Tuesday, September 19, 2017

Selective Implementation Of The Algorithmic Ideal Is Intellectually Dishonest: Hillary Clinton Book Review Edition

The thing that got me wildly, disproportionately pissed off this week was the news that Amazon had been deleting one-star reviews of Hillary Clinton's new book. What the actual fuck?

The reason this enraged me so much didn't have anything to do with Hillary Clinton or the content of her book, but rather the way it laid bare the depth of intellectual dishonesty and cynicism of the tech-retail industry. I used to think the algorithmic approach to the world was coherent and reasonably well-intentioned, while also massively and dangerously flawed. But this episode shows that "coherent and reasonably well-intentioned" may have been giving people too much credit.

As we've discussed before, and as you can learn much more about from the master herself, Cathy O'Neill, the algorithmic approach to the world is massively and dangerously flawed along several dimensions. There's a pernicious illusion that algorithms are "neutral," when they're anything but. The problems of legitimacy are problems of trust, not data. The attempt to draw lines while avoiding ethical judgment leads to policies that disproportionately target people who are already oppressed.

Still, I was able to sort of see the inner logic of the thing. I mean, I believe algorithms are problematic. But I could see how someone who didn't believe that might push forward toward more data and more rules and less judgment and so on. I think they'd be wrong, but they'd have a coherent position on the world, and might be acting with integrity.

But then going on and making an exception ... for Hillary Clinton? Are you kidding?

As I understand it, the main reason was that they thought the reviews were posted by people who hadn't read the book. As the publisher is quoted saying here, “It seems highly unlikely that approximately 1,500 people read Hillary Clinton’s book overnight and came to the stark conclusion that it is either brilliant or awful.”

But ... this a problem for all Amazon reviews. Admittedly, it's usually a less dramatic problem. But there are plenty of stories of people trying to sandbag other people's books for all kinds of reasons, there's sock-puppetry, and there are all kinds of other problems with the review system.

While yes, Hillary Clinton's book prompted a dramatic showcase of some of these issues, she's also like one of the most powerful people on the planet. She hardly needs protection from her critics to get people interested in reading her book. It's bizarre to me that it would be like, well, when it's everyone else, sorry, but when it's a power player politician, oh, then we have to do something.

When things like this happen, I always try to imagine how these decisions get made. Does Amazon have a hundred meetings in a row where someone brings up fake reviews and blah blah blah, and then on the 101st they bring up Hillary Clinton and people suddenly say OMG we have to do something? Is it that the issue doesn't come up those 100 previous times because the incentive structure of power players at places like Amazon discourages it?

Or is the ultra cynical interpretation the right one: that PR with prospective customers will work best when they intervene selectively in only the most high-profile situations -- i. e., just when it is the power players of the world?

In any case, I'd just like to remind everyone that they can shop for books at Indigo, Kobo, Barnes and Noble, and Powell's. I'm not saying these are some kind of do-gooder organizations: in a dog-eat-dog capitalist society, being less than ruthless than your competitors is death, so bad behavior is kind of baked into the system. But isn't it better to have several flawed tech-retail companies than just one flawed tech company, which can then go ahead and rule the world?

No comments: