Hacker News Reader

Comment by aero142 | Hacker News Reader

I still don't get this. What does my social circle have to do with my Google searches. I am friends with lots of people who I don't trust to influence my search results. The Hunch approach makes a lot more sense where it uses recommendations of people who have similar tastes to me. People I am friends with is not a good proxy for people who's opinion I care about when searching for camera documentation, or dog breeds, or Python libraries, or anything at all. I hope I'm proven wrong, but I just don't see it.

View on HN

I think it would be more useful to show +1's from people who have similar search histories to you. That way if you search a lot for Python libraries and +1 useful content, the next guy doing a search for that will get useful recommendations. I don't see how non-techie people in your social circle could contribute in that regard.

This is what I am trying to do with my project with GAE: http://ting-1.appspot.com/rt?rt=python

At this point, I only have my own links; but if other people submit their own python links; and I figure how to sort by relevance, ting may be a good resource for search results without spam.

Problem with finding similar people with similar preferences in Google's data is that the similarity matrix of any 2 people is very sparse.

I search for Python-stuff a lot, but I also search for stuff related to things I want to buy, or for people, or for my hobbies -- photography and now drawing.

The algorithm for finding similar people is going to have to be very smart (i.e. it should not rely on URLs alone), if at all possible.

I think there's two things here.

1. they're trying to match the 'like' button. So you '+1' something, it goes in your profile and your friends can see it. It also would surprise me if there weren't a widget made available to publishers, so viewers of a source article could see if their friends '+1'd it.

2. They'll be able to use '+1' score as a factor in search result weighting along the lines you're suggesting. While they may still weight toward +1s from your social circle, this being Google, I doubt they'll settle for the simplest statistical use of the data.

As for the browsing flow problem -- i.e. how often are you looking at google search results after you determine the result is good -- I'd be surprised if the seemingly inevitable package for site owners didn't allowing people to '+1' directly at the source.

well i think they have to get some amt of data first... obviously they'll add this to the algorithm and let the machine learning decide how much to weight it. I think it will likely increase click throughs by a decent multiple (for the most +1'ed results). I actually did a proof of concept on organizing search results by facebook like counts last year http://news.ycombinator.com/item?id=1372311 (and oddly enough i did see a spike in mountain view traffic this past december :).

According to the blog post, "if you’re looking for a new pasta recipe, we’ll show you +1’s from your culinary genius college roommate."

From the example, what you see is "Laura Delaney +1'd this", so at the very least you can evaluate the "+1" based on what you know about Laura.

Better still would be if they include Laura's name because she's +1'd (what an awkward phrase) a lot of cooking results, maybe in proportion to links on other topics.

The problem with seeing strangers' +1s is that you can't validate their authority. However if I were to see that my friend who is an excellent photographer has +1'd something photography related, I would trust it more. Alternatively, if a friend who's taste in movies I disagree with +1's a movie, I'll know to ignore it.

If people use this, I see how it could be useful.

Exactly. Much of what's obligatorily "social" on the web is orthogonal to who my friends are, why they're my friends, and what I'm doing on the web at that time.

Reply on news.ycombinator.com