Cathy O'Neil
TT

Biases Hidden in Algorithms

Apparently, President Donald Trump doesn’t like the news that he sees about himself on Google, Facebook and Twitter. So he’s accusing them of being rigged against him — and against conservatives in general.

Absurd as Trump’s claim may be, the tech giants have it coming.

Trump is probably wrong that the gatekeepers of the internet are suppressing conservative views and positive stories about him. If anything, the likes of Twitter have amplified his voice. But he’s right to be worried about bias.

It absolutely exists, it’s a problem, and the tech companies are not being nearly forthcoming enough about it.

The companies are most likely being sincere when they say they’re not politically biased. The source of their bias is strictly commercial. When they help advertisers and others exploit people’s weaknesses and baser instincts, they’re optimizing to profit, not to a particular ideology. But even if the motivation is politically neutral, the result might not be. It can privilege Alex Jones-type conspiracies, encourage anti-vaxxers or propagate hate speech aimed at getting people to kill Rohingya Muslims in Burma.

The biases could be manageable if the companies were capable of accepting responsibility. So far they haven’t been. They dismiss concerns out of hand, so there’s no true monitoring. When biases are discovered, as they frequently are, the companies find ad hoc solutions and do their best to pretend it’s not a big deal. That makes them perfect targets for the kinds of outrageous claims that Trump has made.

The danger is evident in the way the extreme right has coopted the message of my last book, in which I warned about the biases hidden in the algorithms that influence everything from teacher promotions to prison sentencing. My prescription was to hold the algorithms’ creators more accountable. I outlined some of the ways this could be achieved, including federal regulation.

In its review of the book, headlined “How Algorithms Secretly Run the World,” the far-right website Breitbart focused entirely on the first part: how privately held, secret algorithms are feeding us biased information. In other words, it selected precisely the elements that would most undermine trust in any kind of objective truth — which is what I believe to be Breitbart’s mission.

If the big tech companies keep avoiding responsibility, they will play right into the Breitbart agenda. Biases will keep popping up, and people will assume that nothing is real or reliable unless they experienced it directly. If that sounds preposterous, note that even fake video and audio can now be entirely convincing — a fact that Trump is already using to suggest that his self-incriminating statements on live television were somehow “fudged.”

The companies’ defenders will claim that bias has always been with us. That’s true, but in the past we had editors and other human gatekeepers to hold responsible. Now we have computers, controlled by people to whom we have handed far too much power. They need to reinvent their business models with a clear emphasis on policy and accountability, for their own sake as well as for the country’s.

Bloomberg View