Technology Meets Behavioral Science for the Social Good: Q&A with Marta Milkowska
Stanford MBA student Marta Milkowska has followed an unusual path that established her credentials as one of her generation’s leading innovators in economic development. Her experience includes work at the United Nations and the Polish Embassy in Kenya, at the World Bank as a private sector development specialist, as a member of the team that launched the World Bank Innovation Lab, and as a serial social entrepreneur.
Early in her career, Milkowska became convinced of the need to adopt cutting-edge technology to solve development problems. But she came to realize that advanced technology by itself was not enough. To be effective, she says, technology has to be integrated into a complex social and cultural context and be used in ways that take human behavior into account.
She has founded several social ventures, including Dignify, a mobile platform linking refugees to digital micro-work, and yoni.io, which helps women suffering from pelvic pain take control of their sexual health. And working with HealthTech startup Dimagi, Milkowska helped develop a machine-learning risk prediction tool for HIV and TB patients in South Africa and Lesotho. With the UK’s BI Ventures, she also designed a mobile app that uses behavioral insights to help young people start saving.
Golub Capital Social Impact Lab faculty director Susan Athey met with Milkowska to talk about behavioral change and technology in designing development strategies.
You’ve carved out a fascinating niche for yourself. What are you doing now?
I’m getting a master’s in business administration at Stanford and a dual degree in public administration at the Harvard Kennedy School. My focus is on the intersection of technology and behavioral science used for social good. While working in international development across Africa and Asia, I realized that we focus primarily on access and infrastructure development and much less on behavior, the things that enable people to actually use our services or products. At the same time, the focus in emerging economies has been shifting from mostly building bridges and roads to developing effective service delivery (healthcare, education, job creation). With that shift, we have to address those soft needs in order for our work to succeed.
What does that mean in practice?
I will give an example of a project. Open defecation is a major healthcare and environmental problem in India, affecting 40 percent of its population. To solve that problem in a small village, a few dozen toilets were constructed. The toilets worked well, but the villagers were not using them.
Why didn’t the villagers make the switch? When researchers investigated, they discovered that the morning open defecation practice was the only social activity for the local women, who otherwise spent all their time under the guardianship of their husbands. The morning walk to the forest was their mini-vacation. They laughed, gossiped, shared information, without any constraint from the larger community. Women knew about the health benefits of using the toilets. But the social need for connection was much deeper.
In my experience, the international development actors too often focus solely on providing access and developing infrastructure. But in today’s world, faced by increasingly complex challenges and deeply dependent on people’s behavior, getting behavior change right is essential for development.
So are there generalizable lessons for intervention design?
We have a few of them already. Interventions must be designed to take into account the social and cultural worlds people live in. One of the important aspects to consider is who is making decisions at the household level. When Dimagi was helping pregnant women in India get more sleep, change their diet, or sign up for maternity counseling and vaccinations, success was limited until they got husbands, grandmothers, and mothers-in-law involved. Pregnant women had very limited ability to make decisions themselves. Dimagi had a breakthrough when their community health workers started showing women simple videos on their phones. Because technology (even simple phones) is fun and aspirational, the whole social circle came up. The husbands came, the mothers-in-law came — everybody watched the video. That helped change the behavior of a woman. But the whole household had to be involved.
Another way of thinking about generalizable lessons that we talked about at the World Bank was the replication of a process versus the replication of an exact intervention. In terms of a process for behavioral intervention design, we found several principles were important.
Most importantly, don’t try to fix the person you design for. Fix the context. Contextualize an intervention to a particular social and cultural environment, as Dimagi did.
We development specialists need to examine our mindset. We often think in terms of fixing a person; that is, changing their behavior by educating them about good health or financial practices, or things like that. That’s based on the very wrong assumption that we’re wise and they’re ignorant. We can’t think about development in terms of educating the unenlightened. That’s disrespectful and it will be recognized as such. We need to listen and assume there’s a good reason why people do what they do. Instead of fixing people, let’s fix the context.
