This initiative focuses on creating and assessing new methods that harness data to positively impact society. Selected projects include:

• Developing interpretable statistical models to aid judicial pretrial release decisions.
• Implementing predictive machine learning techniques to allocate beneficial services to children and families with high need, and to avoid burdening low-need families with unnecessary services.
• Improving the reuse of open data and measuring the demand for open government data.

Traffic stops are one of the primary ways in which the public interacts with law enforcement, yet there is little easily accessible information on the nature of the practice. In particular, though there is enormous public concern regarding racial profiling in such interactions, a lack of data has made it difficult to rigorously investigate this issue. To help individuals and communities understand police practices, we are collecting, cleaning, releasing, and analyzing over 100 million highway patrol stops conducted over the last several years across the United States, one of the most comprehensive datasets of police interactions with the public ever to be compiled and released. These records -- which we are obtaining through a series of public records requests filed with each state -- often provide detailed information about each stop, including the age, race and sex of the driver, and the outcome of the stop (e.g., whether a search was conducted or a citation was issued). A key challenge in compiling these records is organizing the myriad formats used by each state into a single, coherent dataset that is amenable to analysis, exploration, and visualization. Funded by the John S. and James L. Knight Foundation's News Challenge.