May 28, 2015 · 25m · mad
Jake Porway, DataKind // Data For The Greater Good (FirstMark / Data Driven NYC)
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
At Data Driven NYC, DataKind co-founder Jake Porway presents how data science, machine learning, and human-centered design can be repurposed from commercial consumerism to solve complex global social issues. Through real-world case studies and ethical frameworks, Porway demonstrates the power of pro bono analytics to drive meaningful, life-saving impact.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 1.1% of the talking time here. How this is scored →
Matt as informed peer 0.0 Guest teaching 2.2 Guest disagreement 0.0 Matt pushing back 0.0
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 0:05 Playful rejection of TED tropes
Jake playfully clarifies that having the audience stand up is not a canned TED-style engagement trick, establishing a casual tone.
Hardest push from Matt ▶ 20:08 Host question on project selection criteriaMatt asks how DataKind manages to select projects given that they are overwhelmed with inbound requests from both volunteers and organizations.
Biggest teaching moment ▶ 12:50 Algorithmic flaw in GiveDirectly satellite modelJake educates the room on how high algorithm accuracy can be misleading when machine learning misses local cultural context like thatched kitchen huts on wealthy Kenyan properties.
Matt holds his own ▶ 0:26 Host icebreaker jokeMatt humorously interjects with 'Do we say om?' during the guest's audience stretch exercise.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Event Opening and Audience Stretch Icebreaker | 0 | 1 | 0 | 0 | Jake opens the presentation with an audience stretch icebreaker and brief banter with host Matt Turck. The atmosphere is warm and collaborative with no host pushback or domain grilling. | |
| Foundations of DataKind and African Microloan Data Case | 0 | 2 | 0 | 0 | Jake presents the background of DataKind and a DonorsChoose case study analyzing public school resource requests. Host scores are zero as this is a monologue presentation. | |
| Case Study: Crisis Text Line and Repeat Texter Optimization | 0 | 2 | 0 | 0 | Jake details data science applications in Crisis Text Line and GiveDirectly cash transfer targeting. Host scores remain zero for this uninterrupted guest monologue. | |
| DataKind's Expansion and the Necessity of Human-Centered Design | 0 | 3 | 0 | 0 | Jake discusses DataKind expansion and human-centered design, explaining how a 95% accurate satellite model failed due to missing local contextual knowledge regarding thatched kitchen huts on wealthy estates. | |
| Grassroots Data Hacks: LA Swimming Pool Mapping & Malaria Application | 0 | 2 | 0 | 0 | Jake highlights grassroots data science projects, such as an LA swimming pool detection algorithm redirected to combat malaria stagnant water. Host does not intervene during this segment. | |
| Algorithmic Ethics, Bias, and the 'Macroscope' Analogy | 0 | 3 | 0 | 0 | Jake wraps up his presentation discussing ethical pitfalls in algorithms and framing data science as a macroscope for society. Host scores remain zero throughout the monologue portion. |
Statements from this episode (14)
Jake Porway: Social sector data science depends more on politics than technology.
“Applying our trade to social sector issues, which often has less to do with data and technology and more with politics and people”
Jake Porway: Placing data scientists in non-profits fails without structural support.
“Data scientists are expensive, and even if you just plopped a data scientist in with a lot of these non-profits, they wouldn't know what to do with one.”
Jake Porway founded DataKind to match volunteer data scientists with non-profits.
“So that's why we founded Datakind. It's a non-profit that would team up pro bono data scientists. People who wanted to volunteer their time anytime from a weekend to a six-month project, all the way to working on a full-time, full-day job project to use data i…”
DonorsChoose holds the largest US database of public school resource requests.
“What DonorChoose realized is they're actually sitting on the largest database of school requests around public schools in the US. No one's ever looked at that data before. No one else has it.”
DonorsChoose uses its classroom resource data to actively lobby US lawmakers.
“DonorsChoose is now taking these findings, and they're actually becoming an advocacy group. So they are lobbying lawmakers and decision makers to say, hey, did you know this is going on in your school systems? We have this data that once we analyze, we're able…”
Five percent of Crisis Text Line users consume 40% of counselor resources.
“So it's something like five percent of the people that call into these crisis lines take 40% of the resources.”
DataKind volunteers mapped poverty across Kenya and Uganda in two hours.
“So a team from IBM and Mozilla basically built a classification algorithm that would find the different roof types. Interesting regression model we could talk about later. But what it allowed them to do is basically build this map of roofiness be a proper stat…”
DataKind has completed over 60 projects with 7,000 global volunteers.
“We've done over 60. We've got over 7000 volunteers around the world. Recently we expanded to six international chapters.”
Jake Porway: Most social sector groups need basic exploratory data analysis.
“Most of these groups really just need basic exploratory analysis, data that's never been mined that could be really effective.”
Accessible computing enables independent social impact projects outside of traditional institutions.
“We don't have to make change just from work. We have the computing power and the data that's out there to build things like this just sitting at home in our underpants.”
Swimming pool detection algorithms were adapted to identify stagnant water for malaria.
“The one that actually this ultimately ended up getting used for was outside of the US for actually doing malaria prevention. So it turns out that stagnant pools are a huge issue for malaria. And so people use a similar algorithm to find these pools and drain t…”
Jake Porway: Unintended algorithmic biases will dictate how humans actually work.
“These unintended cultural and social biases are making their way into tools that are going to shape the way people work.”
DataKind built a logistic regression model for Amnesty International human rights tracking.
“We face this problem building a project with Amnesty International predicting human rights violations. Built a great logistic model that would help us understand, based on the severity of a report of a human rights violation, which ones Amnesty International g…”