hindi publishing sources that had a Twitter presence, they collected and manually annotated a considerable number of headlines associated with tweets. The most interesting finding was that Hindi has as much of a clickbait problem as the English language. Calling it a huge issue, Vivek says that while machine learning models trained on the English language to detect clickbait headlines exist, there is nothing similar for Hindi. “The ML models are used by both social media and by news aggregators, to filter out clickbait content. But for Hindi consumers of online content, there is nothing. There is no filtering, no optimization happening and that’s because there isn’t enough research being done in Hindi.” While highlighting the gap that exists for vernacular models, the paper Clickbait in Hindi News Media: A Preliminary Study has also created a Hindi database that is now available for other clickbait researchers who want to develop machine learning models out of it. The study was presented at the 17th International Conference on Natural Language Processing (ICON) in December 2020.
Into The Future – Responsibly
For Vivek, all of the research that he is working on adds up to what the future of journalism beholds. “With AI in every field, journalism and automation is not too far behind. Not only will a lot of articles be auto-generated but so also will the headlines. The automated systems may churn them out without any journalistic responsibility and try to capture our attention at any cost – with dubious or misdirectional headlines. Hence checks are needed for publishing houses to ensure that people can trust the information they read.” Kavita Vemuri mulls about the future from the Cognitive Science perspective, saying that it is interesting how credibility and trust structures change over a period of time. “At the basic level, we need to provide empirical evidence if exaggerated headlines translate to reduction in learning and comprehension. In addition to this, there is considerable interest to understand if clickbait also leads to polarization. There is a plan to look at clickbait in text and speech in Indian languages, extending the initial work done for Hindi,” she says.