The LA Times' AI ‘bias meter' looks like a bid to please Donald Trump

The Guardian

05-03-2025

The past few months have been brutal ones for the readers and journalists of the largest news organization in California, the Los Angeles Times.
Since he bought the paper in 2018, the billionaire and medical entrepreneur Patrick Soon-Shiong has become something of a Donald Trump acolyte.
That's his right. Many media owners have political views; but the best keep those views to themselves, or at least allow their news organizations to exercise editorial freedom.
But Soon-Shiong, who took over promising to steady the ship and return it to financial health, has turned out to be a deeply flawed leader. You might recall that many longtime subscribers canceled their subscriptions months ago when Soon-Shiong blocked his editorial board's decision to endorse Kamala Harris for president.
Then he reportedly told his editorial board to 'take a break' from writing about Trump, and, according to a staff memo signed by members of the opinion section, instituted a policy in which articles critical of the newly elected president were to be published side-by-side with the opposing, pro-Trump, view. That's straight-up meddling.
But now, he's taken a more public-facing step by inflicting what's become known as a 'bias meter' on some LA Times opinion pieces. Its findings are generated by artificial intelligence, without human intervention or review.
If there's one firm rule about the use of AI in journalism, it's this: there should always be a 'human in the loop' before publication. Why? Because AI, at least at this point, is often wrong on the facts, and because many news consumers are suspicious of it.
At the LA Times, the AI-powered 'Insights' feature evaluates opinion articles and puts a label on them – for example, 'center left.' Then it provides 'different views.'
Articles about Trump-related policies have gotten the bias meter treatment – for example, an opinion piece on Ukraine that stated that 'Trump is surrendering a century's worth of US global power in a matter of weeks.' According to the Guardian's Lois Beckett, that piece is followed by an AI-generated summary of 'different views', such as describing Trump's policy as 'a pragmatic reset of US foreign policy'.
Soon-Shiong called the new feature a victory for viewpoint diversity. 'No more echo chamber,' he crowed on social media.
It looks more like a way to avoid offending President Trump.
Let's get real. Many opinion pieces at legitimate publications these days are critical of Trump – for good reason, given the chaotic damage he and his helpers have unleashed.
So this effort is less a rooting out of lefty bias than a way to give a platform to pro-Trump views.
At well-run news companies, it is journalists themselves – editors, in particular – who can point out unfairness, inaccuracy or bias. And they deal with that, editor to writer, before pieces are published.
'Our members – and all Times staffers – abide by a strict sense of ethics guidelines, which call for fairness, precision, transparency, vigilance against bias, and an earnest search to understand all sides of an issue,' the LA Guild, the union representing the paper's journalists, said in a statement objecting to Soon-Shiong's idea.
These days, many of the opinion-side journalists at the LA Times have fled. This is apparently no longer a place where they feel they can do their jobs.
Soon-Shiong's gambit is happening in a broader context of media companies yielding to Trump's will, as Axios's Sara Fischer aptly noted.
Journalists are doing their jobs, but owners are 'compromised', she wrote, listing some of the most prominent examples: ABC News settled a defamation suit by Trump it could have won; CBS seems poised to settle Trump's absurd claim against its flagship 60 Minutes show; Disney and Paramount have rolled back some DEI policies; the Washington Post's opinion section will reflect owner Jeff Bezos's beliefs about 'personal liberties and free markets'.
Some of the bias-meter results so far are simply weird, as in an AI response to an article critical of AI itself. The original piece, by two experts in film production, explored the dangers of AI-generated footage within documentary films and how it could shatter audience trust in the visuals they see.
The AI-generated bias meter labeled this piece 'center-left' and provided 'different views'.
Another piece, reflecting on the history of the KKK in Anaheim, California, included an AI-generated defense of the Klan at the bottom, as the tech journalist Ryan Mac pointed out. It's since been removed.
I can't imagine what reader would want to trot around in this silly circle like a horse on a lead line. Most of us can read a viewpoint article and decide, all by ourselves, without a helpful robot, whether we agree.
In the name of viewpoint diversity – but really to push his paper Trump-ward – Soon-Shiong has done far more harm than good. His bias meter should – quickly – go the way of hot type, the manual typewriter, and the dodo.
Margaret Sullivan is a Guardian US columnist writing on media, politics and culture