Machine Learning: How Algorithms Get You Clicking

Machine Learning: How Algorithms Get You Clicking

Technology may be able to free journalists from creating sensationalistic stories to generate online traffic, but it doesn’t alter what people choose to read.

Click-o-tron, as described by creator Lars Eidnes, “gives us an infinite source of useless journalism, available at no cost.”

It is a website that, when not overwhelmed with traffic, employs machine learning to generate clickbait headlines for news that never happened, such as “New President Is ‘Hours Away’ From Royal Pregnancy.”

It barely makes sense but you want to read it anyway. It’s almost as astute in its randomness as The Onion manages to be through deliberate humor.

“Clickbait” refers to sensationalist stories that aspire to maximize online visitor traffic by appealing to emotion rather than reason. Though a source of shame among many journalists, it’s also a source of revenue, something scarce in a world where few willingly pay for content. It has become the foundation of a handful of successful online publishers such as Buzzfeed, which subsidizes substantive reporting with pablum tuned to maximize attention and ad impressions. For example, consider Buzzfeed’s “26 Adorable Dogs Who Are Slaying The Fall Fashion Game.”

Eidnes, a software development consultant at Itema in Norway, created his system to automatically create clickbait news headlines in order to free writers to do more meaningful work. “If I remember correctly from economics class, this should drive the market value of useless journalism down to zero, forcing other producers of useless journalism to produce something else,” he explains in a blog post.

It’s a noble goal, through arguably misdirected. Driving the cost of clickbait creation to zero doesn’t change the consumption preferences of the online audience or the economic incentives that follow from those preferences. Rather than teaching machines to craft clickbait, we might do better to teach humans to demand a higher-quality media diet.

[Read about how software-based machine learning attempts to emulate the same process that the brain uses.]

Using machine learning turns out to be easier than reeducating tabloid tastes. By training his machine learning system on roughly two million headlines harvested from Buzzfeed, Gawker, Jezebel, Huffington Post, and Upworthy, Eidnes has created a website that generates irresistible, though sometimes nonsensical, news headlines.

“World’s Most Dangerous Plane” and “How To Get Your Kids To See The Light” could be real clickbait stories. “Obama Has More Surgery, More Xbox One Coming?” doesn’t quite work, but Eidnes suggests that crowdsourcing story prominence through voting can weed out the incoherent headlines.

The headline, however, is only the sales pitch. To close the deal, Eidnes will need to generate coherent stories for his provocative links, instead of the gibberish that’s there now. That may be beyond the scope of his experiment, but it’s not as implausible as it sounds. Machines powered by software from Automated Insights are already generating financial earnings stories for the Associated Press — and machine learning keeps getting better.

To read more:

Please follow and like us:

Post a comment

Your email address will not be published. Required fields are marked *

close slider
  • +44 (0)203 004 9596
  • This field is for validation purposes and should be left unchanged.