As popularized by Google Page Rank, the idea of gleaning valuable information about documents (or anything) from traces of user-behavior (e.g. links, footprints, etc…) is quite the rage. Scientists constantly approach the online world with new tools to uncover or analyze traces of human behavior. 

I thought that it is interesting to point out that the online world is a artificial place that was and is continually engineered and reengineered. Since web designers and others are in control of this environment, that means that, in addition to watching humans behave, we can design enviroments that lead users to behave in such a way that we learn more about what we are trying to gather data on.

For example, in a search engine, we can construct the search results page to also collect data on potentially relevant hits. For example, we could occasionally place a search result that technically has a lower score within the top ten results. This “mistake” and the related click-through data can then be used to calibrate the search engine. 

Granted, to build a webpage that collects user data is not a new idea, but it would be interesting to apply that theory to online spaces that scientists have been approaching in a very naturalistic fashion.

From zdnet

In an article in the Asian business publication Tech-on reporter Tomohiro Otsuki writes:

Retailers and contract manufacturers in Taiwan say that novice PC users there, like students and housewives, tend to buy the Linux version of the Eee PC701, while geeks go for Windows XP.

Does that sound backwards?

Yet a quick look at Amazon shows that Asus Eee’s with XP roughly $35-$100 more than their Linux brethern. Housewives know a bargain when they see one.

Or… to take another shot in the dark, is it that students and housewives are so unfamiliar with XP that the difference between XP and Linux is negligible?

Although, another factor to take into account is that the flavor of Linux on the eee pc has been advertised as simple and straightforward.

I just thought of a model for looking at sources of meta-data. Thought I would share it with people who want to employ meta-data from systems like Flickr, Facebook, Twitter, etc… to the design of intelligent and aware information systems.

Marc Davis just gave a very interesting talk at the dub meeting today on social mobile computing and massive social-technical systems that take advantage of platorms like flickr tagging and cellphones with meta-data.

I got distracted in the middle because his (very good) talk got me thinking.

Davis is interested in leveraging two features of the mobile consumer. One is the existence of a powerful, ’sensitive’, and ubiquitous computer. This is the mobile phone: it has cameras, processing, displays, bluetooth, GPS, compasses, and accelerometers. The second is the network effect. There are millions, soon-to-be billions, of these devices canvassing the globe. Intersected, these two factors offer an enormous array of potential. You can calculate Where, When, Who… and infer What and maybe even Why about all kinds of activities and objects. This enormous potential for deriving context can be applied to previously “NP-complete” tasks: computer facial recognition was his very nice example.

He also suggested another application of contextual meta-data: if you’re about to call someone on the phone, wouldn’t you like to know where he is so that you don’t interrupt him? or you might be reminded to talk to him about something (“I didn’t know you were in Tahiti, too!”). This would be the power of applying this meta-data to computer-mediated activities (making a cellphone call is certainly that).

I feel that this scenario surfaces a number of different responses. (E.g. “No, I wouldn’t want to do that.” or “I can totally think of a better application than that.” and etc…). We can revisit those later.

What I ended up thinking about is the source of this contextual meta-data. The two that Davis talked about are the two that I think are the most popularly perceived right now. You can arrive at contextual meta-data through one of two means:

a. User-input or

b. Sensor-data

Let me illustrate this with annotated sketch from my notes at the time:

So, examples of user-input are facebook status messages, twitters, responses to evites, tagging, etc… Examples of sensor-data are exif data from digital cameras, access_logs, and Intel’s mobile sensor platform. Within the cellphone scenario, this fits like so: “I’m about to call Daniel, but the mobile sensor platform on his cellphone tells me that he’s driving a car. Therefore, I decide to wait to call him later.” Or: “I’m about to call Daniel, but he just Twittered that he is in a meeting with his professor. Therefore, I decide to wait to call him later.” This is when I got distracted. I thought, those aren’t the only options.

There is a third source: myself.

Because *I* could input data about what *I* know about Daniel’s context. So, by distinguishing between “me” and “them”, the original dichotomy becomes a triangle like so:

Some clarifications. Two important roles are distinguished. One is the target. This is the person whose meta-data is worth collecting: Daniel, in the scenario. The other is … well, qui bono, the decision-maker, the sense-maker, the user: “me” in the above scenario. The unspoken part of all the scenarios about Daniel are that I am also applying my knowledge about Daniel to the information that I get from sensors or offered by Daniel himself. Suppose that I know that Daniel is avoiding his girlfriend and has been in “meetings” all day. Or that Daniel does or does not have a hands-free device in his car.

We can even go back to one gut reaction to the original cell-phone scenario, “No, I wouldn’t want to do that.” I presume that one objection is that sensors and twittering border on the edge of surveillance and invasive knowledge. A system that is based solely on “personal knowledge” basically sidesteps these issues and allows us to examine the scenario on different merits.

By acknowledging that the user’s mental model of the target is another source of meta-data (and certainly it is equally relevant to the user’s decisions), I think this expands how we think about and design information systems that infer conclusions from a sea of bits.

Where to go from here

Privacy issues: This triangle allows us to look meta-data driven systems that are sensitive to privacy. Or, maybe they raise new questions about privacy – I don’t know. Are personal notes and observations that I make about another person’s activities a violation of their privacy? What if my personal information management system (yes, we’ve been talking about PIM all along!) remembered notes and observations that I would have forgotten long ago? Or made connections between bits that I never would have noticed (“e.g. Didn’t you say, earlier, that she was unemployed?”).

Incentives: When would I decide to manually input observations about another person?

Testing the model: What is data collection that is part-automatic and part-offered? Would that be when I use my Safeway card knowing that it is been recorded? How about other points on the triangle?

Social Theory / Lit. Review: This model could be supported (or challenged) with a better understanding of how people see or model other people in their heads. Right now, all I can think of is Goffman, but I know there’s more on how I perceive people (and how I perceive how people perceive me, ad inifinitum.)

Well, its been fun…

I guess I got bitten by a bug. I got excited and wrote a post — I hope it provokes thought and that the idea holds water. I guess you don’t know until it survives a little criticism. Who knows, maybe I’ll come back to this idea in the future. Peace.

To design novel technologies is a way to advance the functional capacity of the existing technological infrastructure in a user’s everyday life. The contribution of the telephone system, the Internet, and the tools that make it more powerful or easier to use: the cell-phone and the web-browser, for example, is that it changes the infrastructural landscape such that what was previously difficult or impossible became a technological affordance and, eventually, was embedded into the rhythms of everyday life. These tools now exist as infrastructure: taken for granted, appropriated for a variety of tasks, and, often, essential.