In Schelling's segregation model, each agent prefers only 30% of neighbors to share its group. The model produces near-total segregation. What does this demonstrate about ABMs?
Think about your answer, then reveal below.
Model answer: It demonstrates emergence — macro-level patterns that are more extreme than any individual agent's preferences or intentions. The segregation outcome cannot be derived by reading the individual rules; it arises from the interaction of many agents over time.
This is the signature insight of agent-based modeling: micro rules produce macro patterns that often surprise us. The gap between individual intent and collective outcome is precisely what ABMs are designed to reveal.
Question 2 Short Answer
A researcher builds an ABM of opinion polarization. She tunes parameters until the model matches the 2020 US political landscape perfectly. Is this validation? Why or why not?
Think about your answer, then reveal below.
Model answer: No. Fitting parameters to reproduce one observed outcome is not validation — it is curve-fitting. Validation requires checking whether the model reproduces other empirical patterns it was not fitted on, testing robustness across parameter ranges, and grounding agent rules in independent theoretical justification.
The risk in ABMs is overfitting to the historical case. A well-validated ABM should generate insights that transfer to new cases and survive parameter perturbation.