Can you give an example of fixing the context?
Two examples of interventions that focused on fixing the context come to mind. One decreased the “hassle factor.” The second removed cultural barriers to an intervention.
Hassle factors are roadblocks; small inconveniences that may require time or effort and keep us from taking an action. We often don’t do things simply because they’re a pain. We may choose a less healthy meal in a cafeteria simply in order to avoid standing in a longer line. People may forgo applying for food stamps because filling in numerous forms is overwhelming and time-consuming. Patients often drop out of tuberculosis treatment in order to avoid travel to a distant clinic to pick up free TB medicines. All of these hassles make it more likely that we’ll just say, “I’ll do it tomorrow” and less likely that we’ll actually ever do something.
Fia, a social enterprise improving financial inclusion in India, successfully adjusted its business model to decrease the hassle factor for its low-income customers. Unlike other banks, Fia extended its operating hours from 8 a.m. to 8 p.m., allowing working mothers to come to the bank in between their work and home duties. They also included special assistants to help illiterate customers through the process and eliminated several steps that made low-income customers embarrassed and unwelcomed. What Fia does is essentially reduce the upfront cost of the future-oriented behavior (increasing savings). Other programs showed similar results. Providing kids with free school uniforms increased school enrollment in Kenya by more than 6 percentage points. Offering U.S. students assistance with their applications for federally funded college student aid has increased the college enrollment by 24 percent.
India’s GVK Emergency Management and Response Institute is an example of an organization that fixed the context by successfully removing cultural barriers. EMRI provides ambulance service and a 911-type emergency response in places where the state doesn’t provide it. Given a high rate of complications with at-home deliveries in rural India, one of GVK’s objectives is to increase the number of in-hospital deliveries. But simply providing a 911 number wasn’t enough to make people call. While working with EMRI, one of the barriers we recognized was that an average husband didn’t want to leave his pregnant wife in an ambulance alone with a male nurse. Hence, EMRI adjusted its model and included female health workers in its ambulances. Understanding and responding to the social and cultural dynamic was key to a successful program.
Are there techniques we can learn from behavioral science about making interventions successful?
Yes. We observed an increasing number of insights from behavioral science research to learn from. One of the most famous examples is using defaults in intervention design. Several programs in the United States have successfully increased employees’ 401(k) savings by defaulting new employees into the savings plans. Instead of a standard opt-in, new employees were automatically enrolled in the 401(k) plan, which increased the participation rates from 59 to 95 percent. The program also successfully increased the contribution rate.
Lotteries are another efficient and cost-effective way of getting people involved. (And they like to play.) Lotteries have successfully increased employee vaccination rates and health risk assessment completion rates (by 20 percent). They were also more effective than a financial, or in-kind award of the same value. (Most people simply overstate their probability of winning.)
Using social norming is one of my favorite techniques. OPower, for example, managed to decrease electricity consumption by sending its customers home energy reports benchmarking their use to those of their neighbors. The new design saved the utility customers an estimated $1 billion. Accountability has been successfully used by many programs ranging from reducing HIV treatment default rates to decreasing college dropout rates in America.
So informing people about the benefits of development isn’t always enough to change behavior?
That’s right. As we mentioned before, that’s one of the most important messages (that we often get wrong). Simply educating people on the benefits of the new behavior is rarely enough. If it was, all of us would probably go to the gym much more often. People (whether low income or not) often understand why they should change their behavior. And they have a serious reason why it’s hard or impossible. We already talked about three obstacles to effective behavior change programs — competing social priorities, the hassle factors, and inadequate cultural or social intervention design.
What other obstacles have you found that inhibit people from changing behavior?
Another one to consider is the context of scarcity — a state in which people experience substantial stress caused by a lack of something important: money, time, attention, etc. Scarcity decreases our cognitive abilities. Most of us encounter that. For example, during the exam or recruiting periods, most GSB students’ ability to empathize with others or take on new projects is likely decreased because our attention, time, and sleep are all a scarce commodity.
