Papers and Findings
A cross-disciplinary team of researchers has developed an NLP method that can predict judgements in the European Court of Human Rights with 79% accuracy, based on an analysis of case documents. AI to replace judges? Perhaps. More comforting is their conclusion that this finding supports the theory of legal realism, “suggesting that judicial decision-making is significantly affected by the stimulus of the facts.”
100 Stories: The Impact of Open Access. That’s the bombastic title of a forthcoming paper focused on public access to scientific research, aiming to change “how we talk about the impact of open access.” They do indeed present 100 stories, but in short format and not of “impact” per se. The bar for inclusion is unsurprisingly low. An example (in full) from the category of advancing innovation:
“New patents matched against University’s patent portfolio Iowa State University patents have been downloaded over 16,000 times by 275 institutions. 35% of the patent downloads have been from high-profile corporations such as IBM (33), Unilever (11), Dow-Corning (7), Hewlett-Packard (6), and Deere & Co (5).”
Also, it’s rife with depressing word clip art charts.
A survey of how usability impacted the use of Dutch political party websites coined my favorite new term in it’s title: “Return on Interactivity.” ROI, get it. They find that national parties have more interactive sites, as do winning parties.
An online survey of the UK electorate (stratified sampling, n = 1982) suggests that online political activism is more motivated by external rewards and social acceptance than offline political activism. The paper also offers an interesting discussion of how this finding can be used explain NGOs online campaigning strategies.
Meanwhile, a Pew survey of civic engagement in 8 countries (KE, NG, ZA, US, IN, GR, IT, PO, HU) finds citizens frustrated with politics but optimistic about their influence over politics, finds generational divides along means of political action, and finds that people politically active online tend also to be politically active offline.
Community and commentary
A new comparative index enters the scene: the Right to Education Index is launched in collaboration with Open Knowledge International.
IDS announces that it will be studying “adaptiveness” in MAVC’s projects, without saying much about how (though there will be a workshop and a focus on Kenya, phew).
An op-ed in Governing suggests that US city government approaches to citizen engagement are constrained by measureability. My thoughts here.
Andrew Schrock is kickstarting a book on civic tech.
In the methodological weeds
HCR researchers are working to figure out what people really think about privacy norms. To do so they developed a method that conceptualizes privacy around sharing interactions, and codes different characteristics of those transactions. This allows them to distinguish between sharing norms in different types of relationships, and potentially, to use crowd sourced data collection to assess things like the role of perceived power relationships in civic interactions. They build the method on a conceptual integrity framework, then tested and refined it on Amazon Turk, with promising results. What seems most exciting about this method and it’s potential application to other contexts, is how it might be able to uncover communal norms by only bothering individuals with one or two questions, which can cut away a whole host of incentive challenges.
Sunlight Foundation has a couple of fellows working on “measuring the impact of open data policies.” In the first description of their work that I’ve seen, they talk about mashing Sunlight’s database of US cities’ open data policies, with cities’ rankings on the US City Open Data Census. As they note, this gives an imperfect measure, because the coverage doesn’t match up perfectly, and because the census is itself an imperfect measure. But it’s a preliminary method that should be applied more often, also at the national register. There’s enough secondary comparative indices for countries out there to begin comparing how contextual factors influence the implementation of things like open data initiatives. There’s lot’s more to say about that, but this study is most interesting for the methodological issues it raises, and which would be nice to see more carefully described. Their findings are unsurprising: strong open data performance is associated with having an open data policy, with policies that are strict, actionable and established, and with the allocation of roles and responsibilities.
One more shocker: new study finds that city websites suck.
Meanwhile a couple of US researchers tested the impact of transparency on policy deliberations (recall earlier claims regarding input transparency), by assessing a federal body that released approximately 2 decades of deliberation records. Their regressions found that transparency has no impact on policy deliberations, understood as “the use of reasoned argument.”
CIVICUS has released a shiny new Monitor to track civic space around the world. The interactivity has gotten a lot of attention; it’s always fun to compare countries with clicks and colors. The methods for these comparisons are less straightforward, though. According to the methodology page, the Monitor continues in the tradition of the Civil Society Index, to combine the work of existing secondary and primary indices with network inputs, consultations and alerts. It’s essentially a composite expert report, producing subjective indicators in a traffic light format. That’s not a perfect method, but nothing is, and it’s adaptive, and pretty comprehensive at least.
Oxfam had two blogposts up last week on measuring inequality. I can’t possibly sum them up, but I heart them muchly. Favorite quote: “…measures are estimates. They’re never just true.”
Miscellanea & absurdum
- Africans prefer China to western democracies (Afrobarometer )
- Ushahidi has launched a USA Election Monitor.
- Preprint: Gender differences and bias in open source: Pull request acceptance of women versus men
- The Economist crunches data to conclude that higher pay for actors correlates with shooting lots of guns in films
- Someone mapped all the weapons confiscated in US airports in 2015
