I'm researching attention data and privacy issues just keep jumping out at me.  If an attention system is implicitly monitoring everything I say in IM and Email and when I browse theb, read my feeds there are huge privacy implications especially if that data is stored on someone's server somewhere. This is augmented with the likelyhood that good attention systems will be monetised through advertising. At this stage, what is the difference between an attention system and adware like Claria ? Can I own the data? Keep it on my system? Approve what data is added to it? Are there any good debates on this subject already anywhere?

I like this concept:

"We can make them pay for it, lease it, scream for it "show me the money",
barter for it, whatever. The important point is that we get to decide
who has access to it, how long they have access to it, and what we want
in return." http://worcester.typepad.com/pc4media/2005/07/attentiontrusto_1.html

But how do you practically do that? 

From the same URL though, here's what sounds yucky:

"If Bob Saget is looking for a potato slicer, Acme Potato Slicer Inc may
send him one for free if he lets them tell the world they sent him one.
Bob Saget's attention is worth a whole lot. If you are Kevin Federline
and you are looking for condoms, (Please look for condoms, Kevin.)
Trojan will probably pay you to use them. Who knows what they'll do?" 

I've invited Chris Saad into this group. He's the CEO of Touchstone. I'm sure he has a lot to say about the subject.

Thanks for inviting me, Mick. Interesting topic

I think you hit the nail on the head Mr Phil. I think it's important for us to be able to manage our own data (add, remove, modify, delete, etc
.). To get the benefits that attention data afords, we need to allow access to that data, but it doesn't mean we need to give up complete control.

I blogged some stuff about my 'Attention' thoughts so far: http://www.morle.net/phil/archives/2007/02/the_attention_d.html#comments

Chris Saad has commented. 

Hey guys - sory for being late to the party.

Its OK Chris - this is an all nighter!

As my comment on Phil's blog suggests,we have been thinking about these issues for a long time and decided that the first step to protecting privacy was to store the data locally. The second step was to define and declare an open, human readbale format so that end-users could look into, modify and distribute their Attention Profile as they see fit - no servers involved unless they choose to submit it to one - just like OPML

we call it APML - www.apml.org

The difference between Claria and a real attention management product is that Claria is designed to give you ads for the benefit of advertisers and the distributors of the software. Real Attention Management Engines should be designed to help you manage your attention for your benefit.

For Attention to reach its full potential IMHO there comes a point where the user needs to deploy their APML into hostile territory and it is this moment I am most challenged by. Many of the use cases for attention data are around advertising. User makes their attention data (or a subset of it) available and the web site can optimise the experience to make it relevant for them. But now I have given a website the crown jewels - like giving some sites you email address and getting spam from 100 other people from there on in...

Well yes and no.

First your attention data can be served from an "Attention Brokerage Service" that allows granular control over access rights

Second giving an application access to your attention profile only helps it create a more personalized experience for you (yes including advertising) - which can not really be a bad thing - if you're going to see ads, they might as well be relevant

Third just because thy know what you're interests are, does not mean they can reach you - they stll need a conduit like email which is already filled with spam - maybe at least the spam will be better targeted

A brokerage can also support a blacklist to help users identify suspect behaviour on the part of requesting services

Can you say more about how a brokerage service might work?

What about creating a link between open id and apml?

Open ID already supports brokering permission based data, and has the authentication ready.

The party's just begun

What..a...great...idea

We have been planing to do it for quite a while - just a matter of time

So if anyone wants to get in on the action and join the team I'm all for it

Let's take an example of the problem and get into some specific solutions. Person A and Person B want to compare their interests. Let's assume there is a badge on their blogs like MyBlogLog. Person A is at Person B's blog and sees that the badge indicates that they have similar interests. After pressing a magicbutton, APML files are compared via a brokerage service and Person A (who is looking to hire person notices that they both like the same kind of pornography and recommends a site... how do we stop that happening?

Of course, this is an exaggerated example but there are many more subtle ones that woul dbe hard to guard against... 

Well depends on implementation - but I would imagine that the comparison could be blind - so you don't know what the similar interests are - only that there are some - so u could get a 'compatability rating' or something

also Touchstone blocks a lot of porn out of your attention profile with an adult filter

and also why is having porn in common a bad thing

Chris, I am extending APML in my mind beyond Touchstone.  Currently I am using Touchstone as an attention filter to see how well it sorts the wheat from the chaff.  It seems to be doing this really well. And there are no privacy issues at all that I can see.

But I want to use APML to find new stuff, new friends, tighter relationships. In that model the comparison can't be completely blind.  I need to know more than the fact we are connected. For example, when we connect in LinkedIn - I get all your connections. I get some data.

 Porn is just one extreme example so is easy to filter (at least the obvious terms) but what about more subtle things?

Probably the main need I have as a user is some process of approving what it deployed into my APML file. That will also get over the problem that I am not at all interested in pig farming in Adelaide, I just needed to research it my son's school project.

Yes I see what you're saying - I only use Touchstone as an example because it's my best frame of reference. But speaking more abstractly, I don't think it's possible to build in privacy and security in by default. Any solution will require the right kind of implementation to respect privacy and security concerns.

Like anything, you can use or abuse it.

A brokerage service can be as safe, secure, interoperable and federated as you like. But as soon as you give access to your profile to another service - that service has (by definition) an ability to retrieve information about you.

It is up to the user (common sense) and perhaps some sort of community rating/reputation system to help weed out the offenders.

The same is true for an implementation of relationship system whereby two people's APML is compared. The implementation could be to expose the matches and reveal the APML files to each other - but that's just bad design.

A better implementation might be to reveal a percentage match (or put it in terms of 6 degrees of separation) without actually revealing the whole profile. Maybe the user can tag part of the profile as public and part as private as well.

The issue of Pig Farming in Adelaide also comes down to implementation. If the app is only profiling you based on your interactions with a narrow silo, then sure the results may be easily skewed by aberrations. But if, like Touchstone, it uses a larger data set of information, skews towards Pig Farming are less likely or quickly corrected.

So what I guess I am saying is that while these issues are great to discuss, and best practices important to establish, the ultimate responsibility (as always) lays with the people you trust with your APML file (or email address or credit card or home address).

I like the idea of building the best practices, a distributed reputation layer and an API to determine the level of repuation - some sort of SSL for Attention

Can we win an award for longest msgs?

I think Marty and Mick may well we overjoyed with out long messages!

I like long messages!

I like long walks on the beach

Great talk the other day Phil. I hear you changed someone's research plans.

Did someone Record it for us loosers who live in Brisbane