Columbia Student Brings Data-Driven Focus to Policy Negotiations, Journalism
Two decades later, computers are now embedded in every aspect of modern life. As they generate ever-increasing amounts of data, Imani has found a new enigmatic world to explore. Graduating this spring with a master’s in data science from Columbia Engineering School, he looks forward to mining massive data sets for social good.
“Data science has the power to make institutions more transparent and improve the lives of ordinary people,” he said. “I’m interested in this public-service mission more than using data science to optimize ads or make more money for shareholders.”
Born and raised in Iran, Imani earned his undergraduate degree in mechanical engineering at Teheran’s Sharif University of Technology. A short stint in the oil and gas industry made him rethink his career path. He studied for a master’s in sustainable development at the University of Edinburgh in Scotland, and eventually found a job at the UN organization that oversees the global chemical weapons ban—the Organisation for the Prohibition of Chemical Weapons (OPCW). There, he encountered his first real-world Big Data problem.
The OPCW at the time was stalled in negotiations over a universal ethics code for chemists. Most of its member-states wanted to stick with the code they deemed best — their own. It was unclear what common language, if any, existed.
Imani developed an algorithm to sort through 140 conduct-codes covering thousands of pages of text. To his surprise, the algorithm uncovered relatively little variation among national codes; Wording varied mostly between universities and industry within a country. “We presented our findings to the OPCW member-states and said, ‘Look! You guys are mostly saying the same thing,” he said. “The similarities were hidden under the politics of negotiation.”
Working from this basis, the member-states moved quickly to adopt a common code. In late 2015, the Hague Ethical Guidelines for chemists was signed.
By then, Imani was at Columbia, exploring new ways to use powerful computers and big data sets to cut through personal bias. One of the most polarizing presidential races in U.S. history had just begun. At a hackathon in early 2016, Imani and three friends wondered if machines might do better than the political pundits in evaluating the candidates.
“There was lots of talk about media bias, and people dismissing information from the other side,” said Imani. “We thought computers could provide a new layer of objective information.”
In two days, the team developed an application to ‘watch’ Democratic rivals Hillary Clinton and Bernie Sanders debate on TV and track their emotional state. By measuring changes in their facial expressions throughout the debate, the app was able to infer the likelihood each candidate was feeling happy or surprised, sad or contemptuous, and disgusted or angry, as they spoke. Encouraged by the results, Imani worked over the next several months to fine-tune the app on his own.
Eventually he partnered with journalists at Quartz to evaluate Clinton’s debates with Republican opponent Donald Trump. Again the app picked out clear patterns, with Clinton coming across as mostly happy, and Trump, mostly sad or angry, they reported.
“Everyone knew Hillary is a well-behaved politician,” said Imani. “She smiles while taking criticism. Even while she’s being attacked she tries to keep her cool. It was interesting to get that confirmed. On the Trump side I was surprised to see a lot of sadness.”
In a follow-up piece comparing Trump’s first two speeches as president, they found a high level of surprise as he gave his acceptance speech, and more sadness and anger at his inauguration. The computer detected anger, for example, as Trump uttered the words, “It is the right of all nations to put their own interests first.”
A third story written in collaboration with the Associated Press, not yet published, will focus on others in the Trump administration. “The analysis is a new way of writing stories,” said Imani. “It augments the journalist’s work — it’s not meant as a stand-alone analysis. The journalist needs to bring additional context.”
In the coming months, Imani will work on a project, Data Interrupted, exploring the legacy of the 1994 Rwandan genocide on the availability of weather and climate data. Funded by a year-long $35,000 ‘Magic Grant’ from the Brown Institute for Media Innovation, the project will be led by staff at Columbia’s International Research Institute for Climate and Society.
As artificial intelligence makes its way into newsrooms, Imani says he is considering data journalism as one possible career path. But plenty of other fields have technical gaps waiting to be filled, he says. “My main goal is to do something meaningful that will have a broad impact on society.”