Creating a Systems Change Approach for Data Science & AI Solutions

By Jake Porway, Founder & Executive Director, DataKind


At DataKind, we’re on the precipice of a significant strategic shift. Over the last seven years, our intrepid community of data scientists has generously donated their time and skills to complete over 300 projects with social good partners through our DataCorps and DataDive programs. The results have been impressive, ranging from saving nonprofits millions of dollars so they can provide more clean water to transforming government policy to crack down on corruption. These projects are just a step toward our vision of a world in which everyone working to improve society can use data science and AI ethically and capably.


We asked ourselves, what would it take to get there? 300 more projects? 3,000? 30,000? There was no number of individual DataKind projects that could make a significant impact. If the world is relying on individual volunteer projects to tackle the most pressing humanitarian challenges, from DataKind or any other organization, then something is severely broken.  


The good news is that we’re seeing a shift in thinking in the social sector that opens a pathway for this greater change. Increasingly foundations, governments, and companies are ready to engage in using data science and AI for social impact, and they can put major resources behind that. What they struggle with, however, is figuring out exactly what the opportunity is. How could AI help reduce infant mortality? What data science opportunities are there for supporting NGOs working for a more equitable justice system? Right now, there are too few entities that are resourced to marry the needs of different issue areas with their data science opportunities. 


We’re therefore moving from a project-based model to a practice-based model, where we focus on portfolios of data science projects connected around common themes or issue areas, from one-to-one relationships across every thematic area to one-to-many relationships in a few thematic areas. That’s why earlier this year, with the generous support from The Rockefeller Foundation and the Mastercard Center for Inclusive Growth, we announced that we’re looking to identify data science opportunities to advance entire sectors with the launch of DataKind “Impact Practices”.


So what is an Impact Practice exactly? And what can you expect? Read on! 


Defining an “Impact Practice”


Impact Practices are data-rich issue areas where multiple organizations have overlapping needs. Organizations could have many types of overlapping needs, like:


  • A common outcome, such as reducing the time to detect disease outbreaks. 
  • A common technology, such as drone and satellite imagery. 
  • A common dataset or data system, such as the US Homeless Management Information System


By identifying similar opportunities, shared pain-points across several social change organizations, or an overall theme under which many organizations work, DataKind can target a common data science capacity boost with the aim of generating solutions for multiple partners and potentially sector-wide system change. The success of an Impact Practice is that we’ve gained and shared enough hands-on knowledge about the datasets available, the appropriate ‘data science-able’ challenges, and the resourcing needed to derisk the space for future development.


Let’s take an overly simplistic example. A key challenge in disease prevention is reducing the time to detect disease outbreaks. If we were working with a single disease prevention organization, we’d be able to dig into their goals, their workflow, their available data, and then run a traditional DataCorps project to test data innovations that could reduce their time to detect disease, but that would just support one organization. What if we wanted to reduce the time to detect disease across many organizations or even countries? Understanding the design of that system would require working with multiple participants with that same challenge to understand their workflows, understand what data is available, and test real prototypes in the field to understand what’s really possible. Thus, an Impact Practice would consist of many projects across organizations that could benefit directly from reduced time-to-detect and who may be participating in the same, or similar, digital systems. If successful, this portfolio would result in learnings, prototypes, and “validated scopes” of the developments needed to advance the space with data science and predictive analytics.


It’s worth noting that we first tested this model with the support of Microsoft when we ran our first Labs projects. You can learn more about our work with Vision Zero cities to see an example of a proto Impact Practice.


Identifying Impact Practices


We’re looking to identify opportunities with partner organizations that help address recurring needs across the sector in order to develop prototypes which will generate sector-wide impact. Developing an Impact Practice that has the level of impact we seek will require significant testing up front. We don’t yet know what attributes indicate, early on, which groups are most likely to have shared needs. We’ll do our best to select into the right set of groups by considering the following factors:


  • Is this a pressing humanitarian issue?
  • Are there quantifiable measures of success?
  • Are there rich data sets or digital technologies in use in the space?
  • Are there funding institutions, foundations, or corporations that are committed to finding scalable solutions and applying cutting-edge technologies?
  • Are there organizations open to experimentation with new technology?
  • Do we have past project experience or a group of volunteers with expertise to help answer these questions? 


