Apple News algorithms pick more celeb stories than human editors, says study
When algorithms replace humans in charge of editing news, they tend to select more stories about celebrities from a less diverse selection of publications, according to research.
Jack Bandy and Nicholas Diakopoulos of Northwestern University in the United States analysed Apple News over a two-month period from mid-March, collecting every headline published in two sections of the popular and influential news app: Top Stories, which contains content selected by a team of human editors employed by Apple; and Trending, which contains stories selected by the company’s proprietary algorithm.
Across more than 4,000 headlines, the algorithms selected far more stories, from a far more concentrated selection of publications, than the human editors. Nearly a third of all the “trending” stories came from just two outlets: CNN and Fox News. Adding in the next two, People magazine and BuzzFeed, accounted for half of the 3,144 headlines selected by algorithms.
By contrast, the top 10 sources used by human editors – which did not include Fox News, Buzzfeed or People – accounted for just 56% of the headlines. “On the surface, this suggests that the algorithm could be choosing more stories about celebrities and other ‘soft news’,” wrote Bandy, which was backed up by an analysis of the topics: where the human editors were most likely to pick stories containing phrases such as “Measles cases” or “Brexit deal”, the algorithms preferred “Justin Bieber” and “Florida Man”.© Photograph: M4OS Photos/Alamy CNN was selected by Apple News’s algorithm for its trending stories more than any other publication.
Meghan Markle and Donald Trump were rare topics of interest for both sets of editors.
The findings underscore the problems with relying on automatic curators to do the work of human editors, even for services such as Apple News that try to use both in concert.
Algorithmic editors can also fall prey to other issues, such as the “Scunthorpe problem” – a form of accidental censorship that sees stories blacklisted because of their similarity to obscenities, as with the four-letter word in the middle of the industrial town. In May, for instance, Twitter’s algorithmic filters barred Dominic Cummings from automatically appearing in news roundups because of his surname’s pornographic potential.
However, the problems haven’t stopped publishers from embracing the cost savings automation can produce. Last month, Microsoft decided to replace all of the human editors who curate stories for the company’s MSN News service with algorithms that could do the same work.
Less than a week later, however, the robot editors had embroiled the company in a race row after illustrating a story about Little Mix singer Jade Thirlwall with a picture of her bandmate Leigh-Anne Pinnock, the only other mixed-race member of the group. The algorithms then attempted to post a story about their own error, forcing the few remaining human editors to intervene.
The Guardian – Alex Hern