The Kudasu experiment relies on the simple premise that the information we share with the devices we use has much to say about us. We think that by interpreting this information we can gain insight about the kinds of persons we are, and come to a deeper understanding of ourselves, our actions, and our future.
Search engines are used every day by millions of people for a wide array of purposes. Besides e-mail, search queries are the most revealing snippets of information that we share with our computers. Just think for a moment about the searches you've made in the last few days and you'll realize that some of them are related in some way or another to your deepest desires, fears or aspirations.
It would, nevertheless, be near to impossible to learn much about a person only by looking at what he or she searches for, even over the course of years. However, if we multiply this single user by the millions of people that search the web every day, we can rapidly accumulate enough information to learn a lot about this whole collectivity. This is what Kudasu is about: learning about people considered as a whole.
Since the experiment will need time to gain regular users, we have set it's time span to a whole year. Hopefully by then we will have millions of queries of raw data to interpret. This is where the Kudasu algorithm comes into the picture.
At the heart of the experiment lies the Kudasu algorithm - the computer program that will analyze the search queries we receive looking for an answer to the question: are human today good naturally evil?
The Kudasu algorithm is being programmed to make sense out of all the apparently chaotic and unrelated information inputted by search users and attempt to uncover patterns that will help it decide this question. It will develop and evolve along the course of the experiment: as the volumes of data increase in order, old data will be re-interpreted in light of new data and the algorithm will modify itself accordingly. The Kudasu algorithm is therefore a dynamic, changing thing, and it is determined to a large extent by what you search for.
We have modelled the evolution of the algorithm into three phases. We estimate that the first two will take around five months to complete. The final phase will last around two months.
Goal: To build up a critical mass of information that makes refinements to the Kudasu algorithm feasible. Below this critical amount there is not enough information to justify altering its initial form. The focus of the experiment in this phase will therefore be that of gaining enough daily users to reach the volumes of data required for statistical accuracy.
Goal: Using the data gathered in phase 1 and continuosly during phase 2, the Kudasu algorithm will be modified and refined. This will be the most important phase of the project, since we will begin to see the algorithm perform with live data.
Goal: To finalize the Kudasu algorithm while gathering the final set queries. At the end of this phase, on March the 1st 2008, the algorithm will be run for one last time and its final judgment will be broadcasted to all users.