Distance to an event changes how people describe it, Penn research shows
“The way people think about events varies as a function of how far away or close those events are,” says Sudeep Bhatia, an assistant professor in the Department of Psychology in the School of Arts & Sciences. The more distant an occasion, the more abstractly it’s described. As it gets closer, the portrayals become more specific.
That well-established conclusion—a key prediction of a concept called the construal level theory—had to this point largely been tested in a laboratory setting. Bhatia and Lukasz Walasek of the University of Warwick confirmed it using natural language processing techniques from true-to-life examples. They published their findings in the journal Cognition.
“You get someone in the lab and you say, ‘Imagine you’re going on a vacation in two years. Write down what you’re thinking about.’ That’s the sort of the approach that’s been used,” Bhatia says. “Wouldn’t it be amazing if these effects also manifested themselves in the way people talk about events in the real world? We tested that.”
Bhatia and Walasek honed in on phrases used in two mainstream media outlets, The New York Times and Twitter, looking for terms from a 30,000-word collection rated in previous research as either “concrete” or “abstract.”
Penn has a database of every Times article from the 1980s to 2008, each tagged with keywords. Because presidential elections happen cyclically and receive an excess of media coverage, the researchers opted to analyze any article leading up to and immediately following a presidential election. Time to voting day determined how precisely articles explained candidates’ positions or the overall race, for example. As an election approached, ideas solidified—until Election Day itself.
“There’s a sharp drop-off right after the election,” Bhatia says. “Leading up to it, concreteness increases, increases, increases, and then right after, boom, it drops.” The researchers replicated the findings using articles about annual holidays like the Fourth of July and New Year’s Eve.
To test the theory with another broadly used tool, the researchers collected every tweet posted on Twitter during a one-month period in 2014, pulling out any containing “2015,” “2016,” “2017,” or “2018” and “next week,” “next month,” or “next year.”
Just as with the Times articles, phrases signifying something sooner—in this case “next week” versus “next year” or “2015” versus “2018”—were presented more definitively.
“Prior work has shown that if you represent an event concretely versus abstractly, that affects how much you discount the event, how important the event feels to you,” Bhatia says. “You can also have similar results with people.” Social psychology research has shown that stereotyping occurs much more frequently when people think abstractly.
Bhatia is conducting other computational linguistics work, studying how people’s expectations about elections shape the way they discuss and write about these events. For now, he’s keeping a close eye on November—not only to see who is elected president, but also to add data points to his research.