What could Watson do with the budget?

The Energy Department’s former science chief mused Wednesday about what IBM’s Jeopardy-winning Watson technology could do with the federal budget.

Steven Koonin was leading a panel discussion at an IBM Capitol Hill briefing on big data analytics when he paused to half-seriously pitch the concept to IBM executives.

“It would be interesting to be able to pose to the machine: How much did the government spend on oil subsidies in Louisiana in 2007,” he said. “That kind of capability is just going to open up government in ways that couldn’t have happened before.”

The idea occurred to Koonin during a 2011 meeting between Energy Department and White House officials when someone asked for a list of all energy subsidies, he told me later. No one could produce it and the idea of an ultimate source for such questions that could spit out an answer on the spot seemed very appealing.

Watson could bring three important things to bear on questions like Koonin’s.

1. Computing power to crunch through thousands of pages of federal budget documents and supporting materials.
2. A facility with language that allows it to understand that Koonin’s talking about tax breaks for big energy companies not sops to the olive oil industry.
3. The almost human ability to know not just what data it’s looking for but where to look, how to interpret that data and which data to trust.

Sorting out the old political bugaboo -- one politician’s oil subsidy is another politician’s small business grant -- would likely have to wait for version 2.0.

As things stand now, policy experts can spend hours coming up with an answer to a question like Koonin’s and often still not be sure everything’s there.

The Obama Administration has made great strides in making more government-generated information automatically available to the public through application programming interfaces and other systems, but it has generally left the problem of gathering insights from that data -- the Watson job -- up to the public.

Watson would certainly be a boon for reporters and analysts -- something akin to the best public affairs and Freedom of Information Act officer you’ve ever met -- at least when it comes to tracking down hard facts.

“Hey, Watson. Y’know how the government is trying to save money by consolidating all those wireless contracts? You wouldn’t happen to be able to pull together all new wireless contracts, organized by agency for the last year, would you?...Oh yeah? So do you have some historical data I could compare that to?... Awesome! Thanks Watson.”

Of course if Watson started coming up with its own questions and stringing a little verbiage around the resulting data we might be in a pickle.