The first time mathematics professor Amy Langville heard of Tyler Perini, he seemed like the type of academic superstar that comes through the math department every decade or so. But upon meeting Perini, Langville came away with a different impression. Actually, says Langville, “He’s the type of kid that comes around once in a generation.”
That’s high praise for any student, and Perini is quick to return the love to his faculty advisor. He credits Langville as an excellent mentor who demonstrates practical ways in which to apply mathematics. Together the pair recently collaborated with philosophy professor Thomas Nadelhoffer and psychology professor Jen Wright on a research project focused on determining the psychological underpinnings behind people’s abundance of humility or lack thereof. Specifically, Langville and Perini helped create a computer program that can analyze writing samples in order to determine how humble a person is.
For example, research subjects were asked questions about their relationships to others, to the universe, to God and to the environment. Their answers were then analyzed by Perini’s program, which then indicated how humble the test subject was. Sounds straightforward, right?
Not exactly.
At first, Perini and Langville were inclined to have the program ignore what are called function words, which include articles and pronouns. But, after learning of linguistic theories by James Pennebaker at the University of Texas at Austin, they changed their approach. In his book, The Secret Life of Pronouns, Pennebaker argues that function words are extremely important and can be indicative of all sorts of traits, including whether people are truthful, how rich they are or how much power they think they have. In other words, Perini explains, function words act as an “invisible signature.”
So Perini started analyzing every word in a response. He also began scaling up his research, being able to analyze not just a sentence at a time for indicators of humility, but whole paragraphs. Perini made his program 67 percent accurate.
Then came a bigger test: What if you fed his program random samples of text unrelated to the experiment? What if you gave the humility-detecting program eight statements from a paragon of virtue? Someone, say, like Mother Teresa?
Turns out Mother Teresa is not too humble, at least according to Perini’s analysis. Only one of eight statements she made displayed a significant amount of humility, leading Perini to suspect his textual analysis program needs continued refining. Or, as Langville says, maybe Mother Teresa was a humble person, but not everything she uttered would indicate humility.
In any case, Langville says the results of their work have been incredibly satisfying. Already they have applied for a grant to expand their research into analyzing text samples to detect levels of self-control.
A lot of the project’s success can be attributed to Perini’s knack for translating mathematical principles and results into plain language and easy-to-understand graphics. Perini, too, seems to revel in his position as liaison and interpreter.
“I’m caught between a mathematician, a psychologist and a philosopher, and I have to make sense of it all,” says Perini. “That’s my favorite part of this project.”
Photo by Kip Bulwinkle ’04
