A Two-Tiered Formalization of Social Influence

Reflecting on Social Influence in Networks

“…This transparency assumption rules out the modeling of situations where agents act in ways which do not reflect their actual mental states. However, as argued in (Christoff and Hansen 2013), the possibility of this very discrepancy is an essential component of some social phenomena.…”

“…Section 1 recalls the frameworks of (Liu, Seligman, and Girard 2014) and (Christoff and Hansen 2013) and the corresponding notions of social influence. Section 2 discusses what agents can learn if they reflect on the fact that others are following the rules of social influence.…”

“…According to (Liu, Seligman, and Girard 2014), this would be interpreted as follows: all agents agree with the norm (even though they might come to agree with it only as a result of social conformity pressure). The extended framework of (Christoff and Hansen 2013) allows this situation to be analyzed in a slightly more complex way: even though agents are aligning their behavior (that is, enforcing the norm), they might still privately disagree with it, they might just face too much social pressure to dare expressing publicly a discarding opinion.…”

“…(Liu, Seligman, and Girard 2014) proposes a simple model of how agents change their opinions (or beliefs) as a result of social influence seen as social conformity pressure: agents tend to align their opinions with the opinions of agents they are related to within their social network. This framework is generalized in (Christoff and Hansen 2013) to allow for modeling of phenomena where the observable behavior of a group of agents does not reflect their private opinion. Such counterintuitive group behavior is well-documented in social sciences, as exemplified by studies of the particular case of pluralistic ignorance (see for instance (O'Gorman 1986;Prentice and Miller 1993).…”

“…In short, while (Liu, Seligman, and Girard 2014) assumes fully "transparent" agents, whose opinions are always known to agents around them, (Christoff and Hansen 2013) assumes fully "opaque" agents, whose (private) opinions are never known to anybody else. What we want to model here is how agents might "see through" each other's behavior and come to know other's private opinions by observing their public behavior.…”