January 2006
Sun Mon Tue Wed Thu Fri Sat
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30 31        

Editor Login


Convener in chief:


David Lazer
(Methodology, Networked Governance)

Editors:


Stanley Wasserman
(Current Trends, Methodology, Social Networks)
Guy Stuart
(Economic Sociology, Finance)
Allan Friedman
(Simulations)
Nathan Eagle
(Technology, Social Computing, Powerlaws, Current Trends)
Ben Waber
(Technology, Social Computing) Ines Mergel
(Knowledge Sharing, Social Computing, Social Software, Current Trends)
Maria Binz-Scharf
(Qualitative Methodology, Knowledge Sharing, eGovernment)
Alexander Schellong
(Admin, eGovernment, Citizen Relationship Management)

Categories

 

Archives

Recent Entries

Recent Comments



Notification

Enter e-mail to subscribe:
Powered by
Movable Type 4.21-en

Syndicate this site (XML)



Blog Directory & Search engine
Academics Blog Top Sites

globe_blogs.gif
Blog Flux Directory


« Pew Social Ties Report | Main | Follow up: Google bombs and the autonomy of search engine vendors »

30 January 2006

Retrospective vs prospective data analysis

Before I leave the NSA thread, one more thought about the utility of the data mining phone log data, recording phone calls, etc. It is important to note that pattern recognition is only as powerful as the ancillary intelligence that the government would have to complement the phone data. Obviously, in the absence of any indications of who is high risk, the signal in the data is infinitesmal.

A corollary to this is that pattern recognition of today's data will be much more effective in the future-- because the amount of complementary information will only go up. We may have no reason to pay attention to Joe, except that he was 3 degrees removed from someone in some terrorist's phonebook. However, in the future we may receive solid intelligence that Joe is a terrorist. At that point, it would be a pretty good idea to go through all of the data you collected about Joe-- if you have it sitting in a hard drive somewhere. If you haven't been recording, and collecting information from the switches, etc, it may be impossible to reconstruct afterwards. That is, "retrospective" pattern analysis is likely to be much more effective than prospective pattern analysis. Of course, this in turn points to a strategy of collecting data now and asking questions later, which again brings us back to the issues around privacy and collateral usage in aces.

Posted by David Lazer at January 30, 2006 2:29 PM