I had an alarmed email over the weekend from a client who had learned that some of her content had been flagged as “objectionable.”
Not only had it been flagged, but it had been blocked from view, so I couldn’t see it and therefore couldn’t immediately help the client figure out what could be objectionable about it.
“Objectionable” means different things to different people. But there are many sites that will not allow publication of, or will not link to, sites with adult content, or sites that can be interpreted as including hate speech, racist or sexist content, incitements to violence, or illegal activity of any kind.
My client was quite sure she didn’t have anything of the kind.
Now, these are things that are objectionable to humans. But it is often robots that make the initial judgement. Robots are notoriously bad at recognizing offensive content. Robots cannot, for example, identify naked people.
This is why humans are employed to scan through all the Web 2.0 sites where everyone can upload their own content.I read an interview once with someone who did this job (I think it was for Photobucket), and he said that by the end of the day he wished he could wash his eyeballs with bleach.
But the initial flagging is often up to a robot. And robots will “look” at a picture like the one at the beginning of this post and see a general pinkness. General pinkness, or a screen full of other colors that could be skin, are likely to be flagged in case they are photos of naked people.
After a few emails back and forth, I was able to understand the client’s problem. My client, who sells clothing, had a close-up of a pink garment. The pinkness filled the whole screen. She had given it a title with the word “adult” in it. The robots, making their best guess, decided she had photos of naked people — “adult content” — which was in violation of the terms of the site.
I changed the title, and presumably a human came to review it, saw that the robot’s guess about all that pinkness was in error, and the content was reinstated.
My client was lucky. The content in question was a blog post, and she uses a platform that quickly flags possibly objectionable content. The error could be fixed quickly. Without that heads-up, she could have lost links to other sites’ robots’ sensibilities.
It’s yet another case in which thinking like a robot has its advantages.
Leave a Reply