How Do Information Cascades Spread in Social Networks? Models & Simulations

Social networks are a primary platform for information dissemination, and their dynamic, complex structures have long attracted multidisciplinary research. This paper explores information diffusion models in social networks, analyzes the intrinsic link between network structure and diffusion, and attempts to quantify and predict diffusion paths and scopes.

1. Foundations of Diffusion Models

1.1 Independent Cascade Model

The Independent Cascade Model describes how a user who adopts information may influence neighboring users. Specifically, after a user adopts a piece of information, they have a single chance to cause each neighbor to adopt it with a certain probability, creating a cascade effect throughout the network.

In this model we assume that each user, once they adopt the information, has one opportunity with probability p to make each neighbor adopt the information.

1.2 Linear Threshold Model

The Linear Threshold Model focuses on social influence factors that affect an individual’s adoption of information. Each user has a fixed threshold; only when the proportion of neighbors who have adopted exceeds this threshold will the user adopt the information. This model highlights the role of social pressure and collective behavior.

Formally, a user’s neighbor set is represented, each neighbor’s influence weight is assigned, and the user adopts the information when the weighted sum of active neighbors surpasses the threshold.

2. Network Structure and Information Diffusion

2.1 Network Communities

Social networks often contain tightly‑connected community structures that act as “resonance chambers,” amplifying and accelerating information spread.

2.2 Bridge Nodes

“Bridge” nodes connect different communities and play a crucial role in cross‑community diffusion. Identifying and leveraging these bridge nodes can make information propagation more effective.

3. Influence Maximization Problem

The influence maximization problem seeks a set of initial nodes that maximizes the total number of influenced nodes. This problem has wide applications in viral marketing and advertising, where the goal is to select an optimal seed set.

Given a social network G and a diffusion model, the task is to find a seed set S of size k such that the expected number of activated nodes under the model is maximized.

4. Simulation and Analysis

To deepen understanding of information diffusion, we conducted a simulation experiment based on the Independent Cascade Model.

4.1 Network and Parameter Settings

We consider a network of five nodes where each edge has a specified propagation probability. For simplicity we examine a single‑source scenario where information originates from one node.

4.2 Simulation Process

Node 1 is chosen as the initial source. At each step, currently active nodes attempt to activate each neighbor with the given probability. The process continues until no new nodes are activated or a maximum of ten steps is reached.

Example edge‑probability pairs:

<code>import random
import numpy as np

# define edges and propagation probabilities
edges = [(1, 2), (2, 3), (3, 4), (4, 5), (5, 1), (2, 4), (3, 5)]
probabilities = [0.1, 0.2, 0.3, 0.4, 0.1, 0.2, 0.3]
edge_prob = dict(zip(edges, probabilities))

# parameters
num_nodes = 5
initial_node = 1
max_steps = 10
num_experiments = 100

# run simulation experiments
spread_counts = []
for _ in range(num_experiments):
    infected_nodes = {initial_node}
    new_infected_nodes = {initial_node}
    for _ in range(max_steps):
        current_new_infected = set()
        for node in new_infected_nodes:
            for edge, prob in edge_prob.items():
                if edge[0] == node and edge[1] not in infected_nodes and random.random() < prob:
                    current_new_infected.add(edge[1])
        new_infected_nodes = current_new_infected
        infected_nodes.update(current_new_infected)
    spread_counts.append(len(infected_nodes))

# compute average spread
avg_spread = np.mean(spread_counts)
avg_spread, spread_counts[:10]
</code>

An example diffusion process is illustrated below:

4.3 Result Analysis

In our simulation, information starts from node 1. Across 100 runs, the average spread is 1.17, meaning that on average the information reaches only 1.17 nodes. The first ten runs yielded the following spread counts:

This indicates that in most experiments the information remains at the source node, and only in a few cases does it reach other nodes. The limited spread is likely due to the low propagation probabilities set for the edges.

Small propagation probabilities can severely limit diffusion, even in highly connected networks.

Different network structures can lead to varying diffusion outcomes; community structures may create “information islands.”strong>

Choosing different seed nodes can result in different spread sizes and speeds.

Information diffusion in social networks is a multifaceted, dynamic process. By constructing and analyzing diffusion models, we can better understand how information flows and design more effective dissemination strategies.