Modeling efforts in opinion dynamics have to a large extent ignored that opinion exchange between individuals can also have an effect on how willing they are to express their opinion publicly. Making one’s opinion audible is important: Positions that are rarely voiced in public are improbable to be taken into consideration in, e.g., political decisions. We investigate the conditions that promote or impede opinion expression of different opinion groups in a game-theoretically grounded model with two opinion groups on a stochastic block network[1]. Agents can choose to voice their opinion publicly, or to be silent. Updating their beliefs about the opinion climate around them with the reinforcement learning algorithm of Q-learning, they get more (less) willing to publicly express their opinion if they exchange opinions with an agreeing (disagreeing) neighbor. We reduce the N-agent system in a mean-field approach to two dimensions which represent the two opinion groups, for which we carry out a bifurcation analysis over all relevant parameters. We show that increased structural cohesion (taken into account in the structural parameters γ and δ for the respective groups) and/or decreased costs of opinion expression for one opinion group can amplify its audibility in public discourse. Moreover, we identify conditions under which even opinion majorities can be driven into silence (see Figure 1). These findings gain additional traction in light of the advent of social media which reduce expression costs and facilitate connecting to like-minded others – i.e., building cohesive opinion groups. We also present a Twitter case study[2], covering the 2019 Saxonian state elections, where an opinion minority indeed turned out to be significantly more expressive and confrontative in comment sections than the majority, hence distorting the impression of public opinion on the platform in their favor.
