Hacker News Reader

Comment by what | Hacker News Reader

What do you mean by "discovered an effective method for human powered search that covers the whole internet?"

Their indexing seems rather easy to game. From what I understand you have to include some FB specific metadata tags to tell FB to index it and what's on the page. What's stopping someone from putting metadata tags to say it's a page for a hotel somewhere or a movie, when they are really selling Viagra?

View on HN

Isn't it because the "pagerank" of your page increases by how many people "like" your page instead of how many other sites link to your page. This really is quite interesting because it's easier to press Like than it is to write an intelligent post with a lot of words that links to the site that you like. The downside seems that every site has to buy in and put the like button on their site, but it looks like it's already too late and there's not much choice.

It's still easy to game, because there's no "dislike" button. Show 50% of your visitors a really valuable awesome page, and the other 50% the viagra spam. You'll get a ton of "likes" from the awesome page, and hence a lot of traffic for the spam.

Furthermore, with fake accounts, it becomes even easier to game facebook pagerank.

You're assuming that Facebook's ranking algorithm is incredibly naive and is only counting the raw number of likes.

What if you add in whether you're friends with anyone who liked it. Or at least a few degrees away.

Facebook already finds and deletes fake accounts. Trust me -- I've had my test accts blown away regularly!

All good points, but in all of your scenarios having fake/stolen accounts would still allow you to game the system. Friend networks can be staged/stolen just as easily as an account.

Unless they implement trust. "Your friends [in this friend list] like these pages today."

Allowing only trusted search results would greatly reduce the number of links returned. For most searches no results at all would be returned.

When you click through to the page, you'll see if any of your friends liked it.

Sure, but couldn't all the spammers band together and "like" each other's pages? Or set up a bunch of dummy accounts to like their own pages.

Plus I'm not too sure about the types of pages that will get liked and indexed. I haven't logged into Facebook in a ages, but I don't remember people liking or sharing links for things that I would ever use a search engine to find. But that could just be me.

Presumably you can already buy "likes" from someone who has amassed a large number of FB accounts? If not then it won't be long at all.

Carrying that one step further: why buy "likes" from someone if you can steal them? Maybe the headline should be:

Facebook Offers New Incentive for Identity Theft

I was assuming they're going to be including data from the like button.

EDIT: > What's stopping someone from putting metadata tags to say it's a page for a hotel somewhere or a movie, when they are really selling Viagra?

Facebook isn't going to blindly put things into its index because someone tagged a page. I also doubt search results are just gonna be everything tagged with that page but based on how many likes they page has, how simliar those people are to you. In my opinion, Facebook has the potential here to make this into the first innovation in search since Google.

Reply on news.ycombinator.com