Nieman Foundation at Harvard
Holding algorithms (and the people behind them) accountable is still tricky, but doable
ABOUT                    SUBSCRIBE
March 13, 2012, 1 p.m.
Screen shot of Virginia Decoded

Virginia Decoded is the prettiest state code you’ve ever seen

Knight News Challenge winner Waldo Jaquith says a state code should be designed like a social network.

Screen shot of Virginia Decoded

Developer Waldo Jaquith has launched a public beta of Virginia Decoded, an astonishingly user-friendly presentation of his home state’s code. It’s the first iteration of Jaquith’s open-government platform, The State Decoded, a 2011 Knight News Challenge winner.

Because Virginia makes its code available as structured data — which puts it ahead of most states, Jaquith says — he can build a website that looks better than the official one. Compare Virginia’s entry for § 24.2-956, which regulates campaign ads in print media, to Jaquith’s implementation. The former offers no design and little context. Jaquith’s page provides understandable history information, a list of attempted amendments, cross-references to other provisions, and inline definitions that pop up for key terms.

“Think of a code as a social network,” Jaquith told me. “Every law is like a person. Every law is connected to every other law via court decisions, legislation, shared definitions, cross-citations, and a dozen other ways.”

When you start thinking about laws that way, the user experience changes dramatically. Jaquith said every official state code website in the United States is, without exception, “wretched,” in effect making public data inaccessible to the public. Jaquith plans to post his source code on GitHub and help implement the State Decoded platform in other states.

Jaquith explained four key features of the website to me in an email:

The first, and best (IMHO), is the built-in glossary. The software scrapes the text of the code to find definitions, which it extracts, figures out the scope of it, and then stores to apply in the relevant sections of the code. (For instance, “person” has 118 separate definitions in the Code of Virginia. The most relevant one must be applied each time that it is used.) The reason why this is so important is that definitions are wildly important in understanding a law. A law might say “No playing loud music on weeknights downtown,” but unless you know the definition of “loud,” “music,” “weeknights,” and “downtown,” it’s of no value whatsoever. The definitions of those four words might be scattered widely, throughout the code. The State Decoded puts them all in one place. You can see an example of definitions in this section of the code about the state’s FOIA law. The defined words have a subtle, dotted underline, designed to make it easy to ignore for people who have no interest in them, but noticeable by folks to whom they’re very important.

The second is the connection back to legislation. In the sidebar of another section of the FOIA law, § 2.2-3711, you can see “Amendment Attempts.” That’s being pulled via a JSON request to Richmond Sunlight (another website of mine), which is listing every bill that cites § 2.2-3711. That allows people to see how laws have built up over time, and what sort of interest that there is, at least among legislators, in amending it.

The third is the connection to other parts of the code. Keeping open § 2.2-3711, you’ll see “Cross References” farther down in the sidebar. That’s a listing of every other part of the code that cites § 2.2-3711. Sometimes there are connected parts of a code that are far, far apart—in different titles, hundreds of pages away (on paper). This is a first step towards integrating them, breaking down the sometimes artificial barrier of topical divisions and regrouping laws more meaningfully.

And the fourth and final feature that you won’t want to miss is the connection to case law. For any section of the code for which there have been state court decisions that cite that section, those decisions will appear in the sidebar, under “Court Decisions.” (See, for instance, § 18.2-33.)

POSTED     March 13, 2012, 1 p.m.
Join the 45,000 who get the freshest future-of-journalism news in our daily email.
Holding algorithms (and the people behind them) accountable is still tricky, but doable
“We were able to demystify this black box, this algorithm that had very scary connotations, and break it down into what ended up being a very simple linear model.”
Fill in the blanks: What’s still missing from the study of fake news? (A whole lot.)
A big new report from the Hewlett Foundation pulls together existing research on social media, political polarization, and disinformation to show where we still need to know more.
Google announces a $300M ‘Google News Initiative’ (though this isn’t about giving out grants directly to newsrooms, like it does in Europe)
Also: an easier subscription flow, $10 million for media literacy in U.S. high schools, fact-checking efforts in search around health issues, and more.