Library clips

Minekey is a recommendations widget for your blog visitors, just whack in your own feed, or enter many feeds. I’m not really sure how this works, I guess it must match the content on the page to offer other content…so how’s this different from Sphere.

I’d like to see this based on an APML attention file…firstly, the new Amazon recommendation widget works like this, when you visit Amazon it stores a browsing cookie in your cache, it collects your clicks and calculates your preferences as a base to recommend stuff.
Take Amazon to the widgetsphere, if you visit a blog that has an Amazon recommendation widget, the javascript in the widget will check to match you with a cookie at Amazon and can then recommend you stuff, just as if you were at the Amazon site yourself.

But when you are at Amazon it recommends you stuff based based on the page you are on, whereas with the widget it’s just recommending stuff from your pool of past surfing at Amazon, it’s not recomending you stuff in the context of your present attention (page you are on), in that case it may as well be a Stumble Upon widget just for Amazon.

So it seems the recommendations from the widget have nothing to do with the content of the page you are on, whereas it could look at the contents of the page you are on and recommend you stuff filtered through your past surfing and purchases at Amazon…this is recommendation based on where your immediate attention is focused on.

I hope I have understood this correctly, let me know if I’m mistaken.

APML

Amazon is one site, but what if you are not really looking for anything to buy, but you like being recommended stuff while your are on the web.

What if we had a social network, like Facebook, that could synch an APML file from your PC…all your web experiences are stored in this APML file, and when you log-in to the social network it looks for your updated AMPL file.
When you go to your Facebook profile it will recommend people and content based on things in common with your total web experiences.
When you surf around the network (visiting profiles, checking out content) it would recommend you stuff that is similar to the nature of the page you are on and further refined by filtering it through your APML attention file. Contenders for this type of network are Facebook as mentioned, but there are already networks focused around the whole idea of attention, like slifeshare and Clutzr…but if your community is at Facebook, we need these attention networks features absorbed into Facebook.

Since you are part of a network and logged in, when you surf the open web, the page you are on will recommend other pages based on content in the APML base (this would be housed in Facebook), but it would further filter this through people who have things in their APML that are in common with your APML. Not sure if Others Online is similar, BlogRovr recommends stuff for the page you are on, but it’s based on a feed set of your choice, not based on a social attention pool, which is then further refined through you attention filter.

APML widget

Similarly, when you visit a webpage with a supposed APML widget it will recommend you other webpages by matching the content of the page from the social pool of APML data and filter it through your social connections and your personal attention filter.

I like the non-capitalist idea of this as no-one is trying to sell you anything, you are just getting recommendations of stuff from a social pool, based on people with similar interests who having been surfing the web.
Then you can tune it, because up till now the computer doesn’t know whether you have like or disliked the recommendations…voting, rating, etc…

What about an APML widget not based on recommending stuff in relation to the page you are on filtered through your APML file, but instead just a widget in general based on the latest activity at the APML social network.
Instead of having to visit the network you could get a widget for your startpage, and it would be like a Stumble Upon where you can click or auto generate the latest fed recommendations. These recommendations are not related to your immediate attention, it’s just random recommendations or even recommendations based on the last hours activity.

The intimate marketing comes in when services can filter their data through your APML attention file
eg. filter my APML file through the Amazon website data and then give me recommendations…so Amazon would be able to recommend you stuff even if you have never visited their website in your life.

For bloggers you could put an APML button on the footer of your blog posts, visitors just fill in their user name and click (unlike the widget, not sure the footer button would be able to find your APML file on your PC), and it will recommend similar webpages filtered through your attention, then you could tune it by rating it…or it could be site specific stumbling.

Collaborative Micro-Filtering

So far I’ve talked about recommendations based on a pool of data at a social network, filtered through my personal preferences and past behaviour.

These recommendations are offered as I surf around the social network, and as I surf the open web, and also as widgets on a blog or startpage.

Another element to the recommendation filter may be built around trust…this is a meme going round started by Bokardo, and refined by Sarah Cooper…came across this via Green Chameleon.

This meme is more about the circles of trust and how the closeness may differ depending on the context of the request or topic…how does this link to our discussion in this blog post.

Well, so far we have talked about being recommended stuff (from a pool of data stored in a social network) based on our attention filter…what about extending this attention filter to friends in your network, so it is no longer just based on a personal APML file, but on a social APML file. This opens up the narrowness of personalisation, as we still want to be pleasantly surprised by recommendations that aren’t normally what we are into…serendipity.

Or better still when I plug my APML into a webpage, it will recommend me stuff based on my past attention, plus also recommend stuff based on my friends (circle of trust) attention.

What I mean is the pool of data I am being recommended stuff from is from a social network, and these recommendations are filtered through my APML file, but what about also saying “if your friends visited this website they would be recommended these….”

This way we would get a personal list of recommended websites, plus another list of recommended websites based on the mixed APML file of my friends (you could also limit this view to one friend or a selection of friends).

Instead of just a list of recommended pages it could be visual like demonstrated in Sarah’s post, for any web page (represented by a central icon) it could list recommended webpages dotted around the centre, the closer the more heavily recommended.

Then we could also view this for our mixed friends APML file, or it could be in the same circle as your personal APML file with the nodes being in a different colour, and the nodes in common with yours in a different colour again.

What about being recommended people from a social network like Facebook that have common interests as you, by explicit people tags, but also by attention files:

- plug-in your APML file and it would recommend profiles of people that seem similar to you
- when you are on a random webpage, you could view the webpage as a icon in the middle of a display and people from your social network could be dotted around this concentric circle display, showing you the degree of relatedness they are to this webpage.

Coming back to Sarah’s diagram, what about topics, click on a topic and it would recommend people dottted around concentric circles, but these are people from the network, not necessarily people you trust, so filtering this diagram through your friends would be exactly what Sarah has demonstrated.

I also like the idea of AutoRoll…people who visited your Facebook profile also visited these profiles.