Perfect Attention (With a Little Help from My Friends)

Back in the summer of 2007, I wrote this post about how aggregators should be able to find articles that interest you by paying attention to what people like you are reading.  It's a very simple idea – after all, if 10 people you know all like the same article, then there's a pretty good chance you'll like it, too.  But as is often the case, the simplest ideas often take the longest to implement.

Using an aggregator like FeedDemon ensures that you'll see stuff that interests you because you usually only subscribe to feeds you find interesting.  But there are tons of feeds you don't know about that may interest you, so there's a good chance you're missing a lot of articles you'd enjoy.

The traditional approach to solving this problem has been to show you the stuff that's interesting to everyone (even "mainstream" news sites like CNN show the popularity of each article).  But I personally hate this approach because it shows me so much stuff I don't care about it.  For example, a lot of people may find the latest antics of pop culture icons interesting, but I couldn't care less.

I don't want to see articles about Britney Spears, I don't want to know how many women Tiger Woods slept with, and I don't care if Kanye just made an ass out of himself.  But I do want to know if Google has an interesting new service, or if Microsoft released a new development tool, or if someone declared that RSS is dead (again).  And I also want to know when someone has written a great response to an article that interests me.

So knowing what everyone is paying attention to doesn't interest me, but knowing what people like me are paying attention to is very interesting.  I've tried to tackle this in FeedDemon for quite a while with little to show for it, but when Google Reader added the ability to follow people, I knew it was the "secret sauce" to finally solving this problem.

Long story short: the next build of FeedDemon will include articles from people you follow in Google Reader, and these articles will help fuel the "Popular in My Subscriptions" page which until now has shown only the articles that are popular in feeds you subscribe to.  I've been using a build of FeedDemon with this feature enabled (here's a screenshot), and it has brought me tons of interesting articles that I would otherwise has missed (many of which I've added to my Shared Items feed).

PS: I should add that if you only see stuff that's interesting to people you follow, then there's a risk that you'll end up wearing blinders and miss "general interest" articles that may be relevant to you.  For this reason, FeedDemon also looks at the feeds you're subscribed to and figures out which articles in those feeds have been "liked" the most by everyone (not just people you follow), and it interweaves those articles into the "Popular in My Subscriptions" page.

Comments

[wioota]

Sounds great - for me its the feature that puts your reader in front of everyone else's.

It would be nice to have the popular topics feature available per folder as well (and also if you shared the algo with Brent for NetNewsWire - even though your affiliation is not as strong - so I can use it on my mac at home...).

It surprises me that a decent equivalent which can consume a published opml is not available (there are a few out there but the results are average and they require manual upload of the opml file).

[ben]

The "people you follow" feature in google reader is nice, I agree, but it relies on the people you follow "sharing" the articles. So it still isn't a totally automated way of finding interesting articles. You could follow someone who reads very interesting articles, but never clicks "share".

Also, they could share a blog post from a blog you already subscribe to, so now that shared item shows up in both places. This isn't a huge deal, just a minor annoyance.

[Nick Bradbury]

@ben: I agree with you, much as I'd like not to :) The number of people willing to share (or even "like") an article is a fraction of those who actually read it, but we're still a step closer to finding articles relevant to you based on what those you follow are reading.

Really, this all boils down to a matter of privacy. Everything would be drop-dead simple if nobody cared whether everyone knew about the articles they were reading. If privacy wasn't an issue, then software like FeedDemon and Google Reader could do away with manual actions such as "share" and "like." They could just track everything you're reading and use that information to suggest relevant articles to those who follow you.

For example, if FeedDemon knew what all your friends are reading, it could show you articles that interest you - but it could also show you that some of your friends like articles about having sex with lawn ornaments. If you only follow a few people, you'd probably find it easy to figure out which of them are lawn ornament aficionados. And chances are, you'd rather not know about this, and your gnome-loving friends would also prefer you didn't know about it.

Which is why manual actions such as "like" and "share" which give the user a choice as to what is shared are being used in so many places these days :)

[Jack Brewster]

Only really noticed GR's trends section today (inspired by your post) so I started poking around and started wondering, since you're using GR's info, what are the chances of incorporating its trends info in addition to/replacement for FD's least/most attention reporting? I tend to read a lot of posts on my iPhone so some feeds show up as low attention in FD.

[Pit]

well i never agreed on your constant attention to twitter, google reader and other "web 2" stuff that reinvents the wheel.

i begged you for better filters/watches and you said you'll try to fit them into v3 years ago :)

anyway, i downloaded rss-owl and was amazed as it does everything i ever begged you to do :) i remember testing it before and it was a shadow compared to feeddemon, but now - it's years ahead. sorry to say that because i really loved fd. i'm not writing this to promote any product, just to give a feedback why i left feeddemon. maybe it can be useful if you notice declining number of users.

i follow feeds, i don't waste time browsing through sea of ramblings of all internet journalist wannabes. google blog/news search and even stuff like slashdot are in my junk folder and article gets attention only if filters hit it. i needed that, not a google reader frontend.

[Nick Bradbury]

@Jack: incorporating trends in FD could be very nice, especially since it would reflect your feed reading outside of FD. But best I can tell, trends isn't exposed by the GReader API.

[ben]

Ha, good example of privacy implications, my lawn ornaments are now in the garage :)

But, isn't it still tough to tell which articles someone likes w/o their intervention? You could see my opml file and see which blogs I read, but I don't find every article on every blog interesting. Sometimes I just 'mark all as read' or just fly through them to get them off my reading list.

[Nick Bradbury]

@Ben: Yes, trying to figure out what people are actually reading w/o their intervention is very error prone. This is a big reason why FeedDemon relies so heavily on user actions such as starring, tagging and sharing when determining a feed's "attention rank."

[john]

To your point about pop icons, when will you have the ability to delete posts based on certain keywords so that I never see them? I want to be able to delete every post with "Britney Spears" and "Tiger Woods" (and other keywords I enter) automatically as they come in. Thoughts?

[Alexey]

@john: But then, you would miss this post of Nick's, which includes those keywords.

[John]

@Alexey,
True :). But that's always the case when you filter. You accept that you might miss one interesting post for the sake of missing 99 irrevelant ones. I'll make that trade-off!

JOhn

[benwaynet]

One other place to look for information like this would be friendfeed and items you 'liked'
I use feeddemon and have it sync with google reader, but I don't use google reader to follow people, I find friendfeed much more useful for this.

Love feedemon keep up the great work.

thanks,jb

[Povl]

when are you making a commercial version ??
i loved yor RSS reader but is now looking for other

[Nick Bradbury]

@Povl: by "commercial version," do you mean "ad-free version?" If so, you can already get rid of the ads in v3.0 - just select "Purchase" from the Help menu.