Fighting Child Exploitation with Data Science

Every day, 100,000 new escort ads are posted online. That is according to Thorn, a nonprofit that fights child sexual exploitation through technology innovation. Other studies by Thorn on the underage sex trafficking situations of survivors have shown that 63% of them had been advertised online at some point. The massive online commercial sex market is extremely difficult to fight, which has inspired the start of an annual hackathon to bring cross-industry experts together to work on child safety.

The Domino team showed up to the bright and cheery Facebook HQ in Menlo Park, the meeting spot for the 2nd annual 2017 Child Safety Hackathon, where throngs of volunteers poured in to tackle a very dark and real topic. The brightest minds in engineering, data science and policy were combining forces to develop technology solutions for finding missing children and keeping kids safe, and Domino was their tool of choice.

In their everyday operations, Thorn leverages tech companies for their knowledge, time and resources to build products that professionals on the front lines can use to fight child sexual exploitation. A major component of the products they build manifests from powerful data science techniques. This is why for the hackathon in particular, Thorn played an important role in providing hard-to-get data around child exploitation.

With more than twenty different challenges that anyone could join, volunteers with engineering and data science backgrounds gravitated to Thorn’s projects because most of them required web-scraping, machine learning, natural language processing and image recognition expertise. For Thorn, working on challenges like this is not unique to hackathon events. They rely on volunteers to solve these kinds of problems every day by developing real tools for law enforcement use in the field.

For many companies, even with the luxury of having an in-house data science team, it can be difficult to iteratively experiment and then get insights for stakeholders and models into production (ie, into downstream tools and operations).

The complications are magnified for Thorn, because contributions are erratic and volunteers come and go. It is therefore critical that Thorn has a place for volunteers to get onboarded and start work on a project right away. It should be easy for volunteers to pick up existing work where another volunteer left off, and save their work in a central place so that Thorn can put it into production even after they are gone.

This is why Thorn uses the Domino platform. It serves as a central hub that houses all the work from both internal and volunteer data scientists.

For the Domino team, the Child Safety Hackathon was an inspiring and gratifying experience. We had the opportunity to watch how Thorn uses the Domino to get leverage from amazing talent in order to develop innovative ways to save children from sexual exploitation.

It was a great experience for the volunteers, as well. As they self-selected into Thorn’s challenges, they had immediate access to the project’s data, environments, templates, and information because it was stored in Domino. They were able to start working and collaborating with their new team right away. Volunteers from other teams even came over to ask how Thorn’s team was able to get started so quickly!

As team members started experimenting and writing code they were able to easily share progress on algorithms and compare side-by-side results with others, all within the platform. Using the AWS instances Thorn had provisioned specifically for the hackathon, the team had instant access to compute resources for fast experimentation and iteration.

In the end, Thorn walked away with reproducible work and prototypes produced by some of the top tech talent in the valley, all captured and recorded by the power of Domino.