����vlog����

As a quantitative researcher and lecturer of public policy at Harvard Kennedy School, CID Faculty Affiliate Teddy Svoronos often asks: To what extent is data reliable? How do we know if a policy is really working? When do we know enough to act?

“When you teach, you realize knowledge acquisition is an intermediate step,” says Svoronos, who also serves as faculty director of Harvard’s (EPoD). “The end goal is to have people change their perspectives, their behaviors.”

Driven as a teacher to empower practitioners to do good work, to learn, for example, enough statistics, not to become statisticians, but to produce better policies, his pressing need to measure effectiveness started over 15 years ago in rural Tanzania.

Svoronos had just graduated from Georgetown’s School of Foreign Service, and was leading a Fulbright study to improve maternal and infant mortality using mobile phone technology.

A key problem was getting expectant mothers to deliver in health clinics rather than at home. Svoronos enthusiastically ran his latest scheme by a local nurse: a paid bicycle taxi service to bring pregnant women to clinics. She quickly disabused him of the idea, and offered a far simpler solution.

“Give them free soap, and my doors would never close,” he recalls her saying. He pooled money from family and friends to buy soap, and ran a successful pilot where local health workers told women they would get a bar of soap for delivering in the clinic. After the pilot, the promise of free soap continued to incentivize in-clinic births, saving lives at a fraction of the cost of anything Svoronos could have conceived on his own.

The lessons he learned about valuing existing community knowledge, collaboration, and human connection (if he had not known the nurse so well, she may not have been as frank about his “terrible idea”) led him to co-found , an organization that funds community-driven social impact projects, and to pursue graduate studies in public health.

During his doctoral program at Harvard, he gravitated toward implementation science, studying whether a given research design provides trustworthy enough results for policymakers and practitioners, particularly in developing countries, to act on – and found a calling in teaching.

His focus today is on using technology to replicate the dynamics of small classes on a large scale, a phenomenon he first witnessed as a teaching assistant for Dan Levy at the Kennedy School. From “the first meeting with the teaching team, Dan asked how specific individuals were doing in a 70-person class,” he recalls. The attention on personal learning journeys inspired Svoronos to experiment with emerging tools to customize learning and deepen connection.

His larger projects include the creation of blended online and in-person training programs for civil servants in India and Pakistan, and, in collaboration with Levy, and ����vlog���� alum Kartikeya (Karti) Subramanian, a widely used app to promote inclusive learning.

Launched in 2015, combines data on student backgrounds and interests with real-time stats on participation patterns and in-class interactions. Originally designed to draw more effectively upon the diverse personal and professional experiences of ����vlog���� students, it has evolved into an inclusive teaching tool available to any higher ed or K-12 classroom.

It aims, says Svoronos, to “make it as easy as possible to know your students well,” to be intentional about who’s engaging in the classroom, to identify participation gaps and implicit biases, and encourage broader engagement.

The default is to call on whoever’s most vocal – the classroom equivalent of the famous streetlight effect joke he uses to introduce data and equity, one of his teaching passions.

“There’s someone searching for his keys on a sidewalk under a street lamp. He dropped them a few blocks back, but he’s looking here because the light’s better,” Svoronos quips.

This is too often how people treat data: what is available is taken as representative and accurate, but “we need to question why some data is more available than others, and how we can use it, despite its limitations,” he says.

On the flipside of “trusting data without scrutiny,” he notes, there are those who implicitly distrust data since it is commonly created by people with an interest in maintaining the status quo. He co-developed Evidence for Equity to bridge these two poles, and equip equity-oriented practitioners with the tools to understand both the limits of data, and how to use it to advance policy.

As Evidence for Equity grows, Svoronos wonders: “How can we make policymakers not just more knowledgeable about data but more likely to use it in their day-to-day work?”  

He sees potential in new AI technologies. While acknowledging a need for regulatory caution, he believes generative AI can facilitate, perhaps counterintuitively, meaningful learning. “There’s a narrative that AI will end learning; no one will ever learn again because ChatGPT will answer questions for them,” he states. But he has already seen through his own classroom experiments that, if programmed, for example, to generate questions rather than answers to set parameters, ChatGPT can complement in-class learning in myriad ways.

Svoronos, who is also a jazz guitarist, likens his best teaching moments to jazz improv. They come, he says, from collaboration, from riffing and responding in real-time to “little pearls of insight that you assemble together,” creating new thinking impossible to impart from a scripted handout.

From tools that provide students with personalized problem sets and feedback to custom ChatGPT bots that role-play real-world scenarios with policymakers, and then use that data to assess how they approach evidence, AI tools can free instructors to focus on co-creating insights with students in the moment – and perhaps most compelling for the quantitative analyst, they can track user experience and  measure outcomes.

Ultimately, he says, “we want to have experiences that make us feel closer to each other, and help us understand things better than we used to … if these tools can help us achieve that, I want to explore how we can use them.”