I get hundreds of comment spams a day for this blog. If it’s not obvious what that means, it means that spammers leave spam comments on my posts. Since the email address they enter isn’t on my “Allowed” list, their comments go to a queue where I decide whether to approve them or not. Obviously, I don’t approve spam.
So let’s say there are a hundred items in the queue. Usually, they’re all spam, but I have to check them, of course, because I want to approve legitimate comments. I can pretty much scroll right through it, letting the comments go past my eye, lickety-split.
Every once in a while, something will just look different, and I’ll stop and see that it’s a legitimate message from someone. Then I approve it.
But isn’t it interesting that I can do that so quickly, almost subconsciously, while there’s nothing in the computer world that can do it at all with anything approaching my accuracy? If I take my time, I can do it with 100% accuracy. (But it’s hard to take time, since they’re almost always spam.)
Going as fast as I do, I probably miss something legit every now and then, but I’m still so much better than a computer that there’s no comparison. Computers are dumb.
There are some pretty good algorithms out there for spotting this stuff. My Yahoo account automatically filters most spam to a special folder. I regularly check it for legit messages that get marked as spam, but very few have been. Meanwhile, my inbox gets plenty of good mail and very little spam.
It does well, but not close to what how well you’d do. That’s why you have to check it.
But even that, I bet, starts with using a white list of email addresses that are ok. But I could pretty much spot spam even without a white list.
So it’s reasonable to assume that there are visual triggers in the structure of your spam messages that tip you off. There’s a skip in the rhythm as you scan down the page.
It’d be interesting to see if a program might have success at analyzing screen shots, rather than parsing the text.
Still, people are even better at processing images than computers are, but it’d be interesting to see where that goes.
Yeah, it’s definitely something visual. That would be interesting.
Computers are fast, efficient and stupid. Humans are slow, inefficient and brilliant.
I have forgotten who i am quoting here
Check out the Akismet plugin for WordPress. Since I started using that I went from a hundred or so a day to less than 3 per month. It works by keeping a database of known spam and looking for matches and automatically trashing them.
Thanks Kevin. I’ll do it.
I don’t know if Yahoo uses a white list. If so, it builds it implicitly.
John, I’d generalize more to say that people are better at pattern recognition than computers are, but agreed anyway.
James, the sad truth is that most people are slow, inefficient, and stupid.