POLITICIANS beware. Software can scrutinise legislative bills, working out the source of the text. It could allow voters to see who actually determines what goes into bills.
There is a long history of companies and other vested interests influencing legislation by lobbying politicians. So researchers at the University of Chicago’s Data Science for Social Good programme have created the Legislative Influence Detector. This scours the text of US bills, searching for passages that have been cribbed from lobbyists or the legislatures of other states.
“Our hope is that the public can use this to keep the government accountable,” says team member, now a graduate student at the University of Michigan.
To get the real story behind a bill, the software digs through 500,000 state bills, as well as thousands of pieces of text drafted by lobbyist groups that were saved into a database. An algorithm then calculates the top 100 documents most relevant to the bill in question before examining each one more closely, searching for passages the two have in common.
What this reveals can be telling, says Katz-Samuels. The software can turn up lines of text originally written by activists and special interest groups. Or it might find that the bill borrows largely from laws already in place elsewhere, giving concerned citizens the chance to explore how the policy worked out there.
“The software can turn up lines of bill text originally written by activists and special interest groups”
In one example, the researchers ran the detector on Wisconsin Senate Bill 179, a bill passed in June that bans non-emergency abortions after 20 weeks. The team found that the bill’s passage on fetal development – which pinpoints 20 weeks as the point when fetuses could feel pain – had appeared in one form or another in a handful of other proposed abortion bills across the country. Further digging traced the text back to an anti-abortion organisation’s website, with the group appearing to be the original author of the text.
“Since these bills are tangled up in policy schemes, we need better methods for untangling them,” says David Smith at Northeastern University in Boston, who has studied how text from failed bills gets reused in later legislation by US Congress.
The detector is the latest in athat try to use computers to . At the University of Texas at Arlington, computer scientist Chengkai Li is building a system to fact-check the statements of politicians in real time. , as he’s calling it, will study work done by human fact-checkers and, with the help of machine learning, start automating some of that process.
Li envisions a final platform that can scan through the transcript of a speech or presidential debate, picking out the lines that we already know to be true or false so journalists can work on checking more complicated claims.
(Image: Sally Ryan/The New York Times/Redux/Eyevine)
This article appeared in print under the headline “Law influencers unmasked”
This entry passed through the Full-Text RSS service – if this is your content and you’re reading it on someone else’s site, please read the FAQ at fivefilters.org/content-only/faq.php#publishers.