Research: Editors, algorithms recommend different stories to readers
Readers First Initiative Blog | 11 June 2020
Two Northwestern University researchers audited two Apple News sections: one curated by editors and one by an algorithm. Humans selected more stories on policy and international events, while machines featured more celebrities and entertainment.
- Jack Bandy, one of the authors of the academic paper scheduled to be published later in 2020, has popularised the findings in a Medium post last week.
- It’s the first data-backed analysis of Apple News editorial practice. The app has more than 125 million monthly active users, or about half of what Google News enjoys.
- Algorithmic curation of news feeds, home pages, or e-mail newsletters is increasingly adopted by news publishers, including the subscription leaders such as Aftenposten of Norway and The Times of London, that aim to offer more personalised and engaging experiences.
- The financial squeeze many publishers suffer from during the pandemic accelerates automation. For example, two weeks ago Microsoft sacked the editors curating home pages of the MSN Web site, planning to replace them with robots.
Academic researchers analysed 4,412 stories recommended over two months of 2019 by Apple News, of which one-third were “Top Stories” curated by human editors and the remaining were “Trending Stories” selected by an algorithm.
- Editors chose sources more evenly: While the app aggregates more than 80 news sources, the top 10 ones claimed 75% of all algorithmically curated Trending Stories. In contrast, the top ten sources chosen by human editors claimed only 56% of all Top Stories.
- Editors chose more diverse sources: Algorithms chose most often CNN, Fox News, People, and BuzzFeed, while human editors were choosing from a wider array of sources — for example, The Washington Post, NBC News, The New York Times or The Wall Street Journal.
- Editors chose less “soft news”: The headlines of the algorithmically selected Trending Stories were dominated by keywords such as “Donald Trump,” “Meghan Markle,” “Kate Middleton,” and “Justin Bieber.” The Top Stories chosen by humans most often mentioned in the headlines “hard news” keywords: “measles case,” “sanctuary cities,” “affordable care act,” or “New Zealand mosque.”
In summary, automated curation generates different results than editorial one. Algorithms such as the one by Apple News optimise for engagement, while editors include other factors in their judgment such as the importance of news events or credibility of sources.
Automation brings risks but also benefits — for example, it frees time of journalists to focus on original reporting and analysis. It is required to personalise recommendations to make the news sites relevant to individual readers. Newsrooms have also the choice of which tasks to automate, and which not to.
As algorithms take over roles of editors, their designers become the new editors-in-chief. Newsroom leaders cannot outsource the job of setting the algorithms’ goals and rules to engineers. They may wish also to look for success criteria beyond engagement, for example, ensuring readers stay up-to-date with the current affairs or growing their knowledge about subjects of importance and interest.
To automate or not to automate? Where do you stand? What are you working on?E-mail me at: grzegorz.piechota@inma.org
Banner image courtesy of Pete Linforth from Pixabay.