Algorithms can be incredibly useful tools for services like Instapaper, which uses them to sort and filter popular articles for its subscribers. But without sufficient human oversight, algorithms can result in major embarrassment, like they did last week when Instapaper unintentionally sent the F-word to its readers.
In a weekly newsletter featuring the most popular recent stories, Instapaper highlighted an article by author Mark Manson, entitled “The Subtle Art of Not Giving a Fuck.”
One week later, the service has apologized for the mistake, blaming an algorithm that was not designed to avoid stories featuring bad language.
“The top highlight in last week’s email contained some … colourful language, and we’re sorry if you were offended,” the team noted at the start of a follow-up email. “The Weekly is an algorithmically generated newsletter based on the most popular articles and highlights saved by Instapaper users, and unfortunately we didn’t build the algorithm to filter profanity in any way.”
Instapaper says it has now made changes to its algorithms. “We’ve added in some filters on our end to ensure that future content remains as interesting as ever, while avoiding any potentially offensive language,” the message continued. “Again, we are sincerely sorry if you were offended, we’re still getting the taste of soap out of our mouths!”
Speaking with Fast Company, Instapaper CEO Brian Donohue says the company is taking precautions to ensure similar mistakes do not happen in the future. “We have users who don’t realize that Instapaper is algorithmic, and think that we actually chose that content for their inbox,” Donohue says. “I don’t want those users to think it was an editorial decision–or that it reflects the principles of the company. We want Instapaper to be a family-friendly service.”
Last December, Uber’s surge-pricing algorithm charged riders fleeing a hostage crisis four times the normal fare. And in 2013, Amazon was forced to pull a clothing line by the company Solid Gold Bomb after its algorithms generated slogans like “Keep Calm and Rape a Lot” and “Keep Calm and Hit Her.”
By those standards, Instapaper’s flub could have been a lot worse. However, it’s yet another reminder of why–despite algorithms’ ability to carry out a great many tasks—we still need humans at the wheel.
[via The Guardian]