What does this mean when we design for the low-income populations? Economists realized that poverty (the scarcity of financial means) decreases cognitive abilities of low-income populations by an equivalent of a 13-point deficit in IQ. Poverty imposes a heavy attentional “tax” that prevents people from evaluating and choosing new opportunities. This decrease, however, can be reversed when the stress discontinues! And we must take it into account while designing the timing of our intervention. To illustrate, if we want to help farmers sign up for an insurance program, we should approach them during their harvest season, when their cognitive abilities are higher than during the winter (scarcity).
How is technology important in solving these problems?
Technology can bring enormous benefits — increased access for marginalized communities, decreased cost, the opportunity to scale intervention much more rapidly. We can leverage technology to gather and analyze data, which in turn can lead to better program design and decision making. And this is the time, given the wide improvements in connectivity around the world. There are 5.7 billion unique mobile subscribers around the world. That’s opening a world of possibilities in development work!
Here’s a great example: 400 million people still don’t have access to essential health services. One of the social ventures World Bank invested in, Dimagi, uses mobile technology to help frontline workers deliver basic health services to the most marginalized communities. Dimagi provides an open-source software application with mobile and cloud infrastructure that runs on inexpensive mobile phones. It uses audio, video, imagery, SMS texting, and data tracking, to empower community health workers to standardize service delivery, improve counseling techniques and patient coordination, and collect real-time data for performance monitoring.
And we’ve seen incredible results. In India, Dimagi increased the share of women who visited local health-care providers by 73 percent! And 58 percent more women took recommended medicines. This is huge in any context. And we were targeting vulnerable populations in the scarcity context. Dimagi and many other digital healthcare providers are present in over 60 countries, so these interventions are also massively scalable.
Do you have examples from other sectors beyond healthcare?
So many! In education, Bridge International Academies operates low-cost schools for kids from slum areas in several countries in Africa. The teachers are provided with tablets with step-by-step teaching guides to help them run lessons. Technology is just a guide, and classes are highly interactive. But these guides allowed Bridge to standardize the quality, lower the costs, and improve kids’ educational outcomes. Based on early evaluations in Kenya, kids gained on average 24 to 60 days of learning through this model.
In agriculture, SMS technology can increase agricultural yields by providing farmers with information about weather or crop prices. It can also increase farmers’ production capacity. A West African social enterprise and a Stanford GSB Impact Fund investee, Hello Tractor, is basically an Uber for tractors, with an integrated mobile payments system.
How much of a difference does it make to be able to make payments using mobile technology?
Enormous. I will name just a few. First, it includes previously financially excluded populations into the economy. I used to live in a Masai village in Southern Kenya. Most of the Masai people were un-banked — they were simply unable to open a bank account due to the fact that most of them didn’t own an ID, were illiterate, and wouldn’t even be able to get to the city. But all of them owned a mobile phone. In 2007, with the rise of mobile money, a cashless, mobile-based payment system in Kenya, I could observe their shift: They could finally perform economic transactions, start a business, save. The World Economic Forum estimates that almost 2 billion people, and 40 percent of the world’s adult population, lack a bank account. As a response, in 2016 there were over 270 mobile services in over 90 countries.
Second, it allows those newly included populations to access basic services. Off-grid startups, for example, are already providing electricity to 50 million people through integrated mobile payments.
Finally, there are behavioral benefits: Mobile payments decrease the hassle factor!
Are there risks in going overboard with technology?
We can’t view technology as a silver bullet. I know, a crazy thing to say in Silicon Valley. But at the end of the day, technology is simply an enabler — the infrastructure, sometimes as misguided as the toilets built in our Indian village. We must think about the behavior as well. Let me give you an example.
