The attention economy is only just around the corner

In this presentation and paper I would run through the short history which makes up APML and the attention ecomony.

Starting with the attention trust’s efforts in this area. Faraday Media a small startup in Australia launched their product Particls (previously named Touchstone) on the rss/news world and didn’t receive much traction due to its Windows only executable. However Particls came with the concept of a single file which would describe your attention profile.

Naturally this file was XML based and with some clear community participation through the newly formed working group and a clear break from the legacy Particls application. APML is now almost ready for mass adoption.

However very few people have heard of it and those who have are sometimes mis-informed about its purpose and use.

One example of APML is music discovery. Both Last.FM (previous Audioscobbler) and Pandora deal with music discovery slightly differently but both generally find music you like by what your playing in your own collection (last.fm) or voting thumbs up or thumbs down for the current song playing (pandora). After a while of doing either, the system/service can start to gage what kind of music you may also like and giving you just that. Because your feeding back again into the system your preference by either skipping the recommended song or voting it down/up, before long the system has a very good idea of what you like. But what next?

I choose to start buying music I like through itunes or napster. I would then need to search through the millions of tunes for certain tunes and music which Last.fm and Pandora are playing. In a ideal world last.fm and Pandora would support APML and the download music stores would also support APML at least for import. This would mean I could apply my attention profile from one service to another. Hence, napster or itunes would be able to show me stuff which I certainly want to buy plus base any other recommendations on what I do want.

APML is rich but no more complex that atom or rss. Its description of implied and explicit concepts is very powerful and allows you to define almost any kind of preference, including in this example music tastes. Implicit data is added by machines and explicit is added by humans. Which translates in this example to, things played a lot out of your collection or voted up by yourself being explicit and recommendations or tunes thrown in by the system being implicit.

The best thing about APML is Napster and itunes in this example would be foolish not to include the ability to upload an attention profile because it contains detailed information about a new customer which could take weeks or months to gain normally. APML works well for the customer because not only can they move from service to service and not repeat the same task over and over but they can also control there preference data (although this should be done via a 3rd party service, as APML isn’t made for hand editing).

There has been much buzz around APML adoption including agreements from Bloglines (Ask.com), Newsgator and others. But its not just big companies, small developers are already playing with APML including a prototype of my example.

From Paul Lamere on the APML list,

I’ve created an experimental web service that generates music recommendations from APML. For example, with this web service I can generate recommendations based upon my last.fm listening behavior with the request:

http://aura.darkstar.sunlabs.com/AttentionProfile/RecommenderService?apmlURL=http://aura.darkstar.sunlabs.com/AttentionProfile/apml/last.fm/lamere

The resulting artist recommendations are returned as APML in a profile called “Music-Recommendations”.

APML 1.0 is still a few months away but prototypes and awareness is growing but is slowly. There are many examples including some interesting mobile application which can be powered by APML. After reading the paper or watching the presentation the audience member will want to run out and try out APML.