Sunday, December 14, 2008

learning, organizing, anticipating: better

I joked to my colleagues the other day that I wanted a personal assistant who would not only accurately intuit what I wanted for lunch, but proceed to feed it to me so I could seamlessly work (with two hands) and eat. Pathetic, yes, but after reading about CALO today in the NY Times, I’m optimistic things might pan out.

CALO is the “Cognitive Assistant that Learns and Organizes. After spending some time on the website, I gather that it will automate many of the ‘mundane’ activities in our worklives (e.g. set up meetings with the right people at the right times, and best yet, prepare you with relevant information), based on interactions, activities, and instructions.

Think: Intelligent Amazingness-- it seemingly learns your preferences and priorities by your desktop activities, online behavior, who you physically come into contact with, and perhaps even how you interact with these individuals. From the website:
“The goal of the project is to create software systems that can reason, learn from experience, be told what to do, explain what they are doing, reflect on their experience, and respond robustly to surprise.”
Although it seems unrelated, in a later post, I want to talk about this in relation to Facebook Connect and Friend Connect. I shouldn’t even make this link because the CALO concept is wholly unrelated, but I think it’s important to call out the differences in predicting future behavior based on our daily behaviors, ‘bumper stickers,’ and our connections.

I’ll just say, I’m a huge proponent of figuring out who people are, and anticipating future behavior by what they do in their daily lives-- in fact, it was the topic of my dissertation. I’ve been anti- self-report measures for several years and several reasons.

So this is an exciting development - a new model of intelligent software learning about your worklife and how you navigate it- bound to play a critical role in the evolution of the web and enterprise, even if just inferring what I want for lunch.


John Sheridan said...

Hi Kate,

This article caught my eye, too.

While predictive bots may make your life easier, they can also provide an unbiased and enhanced perspective for the 'profile viewer', as you point out.

This will be important for the Reputation Index we've talked about before. Of course, personal preferences for data exposure will be controllable, but ultimately, it will become important to share the data for the "leaderboards". No one ever has a problem with keeping their number of Twitter followers private, do they?

And I'll bet you'd like a Hoagie for lunch. (^:=


(Super Self-proclaimed Guru of Lunch Predictors. Esq.) (^:=

kate said...

Interesting perspective, John - Thank you. I think you're interpreting my post in terms of figuring out others, based on behavior - and our willingness to be transparent vs. private in 'committing' said behaviors, given others' ability to know us in that fashion. While this is an implication, what I am more interested in is the knowledge gleaned ourselves-- because of its actionability, in being based on our own behaviors. Along these lines, what about me, my leaderboard or language has led you to believe I'm a hoagie-eater! Back to the drawing board...

johnsheridan said...

Hmmm. Guess I'm not a guru after all. (^:=

I can see where you were going with the value in self-analysis. Many do that now with the limited data available. But, it leads us to the wrong conclusions.

I just also thinking more intelligent mechanisms could be re-purposed.