SMART SOFTWARE

18. The New, New Intelligence

by Stephanie Overby

A year ago, the US Army began using intelligent software agents instead of people to route the background files of soldiers who required security clearance to the proper authorities for review. The result: a process that once took days now takes 24 hours. The Army reduced its year-long backlog, and the Army Central Clearance Facility in Fort Meade, Maryland, can now handle 30 per cent more requests a year. The intelligent agent retrieves the necessary background information from existing records and builds an electronic folder for each case. It then examines the file to determine if there are warning signs - financial problems, arrests, anything to indicate a person might be susceptible to improper influence - or if it's a clean case. Human investigators take closer looks at the tough cases. Intelligent agents are semiautonomous, proactive and adaptive software systems that can act on a user's behalf. Give an intelligent agent a goal - help an ambassador pick a safe evacuation route following a terrorist attack in a foreign country - and it creates the best plan after gathering weather information, news reports, airline schedules, road information and police reports.

The military, intelligence and law enforcement communities see software agents as a tool for dealing with the daunting task of having to retrieve and monitor huge amounts of data in ongoing investigations or to prevent potential problems (from terrorist activity to insider trading).

Such agents can also help investigators identify unusual patterns of activity, says Henry Lieberman, research scientist and leader of the Software Agents Group at the MIT Media Lab. "Law enforcement can say to an intelligent agent: Let me know when any person arrived from a sensitive Middle Eastern country that was recently involved in a large bank transfer. Or government agencies like the Securities and Exchange Commission can use them to monitor financial statements for fraud. Maybe they could have caught the whole Enron thing earlier," says Lieberman.

Government agencies say they want to use the agents, but most are still defining how exactly. Steve Cooper, CIO of the Office for Homeland Security, says intelligent agents will be part of the US government's data-sharing strategy. NASA is exploring using fix agents to help boost mission-launch rates. Oak Ridge National Laboratory researchers are studying agents for use in military command and control centres.

But the issue of trust may deter their widespread adoption in business. "People just aren't used to using these kinds of things yet," says Lieberman. "When you first start using one of these agents, you have to watch it closely to make sure it's doing what you want. But performance improves over time. And the agent just makes a proposal. Then it's up to you."