The Washington PostDemocracy Dies in Darkness

A teenager used AI to study a very real issue and just won $175,000

The Northern Virginia high school student’s winning science project used GPT-3 to examine who gets humanized and who doesn’t when it comes to media coverage of homicide victims

Emily Ocasio placed second in a science competition Tuesday night. (Society for Science/Chris Ayers Photography)
6 min

To understand how Emily Ocasio won $175,000 on Tuesday night takes thinking about how society views people based on their identities.

It takes considering who we view as worthy of victim status after a crime happens. Whom do we see as the most innocent, the most vulnerable, the most deserving of our sympathy? And whom don’t we?

The Northern Virginia high school student used artificial intelligence to examine a very real issue — who gets humanized by the media after a homicide happens and who doesn’t — and in doing so, she made something that is hard to measure measurable.

“I think a core part of my project was essentially developing a novel methodology in social science that allows us to look at text in an efficient way that has never been done before,” Ocasio, 18, explained to me on Tuesday night. “I used GPT-3, which a lot of people know through ChatGPT. Really what it was created for was to produce language, produce sentences. However, I wanted to show that this is a tool that not only could be applied to social science but could be applied as an analytical tool.”

GPT-4 has arrived. It will blow ChatGPT out of the water.

I spoke to Ocasio just minutes after she learned she won second place in the Regeneron Science Talent Search, a competition that aims to identify and inspire the nation’s most promising young scientists. More than 1,900 high school seniors from across the country submitted original research as part of the competition, and on Tuesday night, the top 40 finalists were celebrated during a ceremony at the National Building Museum.

Toward the end of the evening, as the finalists gathered on the stage, the top 10 winners were announced. Three of them were from Virginia. Ethan Zhou, of McLean, placed seventh and received a $70,000 award. Max Misterka, of Harrisonburg, placed fourth and received a $100,000 award.

When Ocasio, who attends the New School of Northern Virginia in Fairfax, heard her name called, her mouth dropped open in surprise. Then came the tears. She wiped them away repeatedly as she stood next to the other winners.

“I’m feeling so beyond overwhelmed,” she told me afterward. “It’s almost like an out-of-body experience.”

Ocasio said she didn’t expect a project focused on social science to win such a competitive STEM competition, but she felt encouraged that it did.

“I picked the subject because I’ve always been interested in understanding people and how they work together, and in particular what keeps certain communities from thriving and succeeding,” she said. “I think the fact that I was able to win second place in a huge STEM competition shows that we’re beginning to absolutely value and support quantitative research about people and understanding the roles of race and gender in society.”

How a teenager got more women of color added to history lessons in one of the country’s largest school systems

It is easy to watch what’s happening in the world right now and feel a sense of doom, but it takes only looking at the students who stood on that stage to gain a sense of hope. Their projects took on issues that ranged from cancer research to climate change to suicide prevention.

Neel Moudgal, of Michigan, who won first place and $250,000, is described as “creating a computer model that can rapidly and reliably predict the structure of RNA molecules using only easily accessible data.” And third-place winner Ellen Xu, of California, is described as “developing an algorithm that uses a smartphone photo of the patient to aid in the diagnosis of Kawasaki disease, the leading cause of acquired heart disease in children between one and five.”

“When we look at these young people, they are the future scientific leaders of this country, and they are just brilliant,” said Maya Ajmera, president and CEO of the Society for Science, which launched the competition in 1942. “They are also looking to solve the world’s most intractable problems.”

As a journalist who used to cover crime in New York and Virginia and still writes about it often, I spend a lot of time thinking about the issue Ocasio’s project addresses, so it stood out. Whenever I’m asked to speak to student journalists, I always advise them to strive to find the humanity in the person they are writing about. That doesn’t mean portraying a person as good or kind if they weren’t. Humanizing doesn’t mean prettifying. It means finding details about a person’s life that will make them more than a number or a name. Did they have a nickname? Did they have a pet? What do their family and friends have to say about them?

A Black disabled teen went unheard in prison. People are now listening.

This isn’t always easily achievable. Journalists often work under tight deadlines and have doors slammed in their faces. But it should be the goal, and measuring whether we are doing that equally across races, ethnicities and gender identities helps us get closer to it.

For her project, Ocasio used FBI data, 5,000 archival articles from the Boston Globe and GPT-3 to analyze coverage of homicides in Massachusetts between 1976 and 1984. She assessed each article for humanizing or impersonal language. She then examined that coverage across race, gender, age and the intersection of those identities, and created a “composite humanizing score.”

What she found was that young Black males under the age of 18 were 30 percentage points less likely to receive humanizing coverage than their White male counterparts. She also found that Black women within the age range of 18 to 29 were 23 percentage points less likely to receive humanizing coverage as their White female counterparts.

Ocasio said she would not have been able to detect that by simply reading the articles, because identifying details, such as a person’s race and ethnicity, were not always mentioned. Matching each article to FBI data helped her fill in those blanks.

She said her next step, beyond getting ready for college (she has already been accepted to Harvard), involves applying what she found to more modern data.

At Tuesday night’s event, many of the speeches set high goals for the finalists. I asked Ocasio about those expectations, and she described the experience as leaving her feeling hopeful about the future. She said she has spent the past few years dealing with a chronic illness that affects her heart rate, and that when her fellow finalists learned about it, they responded in a way that showed respect and curiosity.

“If one day, there is a cure,” she said, “I wouldn’t be shocked if it came from one of them.”