Wright-Fisher-type equations for opinion formation, large time behavior and weighted logarithmic-Sobolev inequalities (Q2334999)

The article investigates the rate of convergence to equilibrium of the solution of a particular Fokker-Planck-type equation with variable diffusion coefficient and boundary conditions. This equation arises in the formulation of an opinion formation model in which the opinion variable is allowed to take values in the bounded interval $(-1,1)$. The value of each agent's opinion is allowed to change through binary interactions between the agents of a society. After each interaction, the agent's new opinion results from a combined effect of a compromise with the other agent's opinion and a tendency to keep the opinion emanating from its own self-thought, which varies randomly between interactions. The resulting Fokker-Planck-type equation is parametererized by magnitudes representing how strong compromise and self-thinking are in the society under study. These parameters also determine the shape of the equation's steady states (i.e. opinion distributions that are reached after some transition time), which can be represented by Beta-type probability densities supported in $(-1,1)$. It turns out that these parameters' values determine whether in equilibrium a strong consensus or a high degree of radicalization among the agents will exist. The article's focus is to reckon how these parameters influence the rate at which the solution of the equation reaches its equilibrium distribution. For an spatially homogeneous Fokker-Planck equation, in which the diffusion coefficient is constant, it turns out that the rate of convergence to equilibrium is exponential and can be explicitly obtained. This result follows from using the Shannon entropy of the solution relative to its steady state as a Lyapunov functional and then using the logarithmic Sobolev inequality. Since the equation considered by the article includes a special diffusion term, this procedure does not lead to the same conclusions. In order to apply entropy methods to the equation considered, the article contributes by formulating a new weighted logarithmic Sobolev inequality that is satisfied in equilibrium, at least in some range of the aforementioned parameters, if the steady-state distribution belongs to $L^2((-1,1))$. This inequality is used in order to conclude that, if the parameters of the particular Fokker-Planck equation are contained within the defined region, then the solution converges in $L^1((-1,1))$ towards its equilibrium at an exponential rate, and the rate is computed explicitly as well.