That Ted +1'd instead of Like made me go looking for info on the differences and I found : http://www.webseoanalytics.com/blog/google-plus1-vs-facebook-like-the-similarities-and-the-differences/ I need to think this through. Has it been discussed in JC?
David Gutelius - I can't recall if your innovation piece on adding context to search talked about looking at the volume of 'general' likes, or the same but weighted more highly for likes by friends / people you follow, is going to be incorporated into search?
And slightly off-topic, the place where I would use this in JC (if provided) would be to semi-automate how I vote on ideas. For example, when I discover an idea - usually because I see it in the All Activity stream because someone, anyone, just voted on it - and the title intrigues me - this is what I do when I view the idea.
1. Stop if I find I have already voted on it.
2. Was it created by someone I 'know' - you can't determine that 'easily' because it's not exclusively people I follow or track, it's just people where there is mutual bi-directional interaction which is of above average sentiment. Note there are times when there is bi-directional interaction with a 'troll' which I don't want the system to use as a positive signal. Whereas mutually supportive bi-directional above average sentiment should be treated as a close tie / close circle.
Basically, in my head there is a 'close circle' of people who fit this category - I could give you these names if you want to use me as a lab rat.
If the idea was created by one of this 'close circle' then I'm more likely to spend time on appraising it.
2. Look to see who I know from the 'close circle' who has already voted (up). I am more likely to spend further time appraising the idea if 2 or more close circle people have voted up. And more so if the close circle people actually commented with above average sentiment.
3. A more minor influencer on my likelihood to vote up are how many votes has it already got (by anyone).
4. A more minor influencer on my likelihood to not vote up are if it was created more than, say, a year ago as it is likely to be attempting to fix aspects that are possibly handled differently in 5.0.
The basic notion we have is that Likes are a relatively weak positive signal of interest proximity, and those will figure in our larger modeling approach. My hunch is taking the simple approach here to begin with (where all Likes have a similar relatively low weight) should yield useful results, when combined with everything else we're taking into account. But automating liking would start to work against that signal as a source for learning more about you. I can't dive in too far into the "how" side of this here but we're brewing up ways of supporting something along the lines of what you suggest - but using a much wider array of signals. This gets to the notion that influence over action really does depend on the extent to which you trust someone on that topic (whatever it is).