This Week in Civic Tech: Indianapolis PD Opens Use-of-Force Data, Machine Learning Leaps Forward

This Week in Civic Tech: Indianapolis PD Opens Use-of-Force Data, Machine Learning Leaps Forward

civic tech

This Week in Civic Tech presents a line up of notable events in the space that connects citizens to government services. Topics cover latest startups, hackathons, open data initiatives and other influencers. Check back each week for updates.

Indianapolis PD Opens Use-of-Force Data

Law enforcement continues to have a tumultuous relationship with transparency: In 2015, the number of officers charged with murder or manslaughter for on-duty shootings and has more than tripled, according to an article in Mother Jones. Perhaps most notable are the ongoing controversies surrounding the death of 17-year-old Laquan McDonald in Chicago and the use-of-force investigation of Freddie Gray, who died in a police car in Baltimore from a spinal cord injury.

To confront this national issue, the Indianapolis Metropolitan Police Department teamed up with the civic tech group Code for America (CfA) to launch Project Comport, an open data portal with data sets that chronicle officer complaints, use of force and officer-involved shootings. With the data freely accessible to the public, the openness aspires to develop public trust through accountability and discourse, a move praised by White House Senior Advisor Denice Ross, who directs the president’s Police Data Initiative.

The data, which also is housed within the city’s open data portal, doesn’t shy away from controversial details either. Statistics on the site show that black residents are tied to 56.8 percent of officer-involved shootings, as opposed to 34.2 percent of white residents. The site also reports demographics of the police officers, who are 82.3 percent white, as opposed to the population of Indiana’s Marion County, which is 57.9 percent white. The city has plans to work with CfA to collaborate on future development of the site and how it might be used in other jurisdictions.

Machine-Learning Leaps Forward

Machine learning hasn’t always met expectations, especially when pitted against its human counterparts. Researchers, however, have engineered an artificial intelligence (AI) solution that not only meets some levels of human learning, but in certain cases, it surpasses it.

Researchers Brendan Lake of New York University, Ruslan Salakhutdinov of the University of Torontoand Joshua Tenenbaum at the Massachusetts Institute of Technology have engineered a piece of AI software that can identify handwritten characters as accurately as a human after seeing only a single instance of the character. The discovery is exceptional considering the most advanced algorithms, well crafted in the architecture of “deep learning,” require thousands of handwritten examples to differentiate one letter from the next.

Put in perspective, the trio said humans typically need only a few examples to not only grasp such visual concepts, but develop rich iterations. In their analysis, published in the journal Science, the researchers said they turned to probabilistic methods to avoid the burdens of heavy data consumption and its sluggish side effects. When confronted with a new or familiar character, the program adapts by generating a unique program to interpret it.

Moving beyond the alphanumeric, the software could be applied within the larger realm of linguistics, enabling it to identify and comprehend new words. If this kind of rapid machine learning were applied in a Google search, for instance, keywords would have a constantly evolving source of synonymous content. In the realm of civic tech, AI like this could eventually become a formidable tool in the fight against paper forms. Whether for accessibility or hard documentation, bureaucratic regulations are unlikely to let written forms fade. With this in mind, the AI could be the premise of a new startup to help process the millions of forms city, state and federal officials collect each year.

Captricity is just one paper processing venture on this front. Under CEO and founder Kuang Chung, the startup uses predictive algorithms and human verification to assist a score of state and federal jurisdictions with data entry. If this new tech is accurate enough to reduce — or eliminate — human help, it could be a strong push toward a more automated government. Expands

The Knight Foundation is expanding coverage of the nonpartisan elections platform with an investment of $100,000. The League of Women Voters announced on Dec. 17 the new funding measure, estimated to provide educational voting resources to more than 1 million people.

The initiative was made in response to poor voter turnout and grouped the U.S. among the lowest voting democracies. The influx of capital will drive development and marketing of  VOTE411’s a shareable app that can be embedded on other websites. The League successfully launched the app in 2014 with a number of media outlets, government agencies, elections officials and nonprofits, which inserted it into their websites.

The app’s growth is expected to broaden the League’s influence to underrepresented voter demographics, with the VOTE411 platform reaching a calculated at 3 million-plus in 2016. Since VOTE411’s launch in 2006, more than 20,000 sites have linked to it and 25 million have used the service for its election information.