Influence of Word of Mouth Interactions vs. Mass Media Effects on Opinion Formation and Accuracy

We are studying a society of agents that are exposed to information directly or indirectly via mass media and word of mouth. This is a social learning study, rather than a study on bounded confidence. Assuming that directly perceived information shows high accuracy, we show how the degree of exposure to the mass media projection and to the word of mouth interactions with neighbours in the social networks is decisive for the accuracy of the final opinion of agents. More precisely, we show that in societies with a higher degree of socialization (frequent contacts with social neighbours) the opinion on any issue tends to be more accurate than in societies with a higher degree of exposure to mass media that may distort the real information sharing ; and this occurs even in the absence of the well-known bias toward face-to-face rather than indirect information. These are implemented through a dual work-leisure dynamics, where during working hours the agents are situated in a a workplace and receive or formulate first-hand information about it, whereas during leisure hours they socialize with other agents or follow the mass media and receive indirect information about other workplaces. The relative amounts of time spent in each of these activities determine the accuracy of the final opinions formed. Because the opinions concern the actual value of external objects and are not arbitrary, no true polarization or diffusion outcomes may be observed, and most of the opinion formation and learning parameters are shown to have no effect on the final configuration of opinions.
Our observations can partly explain the phenomenon of higher resistance to mass media information in certain countries, for example in the European south, that should therefore not be attributed solely to distrust toward the media. In such highly sociable populations, positive or negative propaganda and/or mass control can only be made more effective if the agents have individual direct access to only a small fraction of the total information available, that is by a divide-and-conquer strategy. This translates in our model as a sparser distribution of agents to workplaces that does not allow them to form collective correct opinions in the first place. We finally show how this divide-and-conquer strategy can lead to social automatism and we investigate the kind of social/syndicalist leadership that could help alleviate this phenomenon.