May 30, 2006

Is Mechanical Turk for Digital Human Filters?

The New York Times article, Software to Look for Experts Among Your Friends, brought to mind a couple conversations from my recent trip to Amsterdam. At XTech there were a couple presentations about Mechanical Turk (as well as other Amazon web services) and the big question was "what is a killer application for Mechanical Turk?" Many of us through out suggestions, many good, but I have kept thinking about one.

The one killer application is not really a problem a human solves, but the collective interests and amalgamation of information and trained services humans use and have trained. You see, for me Yahoo's MyWeb 2.0 is my killer search solution as it pays attention to what I and my "community" have interest in and mirrors our vocabulary (we used our own vocabulary to tag items of interest on the web in the MyWeb social bookmarking tool). These items of interest in MyWeb 2.0 are offered to me first in my web searches and more often than not what I am seeking is in this collection of mine or my contact's.

These searches that are based on human filters and collaborative filtering of information and structuring what we find of interest and bookmark bringing the information out from 5 to 40 pages deep in the web. It mirrors our interests and perspectives. But, what if we need information outside of our own interests? Now we need not our own corpus of items and possibly one outside those in our loose collective. Let us think I need information about transmission of disease between monkeys. This is not something I have interest in and I am not sure anybody I know (or would be couple of degrees of separation away) would have interest in. This is where I would love to turn to Mechanical Turk to use the filtering and aggregation capabilities of another person's life filtering. I may not want to add the monkey disease transmission to my own corpus, but I need to use a corpus of somebody who has this interest. In a sense I need to uses somebody's digital brain and information attraction mechanisms.

I am not sure if this is really a job for Mechanical Turk, but it requires another person's interests and permission. The Tacit company seems to have a similar product, but only works on Windows and most of the people I would want to tap are not Windows users, but most have web-based information resources.


Comments are closed.

This work is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike License.