Inference in Opinion Dynamics Under Social Pressure

We introduce a new opinion dynamics model where group of agents holds two kinds opinions: inherent and declared. Each agent's is fixed unobservable by the other agents. At each time step, broadcast their declared opinions on social network, which are governed agents' pressure. In particular, we assume that may declare not aligned with to conform neighbors. This raises natural question: Can estimate from observations opinions? For example, represent true political alliances (Democrat or Republican), while inclinations tweets media. this context, seek predict election results observing voters' tweets, do necessarily reflect support due analyze question in special case underlying network complete graph. prove that, as long population does include large majorities, estimation aggregate individual possible. On hand, majorities force minorities lie over time, makes asymptotic impossible.