How can data science contribute to social good? A curated link list
There is a heated debate raging on whether AI and data science do more to help or harm society.
The optimists see the potential for these practices to transform how we tackle all kinds of social good problems, “from battling hunger to advocating for child well-being and more”. They believe that the use of data technologies can foster innovative solutions to social issues and help social sector organizations achieve more impact with limited resources.
The pessimists worry that existing inequalities will be cemented and even made worse by new data technologies. They argue that collecting data on marginalized people exposes them to greater risks. Algorithmic decision-making on access to social resources and in legal decisions has reproduced old biases that discriminate against marginalized groups.
I consider myself a pragmatist. I believe that to harvest the potential of data technologies for a better society, we still need to figure out many questions. I am excited to dig deeper. To dig deeper into what factors make data science projects help with social issues. And to dig deeper into how we can work around the dangers lurking in that practice.
Here are some links I found valuable on this topic. If you have a perspective or resource to add, please suggest them via e-mail or Twitter DM. I’d love to hear your own perspectives, ideas and feedback on this list!
Jake Porvay started the “Data4Good” movement. He founded DataKind, the first and largest organization bringing together data scientists and nonprofits to work on social good problems. His TedX talk from 2012 is required watching to appreciate the optimistic vision in this area. Follow up his talk with reading this recent update on DataKind’s strategy to see how their vision of social impact through data has evolved.
The McKinsey Global Institute has released a white paper titled “Notes from the AI Frontier - Applying AI for Social Good”. It is one of the most extensive collection of case studies of successful applications of “AI for Good”. The case studies are organized by the different challenges of the UN Sustainable Development Goals.
In this German blog post Daniel Kirsch, founder of “Data Science for Social Good Berlin”, explains his reasons for starting the nonprofit. He aims to make nonprofit work more evidence-based and impactful using data science. Similar thoughts in English can be found in the more recent book “Lean Impact” by Ann-Mei Chang.
Stanford researcher Lucy Bernholz argues that data science applied to social issues will increase the risk of data abuses and biased algorithmic decision-making. "There's no excuse for continuing to act like inserting software into a broken system will fix the system, it's more likely to break it even further.“ She details her concerns with specific examples in another blog post.
Another researcher, Tim Urwin of the University of London, evaluated the impact of investments in broadband internet in aid programs. He finds that richer developing countries have benefitted more than poorer countries. A similar impact could happen for data science technologies. More digitally empowered and privileged groups might benefit more, increasing social inequality further.
Natasha Iskander argues in a HBR blog post that approaches to social issues lead by experts privilege the perspective of the experts over the needs of the affected groups. Her critique is about design thinking but the same concerns apply to data science. The designer (or data scientist) is the one making the final decisions on how the developed solution will work. That puts the expert into a position of power that reinforces existing power imbalances and social inequalities.
Kentaro Toyama has been on all sides of this debate, from optimist to pessimist to pragmatist. In his book “Geek Heresy”, he talks about his experience implementing and evaluating “Technology for Good” projects. He has seen many of them fail, and some succeed. He recommends us to focus on building the skills and networks of organizations implementing the technological solutions, not the solutions itself.
We have seen Lucy Bernholz in the pessimist section, but that is not where she stops. In this blog post, her research lab outlines their research agenda on the risks and opportunities of “digital civil society”, and ways for civil society to deal with these. For example, they released a “Digital Impact Toolkit” that guides nonprofits on how to manage and govern their data.
#MoreThanCode have surveyed the space of “Technology for Justice and Equity” projects. In their research report, they summarize five practices used by successful projects. Their focus is on how to make these projects more inclusive and representative of the groups they are trying to address.
youvo is a German online platform for designers to volunteer their skills pro bono. In their blog (German), they discuss principles of how nonprofits can use new digital technologies in their work. And what policy makers and donors can do to support that.