When we talk about government being more “open” and “transparent,” what information are we discussing? In a digital environment it is possible, and even essential, that we think beyond PDF files and “publications.” For example, the advent of data.gov has made raw data more of an explicit part of public information dissemination than it ever has been before.
In a survey of federal managers earlier this year, NextGov reported a number of roadblocks to transparency. While managers agreed that agencies should post raw data, they balked at doing more than they already have done. They expressed concerns about computer security, and “the loss of control over their agency’s message” and complained that they do not have adequate resources to make more data accessible.
In an article just yesterday (The Risks of Open Government, NextGov, By Allan Holmes, 09/14/2009), that same survey revealed that, while almost 90 percent said transparency involved posting facts and figures that agencies collect, “they are less interested in disclosing how policies are formed, such as providing minutes of meetings, and who were included in meetings and as part of the decisions.”
Although we have heard discussions of these issues before, there is another kind of government information that is rarely discussed: software. NTIS has offered government produced software for years, but the kind of software I’m concerned about is that software that is used to make decisions. The algorithms embedded in software document how decisions are made. If we do not have access to that information, we cannot be sure decisions are being made well.
It is hard to know what we do not have because, as far as I know, there is no inventory of this kind of resource. So, it is particularly revealing to see real world examples of decision making software when they do come to light.
Here is an example that is self-evidently important:
- Eye on the Bailout: The Secret Test That Ensures Lenders Win on Loan Mods, by Alexandra Andrews and Emily Witt, ProPublica
The Net Present Value test is a complex computer model used by loan servicers to determine whether a homeowner qualifies for the federal loan modification program. The test compares two scenarios – modification and foreclosure – and determines which would be more profitable for the lender. If it’s foreclosure, the lender has no obligation to modify the loan. But the model is a black box. What goes in isn’t entirely clear, and what comes out isn’t always reliable.
The Treasury Department has refused to release the exact formula for the NPV model…
Information like this should definitely be made available for examination and preservation.