"In terms of spammy results, if millions of people are applying upvotes to good results every day, true human-eyeball search QA will be working at massive scale."
And if a disproportionate number of those humans who bother to vote are spammers?
And if a disproportionate number of those humans who bother to vote are spammers?
Definite humans do certain things, like browse the web in certain predictable patterns, their +1 votes might cluster around certain interests, and generally "human looking stuff". Without enough data to look at I can't tell you exactly what that means but I'm assuming that patterns would emerge once compared with computers.
Assume that spammers manage to automate a certain number of google accounts to vote +1 on their web properties. At the click of a button (in low orbital ion cannon style) thousands of robot-accounts log in to +1 against an individual link.
Now the spammers may find a way to stagger the automation of the logins, but unless they're particularly sophisticated how do they manage to create the rest of the simulation of a web-browsing individual?
Do they automate "fake humans" to run google searches from time to time? Click about on links? What about the analysis of the 'clustering' of accounts? A given group of 1000 accounts always voting on the same things together? What about the odd similarity that all the +1's seem to be routed through the same handful of IP addresses? What if +1'ing is closely related to other people +1'ing near your own geolocation?
I'm not saying it's going to be perfect, indeed they're buying themselves another "arms race" - it really would be interesting to investigate what proportion of Facebook's like button is being "astro-turfed"
There's definitely going to be mechanical turk jobs to search my link, click +1 and such, but that exists for facebook as well.
I firmly believe that as clever as spammers are, there're not going to be able to accurately simulate the behavior of real humans for long. There are too many factors that can trip them up.
Again, this is probably one of the things that 25% of the bonus is tied to addressing. Clever, interested engineers now have a serious incentive to spending their 20% time addressing your concern.
Right?
They're just not counted.
Ghost banning can give you a a large number of false-positives without the same level of risk of upsetting others.
In fact, with google's version of ghost banning, they could further tie your votes to the accounts of your friends.
Google algorithmically thinks your a spammer, you're ghost-banned - except to you and except to people who have elected to count you as a friend.
It will be very hard to feel slighted if you never know that your votes don't really carry an impact.
Obviously there's work in getting this right, but it's got a good chance of being a very solid quality filter.
I think it's a deliciously evil solution to spammers and trolls, as you've described.