Analysis & Commentary:

MachInt 'CUEING' CAN HELP PREVENT FUTURE TERRORIST ATTACKS
by Inderpal Bhandari

In the wake of the drastic events of Sept. 11, Secretary of State Colin L. Powell summed up the failure in US intelligence in an interview with NBC, stating, "in this case, we did not get the cueing we needed". There followed a clamoring for new and improved HumInt, or human intelligence, the development of networks of covert operatives who could penetrate terrorist organizations and help 'cue' the United States to future attacks. However, even supporters of this viewpoint concede that it could take the US up to ten years to build a network of human intelligence that could be effective at stopping terrorist attacks. The question remains: What do we do in the meantime to prevent such attacks?

What I would like to see is a clamoring for MachInt, or machine intelligence, a term well suited to denote the use of computers to provide cueing. How can this be done? Through the use of smart computer programs that alert decision makers to patterns in intelligence data that they are not aware of, inducing them to think out of the box proactively, cueing them to early warnings of terrorist attacks prior to their occurrence.

But I do not hear anything about such use of computers. Consider the post-mortem analysis after the disastrous attack on Sept. 11. Aside from the weakness identified in HumInt, it has been mentioned that while the technical data collection capability of the intelligence agencies is very strong, the analysis of data is weak. That weakness has been attributed largely to the lack of analysts trained in foreign languages. James Bamford in his book entitled "Body of Secrets" quotes former CIA director, Robert Gates: "Sometimes I think we just collect intelligence for the thrill of collecting it, to show how good we are at it. We have the capacity to collect mountains of data that we can never analyze. We just stack it up. Our electronic intelligence systems appear to produce more raw intelligence data than our analysts can synthesize and our policy makers can use". Hiring more analysts may be a good idea but it is not a viable strategy for scaling the mountains of data that are being collected. That answer lies in a developing a MachInt capability.

I read a book recently by Jeffrey Richelson entitled the "The Wizards of Langley". Two pages before the end of a 300-plus tome, he mentions discussions (in April 1999) within the intelligence community of the potential of data mining techniques such as clustering, link analysis, time series analysis and so forth. Seemingly following in the same vein, the recent broad-agency announcement of October 23rd issued by the Pentagon states that there is a need to "develop an integrated information base and a family of data mining tools and analysis aids to assist the analyst".

These techniques are not new -- they have been part and parcel of back-office analysis for decades. I'm talking about taking data mining to the next level. We need smart computer programs that can make sense of data automatically and advise the decision maker directly. In other words, we need to go beyond just thinking about putting together a family of data mining techniques and start thinking about developing a distinct intelligence capability. Just as today, there is HumInt, SigInt (signal intelligence), and Elint (electronic intelligence), there now needs to be MachInt. It may be that there is presently no clear vision of how to develop such a capability for national security. Let me address that by presenting a concept of how machine intelligence should be implemented.

Every report -- be it data, text, graphics, imagery, or, multimedia, -- upon which critical decisions are based should be accompanied by an intelligent computer program that has a two-fold capability. First, it can automatically refute, support or enhance that report by analyzing the data sources that underlie the report. Second, it can explain its analysis clearly to the decision maker. The computer program provides an independent, impartial, analysis. It alerts the decision maker to what the analyst preparing the report may have missed or dismissed.

You, the decision maker, will decide when to follow its advice. In effect, the program acts like an additional analyst that you can turn to when required. And, turn to it you will, especially, when the chips are down, when your game plan is shot, when you are out of ideas, or when your human process has allowed some subtle bias to filter through making you believe that all is well. In other words, when you need it most.

Furthermore, developing such a MachInt capability will structure your entire analysis process, i.e., it will identify where your objectives, measures, analyses, data organization, data collection, inform each other and where they do not. It will also make sure that all parties feeding into a critical decision provide relevant data to decision makers in timely fashion. It will force the different intelligence agencies to co-ordinate their efforts since senior decision makers and their staffs will have a deeper understanding of important findings that underlie decisions, making it obvious which party should be following up on specific open items, fostering accountability.

I advise the decision makers of the intelligence community to look into machine intelligence as envisioned above. It's impossible for an outsider to tell if you are doing this presently but you can certainly judge for yourselves. Consider if you have a MachInt capability linked to your reporting mechanisms. Compile a list of reports upon which your critical decisions are based and then see if you can match them up with accompanying machine intelligence programs that were made available to your staff. Was there such a capability to advise on airport security on Sept. 11? Was there such a capability to advise on terrorist surveillance prior to Sept. 11? Or, more generally, ask yourself what is your last line of defense, when the reports from your analysts are either telling you that everything is hunky-dory or that they are very uncomfortable but cannot put their finger on what is bothering them?

If you do not have these capabilities, you are in danger of falling behind the times. If you can't conceive how a vision of machine intelligence as above can be implemented, you ARE behind the times. Much of what I outlined above can be implemented today, perhaps, even in short order. If you don't have it, act now. You have the resources. Develop a machine intelligence capability as described above to build on the tremendous advantage you already have in technical data collection. It will play to your strengths. It will structure your analysis. It will ensure that the different agencies involved in national security co-ordinate their activities. Last, and most significant, it will get you thinking outside the box when you need it most. It will be critical for fighting a global terrorism network that evolves its instruments of terror.

Dr. Inderpal Bhandari, the founder and CEO of Virtual Gold, Inc., is an internationally recognized expert in data mining, the art of using computer programs to discover knowledge from large amounts of data. Under his leadership, Virtual Gold developed its powerful patent-pending VirtualMiner technology that enables business managers to make more accurate decisions by alerting them to hidden patterns in business data that are discovered automatically. While other data mining platforms can find patterns that have statistical merit, VirtualMiner goes much further by discovering hidden patterns that can actually impact the business. Dr. Bhandari has built Virtual Gold into a market leading company with strategic partnerships with companies such as IBM and Dictaphone and blue-chip customers like Merrill Lynch, Chubb & Son, Bank of America, CBS Sportsline, Boston Celtics, Phoenix Suns and other teams of the National Basketball Association and the Professional Golfers Association.

About Virtual Gold Inc

Virtual Gold Inc, www.virtualgold.com, is a leading next generation business intelligence and data mining software company. Former IBM scientist Dr. Inderpal Bhandari, one of the world's leading authorities on data mining and its timely application to real world problems, founded it in 1997. In addition to its MacInt line of products, Virtual Gold provides an integrated family of Web-based data mining software and solutions in the following fields: contact center, customer relationship management, enterprise resource planning, decision support systems, e-business, and sports and entertainment. The company is headquartered in Hartsdale, New York.