On August 1st, Jae-Young Son defended his dissertation titled, “Abstraction underlies inferential representation of social networks.” He was supervised by Dr. Oriel FeldmanHall. This fall, he will be starting as a postdoctoral fellow in Dr. Serra Favila’s lab here at Brown. 

More specifically, Son’s research explores how people navigate complex social networks by building mental maps that help them understand and infer connections between others. These cognitive maps allow individuals to make strategic decisions, such as finding new opportunities or avoiding potential risks, even when they can't observe all relationships directly. The findings show that these abstract maps help people efficiently navigate social networks by filling in gaps left by incomplete information. 

To learn more about his dissertation, check out the abstract of his dissertation, which is included below, get in touch with him at jae-young_son@brown.edu, or check out his personal website: https://jaeyoungson.com/

Abstract: 

In the context of vast, complex, and dynamic social networks, strategic decision making relies on knowing how people are connected within their larger social communities. A job applicant might contact her ‘weak ties’ in hopes of being referred to a recruiter; a gossiper might hold his tongue in fear that a particular individual would spread a rumor to a different community; a manager might land a valuable position within the company after she realizes that she can bridge two disconnected teams. Yet in most social networks, the space of possible relationships is too vast for any individual to memorize or even observe. How, then, do people build mental representations of social networks that aid adaptive navigation? Drawing on a long history of research on cognitive maps in spatial navigation, I propose that people build abstract cognitive maps of social networks, which afford efficient and flexible inference of unobserved but probable social relationships, and provide a mechanistic, computational account of how people learn, represent, and navigate social networks. One mechanism, known as feature-based abstraction, relies on learning about abstract relationships between features rather than individuals. For example, upon observing Katherine the mathematician having lunch with Mary the engineer, an individual could use the ‘Katherine-to-Mary’ friendship to drive learning about a latent ‘mathematician-to-engineer’ relation, and generalize this knowledge to other mathematicians and engineers in the network. A second mechanism, known as multistep abstraction, encodes knowledge of network members’ relationships as a weighted combination of direct and indirect connections (i.e., one-step and multi-step relations). For example, upon observing friendship between Katherine and Mary, and then later Mary and Dorthy, an individual could stitch together these observations and infer the existence of an unobserved relationship between Katherine and Dorthy. Using a combination of computational modeling and empirical experiments, this dissertation demonstrates that abstraction allows people to build inferential cognitive maps of social networks that ‘fill in the gaps’ left by noisy and incomplete observation. These abstract cognitive maps, in turn, aid adaptive social navigation by allowing people to represent longer-range connections between network members, as well as the global structure of the social network as a whole.