here are my notes, please edit them or add yours below ross

^^^^Robert Canter, FedEx

-Intersection of the need to perserve customer information in the way confidence will not be broached while sharing with other service providers to enhance services.

-Criminals are not anonymous because of the level of contextual information in their system.

-Several layers of security: logon, intrusion sniffing,

-For shipment have 15 primary databases for all the companies. The web is the best integration tool for us to provide a single access point for customers.

-M&A Phase: how much time and sense is there in completely integrating and removing other databases vs. using abstraction and replacing at the end of natural lifecycles. More a concept of management of diverse technologies with tools.

^^^^Mark Cattini, MapInfo

-Provides locational context to information.
-Different types of information: graphic, business-specific (demographics, etc. Key information is the line information, what customers provide.

^^^^Jeff Jonas, Systems Research & Development

-Non-obvious Relationship discovery by data normalization, record enhancement, degree of seperation (social network analysis) and correlation.

-300 entity revolutions per second in real time.

-Takes customer data and puts it through a one-way hash function, then does entity resolution.

^^^^Gilman Louie, In-Q-tel

-Asymetric and near real-time threats are new. As opposed to traditional batch-based threats allowing monthly planning cycles.

-The information systems in the government was designed to prevent sharing. The Hansen factor. Planned stovepipes.

-In-Q-tel was formed because what was happening on the internet wasnt happening in government.

-Few agencies know what real data they have access to. Have helped government undergo a cultral change, the greatest benefit (not their 40 investments).

-Danger of compromise of integrity is the risk of backlash

^^^^Q&A

RC: Having the advantage of leadership in data management is significant

GL: The real issue isnt knowing everything, its knowing what you need to know to get your job done. Actionable intelligence. Can't overload with information.

GL: Advocated Judicial branch getting involved in privacy issues

Question on the role of analysis. Customers want actionable intelligence, government tends to privatize as many areas as they can, so the analytical role gets pushed to the street (government contractors) which have less accountability. Who analyzes the analyzers?

GL: 90s move to outsource everything. Wouldn't outsource analysis. Takes 8-10 years to train people to analyze data to support policy decisions. Good policy makers will demand to see the data.

But what about the analytical function within the system?
GL: We try to train the agencies that the algorytms matter -- in effect they are policy -- who writes it and validates it against the policy -- needs to be part of the process.

This reminds me of Clay's recent newsletter on how political scientists are best suited for designing the rules within online communities. Similarly, with systems that have privacy consequences, need such people in the loop.

Cory on are we presumed guilty before action? People are frequently not in the loop. No constitutionally guaranteed right to work in a casino. The process starts by putting everyone's name into an investigation.

GL: Data mining/profiling school vs. the data analysis school. Latter starts with lead and follows social network. Former is like the credit scoring system. (he prefers the latter...damn good answer/approach). Its not that profiling doesnt have a place, it shouldnt be the first place to go and we shouldnt take shortcuts This is an ongoing arguement with real debate...people are not just saying "lets roll with it" George Tenant wanted outside influence

JJ: defends the role of data mining as providing valuable info (two people living at the same place)

Check out the Carnival Booth argument that attacks the notion that profiling can stop terrorists:

http://swissnet.ai.mit.edu/6805/student-papers/spring02-papers/caps.htm

I think the point Gilman made was to use traditional analysis, the antithesis of profiling. Follow leads, don't filter.