Applying AI to Societal Challenges in Cities & Communities: Highlights from DataKind’s Virtual DataDive With Microsoft

November 7, 2019

We can’t believe it’s been a month since our first-ever virtual DataDive. We’re still feeling the energy and enthusiasm from that day. More than 100 DataKind volunteers from around the world, hailing from five continents and eight time zones, donated their time and talent on Sunday, October 6, 2019, to eight amazing organizations working in important issues facing metropolitan areas.

This marathon-style event was so much bigger than one day. It was a showcase of the “AI Accelerator” program, a partnership between DataKind and Microsoft, in which we’re providing pro bono expertise and assistance to a cohort of organizations seeking to apply AI to societal challenges in urban communities. For too many organizations, the positive impacts of using AI remain out of reach, so this DataDive was just one initiative in a larger effort to help organizations build the capacity to identify, design, and use AI responsibly and effectively. 

DataDives show organizations how data science and AI can advance their missions and show volunteers just how powerful their skills can be. Much like the weeks leading up to any DataDive, where there’s a ton of preparation to scope the projects and prep the data, there will be months of impact that ripple out after this event. In partnership with Microsoft, we’ll continue to work with the participants, and the cohort of organizations will finish with actionable project plans for taking their AI journeys forward. Read on for some project highlights and to hear from some of the DataDive participants! 

  • The Teach for America project made significant strides in developing predictive models that can help the organization to understand which of their member teachers are facing challenges that might lead to premature departure from the program. Once implemented, these models should directly improve Teach for America’s core training and support processes serving teachers and students across the US.
  • The Trust for Public Land project trained deep learning models on satellite imagery of city parks in order to automatically identify amenities, such as baseball diamonds, an initial step toward fulfilling The Trust for Public Land’s goal of generating a complete catalogue of park amenities for every city in the country.
  • The Code for Miami project developed and validated highly predictive models of permit processing time, and prototyped a web API interface that would allow for the predictions of processing times for new construction projects to be generated and included in various city services in order to inform programs and policies tackling housing affordability and accessibility challenges.

"For us, what was important was getting real data science practitioners. During the DataDive, having 10+ folks working with the City of Miami open data to help tackle affordable housing was a rewarding experience. I learned a lot and got a lot of insights on how to prep data for ML/AI work." -- Gregory Johnson, Code for Miami

"I really enjoyed working with a team of very creative and highly skilled volunteers from around the world. They came up with a multi-pronged approach to analyze a difficult, semi-structured data set for Teach for America that was quite thoughtful in a very short period of time." -- Neal Fultz, Data Ambassador

"The collaboration of the partner organizations, ambassadors, and volunteers is an amazing demonstration of how a range of expertise can together enable AI capabilities - from computer vision, to natural language processing, to predictive models -- to advance initiatives in education, sustainability, housing, and more." -- Elizabeth Grossman, Microsoft

This was our first-ever virtual DataDive, and the success relied on a range of innovative technical solutions. We’re eager to do this again, and we’re looking forward to exploring other opportunities where we can virtually bring together our global network of data do-gooders. And as you know, it takes more than innovative solutions and data to change the world, it takes people. The right mix of people coming together in the right ways, sharing solutions, iterating, and building long-term relationships. Our global network (aka YOU!) is what makes our work possible and allows us to make a global impact.

Stay tuned for more from us on the “AI Accelerator” program and the cross-cohort learnings and models for how to expand the use and impact of AI by community-focused organizations. Huge thanks to all of the volunteers, project partners, and Microsoft who made this event possible!

Image above shows construction and development permits for the city of Miami, FL. The Code for Miami project developed machine learning models that were able to predict how long a given permit was likely to remain in review by the city.