A Bounded-Confidence Model of Opinion Dynamics with Heterogeneous Node-Activity Levels

Abstract

Agent-based models of opinion dynamics examine the spread of opinions between entities and allow one to study phenomena such as consensus, polarization, and fragmentation. One examines them on social networks to investigate the effects of network structure on these phenomena. In social networks, some individuals share their ideas and opinions more frequently than others.These disparities can arise from heterogeneous sociabilities, heterogeneous activity levels, differentprevalences to share opinions when engaging in a social-media platform, or something else. To examine the impact of such heterogeneities on opinion dynamics, we generalize the Deffuant–Weisbuch (DW) bounded-confidence model (BCM) of opinion dynamics by incorporating node weights. The node weights allow us to model agents with different probabilities of interacting. Using numerical simulations, we systematically investigate (using a variety of network structures and node-weight distributions) the effects of node weights, which we assign uniformly at random to the nodes. Wedemonstrate that introducing heterogeneous node weights results in longer convergence times andmore opinion fragmentation than in a baseline DW model. One can use the node weights of our BCM to capture a variety of sociological scenarios in which agents have heterogeneous probabilities of interacting with other agents.

A Multimodal Auxiliary Classification System for Osteosarcoma Histopathological Images Based on Deep Active Learning

“… BALD uncertainty-based strategy (denoted as BALD): in each iteration, we use BALD uncertainty to select samples for labeling. Other strategies based on uncertainty, including entropy-based strategies (entropy) and confidence-based strategies (confidence) [ 18 , 63 ]. Strategy based on GAN reconstruction and BALD uncertainty (denoted as GAN + BALD): simultaneously train GAN and classifier, use BALD to select samples, and then augment the training set with reconstructed and selected images.…”

“…Other strategies based on uncertainty, including entropy-based strategies (entropy) and confidence-based strategies (confidence) [ 18 , 63 ].…”

“…This is primarily because growing the training sample size improves the algorithm’s performance. Uncertainty-based strategies [ 18 , 60 ] (including Entropy and Confidence) are an effective improvement on the Random strategy. Such methods not only end up with higher classification accuracy, but they are able to achieve better accuracy with relatively less labeled data.…”

“…Such datasets are then trained to maximize model performance within a limited labeling budget. The methods of minimum confidence, marginal sampling, entropy, etc., are all based on the uncertainty of the model for the sample [ 18 ]. Such methods are biased and complex, which can affect the representativeness of the sample.…”

An ``Opinion Reproduction Number'' for Infodemics in a Bounded-Confidence Content-Spreading Process on Networks

“…There have been numerous studies of BCMs, which have been generalized in many ways. Recent studies have incorporated phenomena such as media outlets with fixed opinions 37 , polyadic interactions (i.e., interactions between three or more agents) 38,39 , noise 40 , asymmetric confidence intervals 41 , cost of opinion change 42 , agents with heterogeneous activity levels 43 , smooth interaction kernels (in the form of sigmoidal functions) to describe how agents influence each other 44 , opinion repulsion 45 , homophilic adaptivity of network structure 46 , preserving connectivity of the interaction network 47 , and adaptive confidence bounds 48 .…”