Student-led startup harnesses AI to predict political violence

A new Yale startup that uses artificial intelligence to forecast organized political violence was recently awarded $87,850 in seed funding.

The venture, called Helios, initially grew out of an independent summer research project by Jackson School graduate student Alex Casendino MPP ’24. Undergraduates Andrew Coli, a statistics and data science major, and Andy Yang, a statistics major, came on board soon after. Using innovative machine learning tools, the team is designing models that track and predict the behavior of adversarial groups and terrorist organizations, such as Hamas and al-Shabaab.

Helios aims to process open-source data — including validated conflict records, social media activity, and local media reports — related to specific groups. The curated datasets are then fed into a predictive algorithm and “layered on top of each other,” Casendino says, providing a more detailed understanding of potential malicious actions that could be taken by these groups in the future. The students believe these forecasts can provide unique insights into adversary behavior.

“It’s clear that the [current] information often isn’t specific enough to provide actionable insight to someone who’s sitting in the State Department, trying to determine what’s going to happen in a specific country or even down to a local level,” says Casendino. “What we’re aiming to do is get even more granular, identifying the actors, the timing, and location with tangible intelligence to support effective policy.”

Helios is in the early stages of development, building upon a “foundation that’s been established in the research community,” says Casendino. The students currently have access to powerful open-source machine learning packages and are using data from platforms that collect information on conflicts around the world. The long-term plan is to find novel ways to gather, organize, and merge this data while also expanding the model’s scope to include additional datasets, such as economic data and local terrain.

When the project began last summer, Casendino says he was initially focused on finding seasonal patterns of political conflict. In fall 2023, he took a machine learning course taught by Yale professor John Lafferty and enrolled in the yearlong course “Artificial Intelligence, Emerging Technologies, and National Power,” taught by lecturer Ted Wittenstein. It was in these classes that Casendino met Coli and Helios really took shape.

With the help of Wittenstein, the students were able to connect with AI leaders across the public, private, and nonprofit sectors, where they’ve benefitted from expert advice and networking opportunities. It was through the Jackson School’s Schmidt Program on Artificial Intelligence, Emerging Technologies, and National Power that the students refined the concept, designed the pitch for outside seed funding, and secured financial backing to transform their class project into a promising business venture. 

Wittenstein has been impressed with the project thus far. “Helios is a great example of utilizing AI tools to solve concrete global challenges,” he says. “It is exciting to see Jackson graduate students collaborate with Yale undergrads in the Schmidt Program and transform their research projects into promising startups.”

The students also see the benefits of working on this project while studying at the Jackson School, which brings together students and practitioners from the diplomatic ranks, the military, and private industry. The ability to continuously engage with experts from these diverse fields helps ensure that Helios is evolving into “something useful and not just flashy engineering models,” Casendino says.

Looking forward, the Helios team plans to build the project into a small business with the potential to support government partners and NGOs.

“We see a real opportunity here,” adds Coli, “to harness open-source data and develop a model that can potentially change the landscape of policymaking and security strategies.”