Operation Asha, an Indian social enterprise, aimed to use technology to revolutionize tuberculosis treatment. Although completely curable, TB kills one person every 2 minutes. One of the biggest problems was that patients’ adherence to a 6-month painful treatment was increasing cases of multi-drug-resistant TB. Op Asha had a solution: a low-cost biometric technology to manage TB treatment in slums. A simple fingerprint reader and a tablet would allow Op Asha to track TB patients’ compliance with their treatment regimens.
Op Asha opened its first TB treatment centers, soon to realize that technology, although crucial, wasn’t enough to attract patients to the centers. Many patients resisted coming at all; others would routinely walk 2–3 kilometers rather than use centers much closer to their homes. Why? Because of the deep social stigma TB has in India. Fear of ostracism compels many patients to hide their symptoms and not pursue proper treatment.
Based on this discovery, Op Asha revolutionized its business model and community outreach strategy. It rebranded the TB treatment centers into un-branded health centers and placed them within the heart of underserved communities: in temples, homes, and shops. Local champions were hired to personally follow up with patients; opening hours adjusted. Clinics were open early morning and late night, so that privacy is maintained.
Today, Op Asha has treated over 50,000 people, with incredible health outcomes.
Skillfully leveraging technology and behavior change — that’s where the magic happens.
What’s the place in all this for low-tech or non-tech solutions?
That’s a great question. There are some incredibly exciting high-tech solutions coming up. (Zipline uses drone technology to deliver medical essentials to rural health facilities in Rwanda. AIME worked with the governments in Brazil and the Philippines to predict the spread of the Zika or dengue viruses.)
But often, especially in low-income settings, basic mobile phone allows for scalable impact (like with mobile money or digital health). At the bank, we often talked about reframing innovation — often it doesn’t have to be “new to the world.” In low-income settings it’s often “new to the context.” Let’s be practical and leverage what people already have access to. Very often it’s low-tech.
These projects are generating tremendous quantities of data. How can it be used in a development context?
This is one of our biggest opportunities. I’ll just give two examples. First, data can help us understand where people are. Spending patterns on mobile phone, for example, can provide a proxy indicator for income levels. Second, prediction is a huge opportunity. Mapping movements of mobile phones during an Ebola outbreak can help forecast the spread of the disease and guide the governments on where to deploy the very limited resources available to stop the outbreak. Digital healthcare providers, such as Dimagi, are looking at predicting which TB patients have a higher risk to default from treatment. Once you know that, you can design a behavioral intervention to help them stay on track.
Do these uses of technology raise larger questions?
Definitely. First, there are many ethical concerns beyond privacy. For example, Lumiata, a U.S.-based startup, uses predictive analytics to identify people at risk of conditions like kidney disease, diabetes, and congestive heart failure. The question is: What happens to the person who we predict as being high risk? Will the insurer stop covering that person? Or use targeting pricing models based on the risk profiles? Second, what happens if the prediction model fails and we trust it too much? What happens if we mark someone as low risk of TB defaulting but they actually have high risk? When playing with people’s lives these questions have to be very carefully considered.
Lastly, there are capacity concerns. We still don’t have enough good, clean data from low-income settings. Further, data is only as useful as it can be operationalized. And many operational managers of the in-country programs lack data literacy that would enable the breakthroughs. As we design those interventions, we must help increase their technological capacity.
What can you tell us about your latest venture, yoni.io?
We are providing digital physical therapy for women who suffer from pelvic floor dysfunction — a massively under-treated medical condition that affects women’s ability to have intercourse, use tampons, undergo medical exams, get pregnant, etc. An estimated 20 million women in America alone suffer from it chronically. Our mobile app and online training shows women pre-selected daily exercises to alleviate pain.
That sounds like a great example of using technology to bring information to people in a scalable way. Best of luck on the venture!
Thank you.
— —
Learn more about the Golub Capital Social Impact Lab at Stanford Graduate School of Business.
Follow us @GSBsiLab.
Learn more about Marta Milkowska.
With writing help from Sam Zuckerman.
