Figure 2

(a,b) The phase diagram of the model on a first nearest neighbor network, for respectively, the cases when the noise is in comprehension, and when it is in production. The model shows two ordered phases; an informed consensus phase for high observation probability, separated by a first order transition from an uninformed consensus phase. By increasing the noise, the model goes to the disordered phase in which no consensus is formed. Transferring noise from comprehension to production shifts the order-disorder transition line to smaller noise levels and higher observation probabilities h. This shows that the production noise is more detrimental for the collective inference of the population. (c) The comprehension-production asymmetry, defined as \({\rm{\Lambda }}(0.75,0,\eta )\,-\,{\rm{\Lambda }}(0.75,\eta ,0)\), as a function of h and \(\eta \), on a 10 × 10 first nearest neighbor network. The shifts in the transition lines lead to the positivity of the asymmetry. (d,e) The phase diagram of the model on a second nearest neighbor network, for respectively, the cases when the noise is in comprehension, and when it is in production. (f) The comprehension-production asymmetry, \({\rm{\Lambda }}(0.75,0,\eta )\,-\,{\rm{\Lambda }}(0.75,\eta ,0)\), as a function of h and \(\eta \) for a second nearest neighbor network.