by Anna Fiorentino
Each year, thousands of research papers are published in a given academic discipline — more than any scientist entering that field could ever keep up with.
But a new AI tool offers a solution.
The approach uses a review method for mapping out scientific literature, called citation network analysis. It identifies key influential papers, giving researchers a bird’s eye view into their given academic field. The findings by Institute of Experiential AI faculty member Ajay Satpute, director of the Affective and Brain Sciences Lab, and graduate student Alessia Iancarelli were published on April 21 in the academic journal PLoS One.
“There’s so much research on emotion now that it’s impossible to read it all. What does it really mean to be an expert if you can’t know all of the literature?” says Satpute.
Starting with a “seed” article, the new tool searches academic databases like Google Scholar for any piece of literature cited by that author. From there, it traces other papers cited by those authors. By analyzing the patterns and frequency of the citations, it extracts the most relevant papers based on their root, clustering related literature.
For example, Iancarelli identified the most influential research using her area of interest — human aggression — as a basis for the study. The tool came up with 15 research communities on aggression, flagging the most influential papers within each of them. Next up, she’s using the system to explore the roots of a landmark paper on western civilization about the science of emotion, looking back 140 years at who has cited that work.
“From this approach, we can see if there were research communities that emerged in the 1920s and 1930s, for example, that had a little moment, but disappeared,” says Satpute. “Unless you’re lucky enough to have a deep historical scholar in the field, that whole community of research would otherwise be forgotten.”
The team also managed to mitigate gender bias in their citation network analysis by identifying influential articles based on community structure rather than the number of citations, therefore promoting gender equality within the field of aggression.
“Often people only look at the most relevant paper,” says Iancarelli, who is tasked with the challenge of continually updating the AI tool’s code to make sure it’s compatible with the programming interface of each of the databases. “It’s really important for people to be aware of how a field is constituted. That’s where a network approach makes a big difference.”
Learn more about Satpute’s work in the Affective and Brain Sciences Lab and the exciting research taking place at the Institute for Experiential AI. To find out more about partnering with the institute in business or research, contact us today.
Ajay Satpute is an assistant professor in the psychology and neuroscience departments and director of the Affect and Brain Science lab. His research addresses fundamental questions about how the mind relates to the brain, from how language shapes emotional experience to whether brain activity alone can predict a person’s social or emotional state. Through brain imaging data, examining variation across participants, and computational modeling and machine learning, he develops data-driven approaches to advance both theory and research in cognitive neuroscience.
Prior to his current role at Northeastern, Satpute previously served as a senior research scientist for the university. He also worked as an assistant professor at Pomona College and a postdoctoral researcher at Columbia University. He earned his doctoral and master’s degrees from the University of California, Los Angeles, and a Bachelor of Arts from Trinity University.
Alessia Iancarelli is a postdoctoral researcher in psychology at the Affect and Brain Science lab.