Why complex systems are more than just the sum of their parts
In short: How do individual choices create massive economic shifts or biological patterns? Agent-based models (ABMs) move beyond simple equations to simulate autonomous entities that learn, interact, and adapt. By focusing on micro-level behaviors, these models reveal how aggregate phenomena—like market crashes or social trends—emerge from the bottom up.
Unlike traditional economic models that assume perfectly rational actors, ABMs are built on the principle of bounded rationality. In these simulations, agents—which could represent consumers, firms, or even biological organisms—possess imperfect information and make decisions based on limited reasoning and heuristics. Because these agents interact with each other and their environment, they create feedback loops. These dynamics mean the system does not settle into a steady state equilibrium; instead, it evolves, potentially producing non-linear 'boom and bust' cycles that more accurately mirror real-world complexity.
Historically, ABMs have struggled to bridge the gap between theory and data. While they are excellent for building intuition, they often lack the ability to estimate 'micro' variables—the specific, hidden traits of individual agents—making accurate forecasting difficult. Most previous methods relied on 'parameter calibration,' where modelers adjusted a few high-level knobs to match summary statistics like average prices.
However, this can hide underlying patterns, as different internal dynamics can produce the same outward average. Recent breakthroughs involve translating ABMs into probabilistic models to learn these latent variables directly from data. For example, in a housing market study, researchers could observe neighborhood prices (public data) while using gradient-based algorithms to estimate the hidden distribution of agent incomes. By updating these latent variables at each time step, the model can track empirical trends more closely. This shift from 'black-box' data fitting to learning internal agent states allows for more robust out-of-sample forecasting and a deeper understanding of how individual behaviors drive systemic change.
Opens the Intellect+ feed with this card first. Swipe up for the next one.
Sources
- Summary by Intellect · source: On learning agent-based models from data - Scientific Reports
- Also: Agent Based Modelling Comes of Age
- Also: Agent-based model
To cite this card, link to https://intellectplus.co.uk/app/health-agent-based-model.
How we source, check and correct cards: Editorial standards.
General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.
Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?
Related
More in Science · Latest · A–Z · All topics
