Schelling Models in Social Dynamics

The Schelling model is a landmark agent-based framework elucidating how individual preferences can drive large-scale patterns of segregation or integration. In its basic form, agents of two types occupy sites on a grid or network and relocate if the share of like-type neighbours falls below a personal tolerance threshold. Despite the simplicity of its rules, the model exhibits tipping points that lead to either homogeneous clusters or mixed configurations, highlighting the tension between local preferences and global outcomes. Extensions have incorporated diverse topologies, strategic decision-making, learning algorithms and additional variables such as income or emotional attachment to capture richer social phenomena. Analysis of its dynamics has drawn on tools from statistical physics (coarsening and phase transitions), game theory (equilibria and efficiency losses) and machine learning (reinforcement learning). The model’s versatility has spurred applications to residential segregation, opinion formation, wealth distribution and policy intervention, demonstrating its enduring impact across sociology, economics, geography and computational social science.

Research from Nature Portfolio

Recent studies have augmented the classical framework with adaptive learning and individual ageing. One investigation integrated deep Q-networks into the relocation decision, endowing agents with reinforcement learning to balance segregation preferences against rewards for diverse interactions. This approach revealed that interdependent reward structures can foster spatial integration even when agents possess a segregation bias. Another study introduced an ageing mechanism whereby agents’ likelihood to move diminishes the longer they remain satisfied in a location. This attachment effect eradicates the sharp transition between segregated and mixed phases found in the original model and yields a slow coarsening process akin to glassy dynamics, indicating that emotional bonds can stabilise diverse neighbourhoods over time.

Schelling Models in Social Dynamics publication trend

The graph below shows the total number of articles in schelling models in social dynamics across all publications each year (not limited to Nature Index journals).

Technical terms

Agent-based model: Computational framework in which individual entities interact according to predefined rules.

Segregation threshold: Proportion of similar neighbours required for an agent to be satisfied.

Utility function: Quantitative measure of an agent’s satisfaction with its local environment.

Reinforcement learning: Machine-learning method where agents adapt actions based on rewards received.

Entropy: Measure of disorder or heterogeneity within a modelled system.

Price of Anarchy: Ratio comparing the efficiency of the worst-case equilibrium to the optimal social outcome.

References

1. Segregation dynamics with reinforcement learning and agent based modeling. Scientific Reports (2020).
2. Aging effects in Schelling segregation model. Scientific Reports (2022).
3. Welfare Guarantees in Schelling Segregation. Journal of Artificial Intelligence Research (2021).
4. Modified Schelling games. Theoretical Computer Science (2021).
5. Topological influence and locality in swap schelling games. Autonomous Agents and Multi-Agent Systems (2022).
6. Incorporating a monetary variable into the Schelling model addresses the issue of a decreasing entropy trace. Scientific Reports (2020).
7. Dynamics of Transformation from Segregation to Mixed Wealth Cities. PLOS ONE (2016).