To Err is Human, But Algorithms Can Be Screwy Too

The “related reading” you commonly see alongside news articles are based on computer algorithms so that editors don’t have to spend time manually plugging in related content.

The concept is simple: a computer judges — based on keywords, date, and other factors — what articles are “related.” Then it automatically posts the article’s headline and featured image on the margins or at the bottom of page. (At the bottom of HonestReporting articles too, you’ll see You may also like, accompanied by four articles — each just a headline and thumbnail image.

Nobody ever thinks about these things until something goes wrong.

After all, despite the mad mix of keywords and computers, you can still get some screwy results. Case in point is the LA Times.

Can you spot what’s wrong here?

LA Times

No, Palestinian protesters didn’t really build a scale model Dome of the Rock in the Jordan Valley.

So where did things go awry for the Times?

The original article itself featured an image that absolutely had nothing to do with the headline. The photo’s relevance was because the dispatch tied to “in other developments” about a separate Temple Mount clash. So the image was related to the article, but not to the headline. And when the algorithm inevitably kicked in, the result was blatantly noticeable,  but forgivable.

Less forgivable, however, is other instances when algorithms recycle a “related” image that gives the misleading impression of being part of the story itself (as opposed to related content). Sky News, is a great example of that

To err is human, but algorithms can be screwy too.

  Like what you just read? Sign up for more: