Attention - profile first, recommendations afterwards - Jon Galloway

Attention - profile first, recommendations afterwards

What do I mean by Attention?

The concept of Attention (as it relates to the internet) is still a new and amorphous enough concept that it's hard to find a simple definition to link to. Wikipedia has almost nothing to say about it. There are technologies and services built around it, like AttentionTrust and attention.xml. Meme engines (like TailRank and Findory, which personalize news based on user input - OPML import for TailRank, OPML import and click tracking for Findory) get close but aren't the full picture. Robert Scoble is excited about Attention, but he's too hung up on the commercial and marketing side of it. Steve Gillmor , Alex Barnett, and others talk about it pretty often, but without any real practical demonstrations of what they're talking about, it's still pretty theoretical at this point.

A better way to approach the concept of Attention is to talk about the problem we hope it will solve: we're drowning in a sea of information, and while we have pretty good tools to search the information, we're so overwhelmed that we can't keep track of what we've seen or even what we're interested in. The hope is that we can track and effectively direct our Attention instead of just randomly bookmarking links and forwarding e-mails.1

They know us better than we know ourselves...

We've reached an odd place - marketers understand our behavior better than we do. We know Netflix tracks rentals and uses that information to recommend films we'll be likely to enjoy based on the recommendations of members with similar rental and rating histories. If you asked me what movies I've watched in the past year, I wouldn't do nearly as well as the Netflix system. At this point, Netflix's understanding of my cinematic tastes is better in some ways than my own. The same applies in many other spaces, from supermarkets to casino loyalty programs. But, Attention is about more than targeted marketing on the internet. Attention data offers much more than contextualized marketing.

Get your priorities straight

The idea of tracking our own Attention is to use the same datamining approach marketers use to understand and and more efficiently manage our time and, well, attention. We need to spy on ourselves as well as the marketers do to better understand:

  1. where our attention is focused, and
  2. how to use that knowledge to find the information that's of most interest to us.
Those order is important - first understanding what we're doing presently, then (and only then) using that information to focus those interests which are most important to us.Our attention defines us to the extent that in a very real way, it becomes our digital identity. If Attention does nothing more than help us better undersand ourselves, it's been a big help. Jumping straight to recommendations (be they blog feeds or contextualized ads) without telling us who we are ignores a big part of the promise of Attention, and loses some credibility by putting the cart before the horse. It smacks of a prescription without a diagnosis. Netflix does pretty well with recommendation transparency - they tell me why it recommended a movie for me. I'm sure they're not telling me all the details, but understanding a recommendation helps me evaluate it. On another level, this kind of openness demonstrates a recognition that I deserve some visibility into my digital profile.

I realize that collaborative filtering is complex, so it may be difficult to explain the recommendations. At a bare minimum, though, Attention based sites should tell me something I don't already know about my profile - at least some aggregate data. That's one of the reasons I like Last.FM so much - the priorities (and sequence) are in the correct order. The recommendations and personalized radio are great touches, but the real killer feature is the cake is the personalized profile of my musical attention.

1 I'm especially interested in this because I subscribe to over 1500 RSS feeds, most of them technical. It's extremely difficult to effectively process that much information on a regular basis. I'd hoped that AmphetaRate would help with this, but it hasn't panned out yet.

Published Tuesday, February 14, 2006 8:25 PM by Jon Galloway

Comments

No Comments