Addressing Inclusive & Responsible Design


Like our individual projects, our Impact Practices must meet our highest standards of responsible and thoughtful design. If not designed properly, data science interventions can be ineffective or, worse, harmful. That goes doubly when it’s being used to support nonprofits and governments charged with caring for the most vulnerable populations. For those reasons, we’ll continue to uphold our project design principles in Impact Practices:


  • Start with the problem, not the data: The technology means nothing without a clear sense of what problem it’s solving, for whom, and how. 
  • Local context is table stakes: From understanding the social issue at hand to working with the NGOs in that field to including the community’s needs, local context is critical. That’s why we partner with respected institutions in each space, consult with advisors across our network, and facilitate conversations and design between local technologists and community members. When I say “we”, I don’t just mean us sitting here in NYC with our points of view. I mean the entirety of the 20,000 DataKind volunteers across all continents. No matter who’s working on the project, we make sure we’ve got representation at the table throughout, and we don’t go where we’re not invited.
  • Data science is our hammer, but not everything’s a nail: We care about effective social outcomes over “doing projects”, so we’re  just as likely to identify where data science is not useful as where it is. We’re also not naive about the core infrastructural data needs at large in many issue areas – most folks don’t have high quality, or any, data. However, we always push to find the area where there’s enough of a digital infrastructure in place for us to demonstrate the “art of the possible”.


Learning Out Loud


A key element of every Impact Practice will be “learning out loud”. We’ll be sharing our learnings, so others interested in using data to help organizations can build on our work. Doing work in cohort allows for DataCorps teams to share insights and approaches as they develop solutions. By facilitating cross-team learnings, teams can approach each phase presentation together for a share-out and critique. The Impact Practice approach allows stakeholder, partner, and participant organizations to have transparency into the process and development not just of an individual organization’s solution, but of solutions across the space. While DataKind will be pulling together synthesized learnings at key stages of the process, our intent is to encourage these communications as an opportunity for development of organic learnings and collaborations across the cohort.  


Introducing Our First Impact Practice


There are over one billion people on the planet who lack access to healthcare. People may live too far from a clinic to get service, may lack the money to afford care, or may simply not know where to go for help. The Community Health Worker (CHW) model has been a breakthrough innovation for delivering last-mile care. CHWs is an umbrella term for trusted frontline public health workers that serve the-hardest-to-reach communities. Acting as an intermediary between health services and the residents, the CHWs provide healthcare services, encourage their communities to adopt healthy behaviors, and much more. If we want the impact of CHWs to grow exponentially though – in order to account for population growth – we must combine their on-the-ground frontline care with the best cutting-edge skills and techniques available. That’s why, earlier this year, we began our work to improve health outcomes for the hardest-to-reach communities of Africa by using data to amplify the impact of CHWs. You’ll very soon be hearing much more about the launch of our first cohort of projects for DataKind’s Community Health Worker Impact Practice


Unlocking the Potential


To date, DataKind’s activities have answered the question: How could data science be most leveraged to serve this social organization? Yet, the world’s problems will not be solved on the backs of hundreds of individual volunteer projects. We now want to help answer the question: How could data science be most leveraged to serve this social cause?


The aim of our work then is to move the needle on the world’s toughest challenges, not just help nonprofits become data savvy nor just help data scientists find ways to give back. We’re ready to take that next step in creating a more equitable and prosperous world with data science, and we hope that you’ll take that step with us. Volunteer to join us for the DataKind Community Health Worker Impact Practice when the call goes out, reach out to us if you run a major initiative that you think could take advantage of data science, or apply to work with us for one of the many roles we’re hiring for in this next phase of our journey. Data is about all of us, so it requires all of us to be involved to be used well. See you out there! 



Quick Links


1 thought on “Creating a Systems Change Approach for Data Science & AI Solutions”

  1. Pingback: Uniting Data Science and Social Good, with DataKind co-founder Jake Porway - Tech Matters

Comments are closed.

Scroll to Top