agent-based model
A computational economic model that simulates the actions and interactions of many individual agents to study emergent macro-level behavior.
As defined in AMA | November 2024
Where it comes up · 5
- 293 | Doyne Farmer on Chaos, Crashes, and Economic Complexity7 claims
- 161 | W. Brian Arthur on Complexity Economics4 claims
- AMA | November 20243 claims
- 330 | Petter Törnberg on the Dynamics of (Mis)Information2 claims
- Man + Machine, Moats, and Power of the Outside View1 claim
Most important claims · 10 of 17Search all claims for “agent-based model” →
- An agent-based model using LLMs with survey-derived personas on a minimal social media platform—with no engagement algorithm—spontaneously produced all three pathologies of real social media: echo chambers (Democrats and Republicans not following each other), power-law attention inequality, and the 'social media prism' where more extreme/polarized users gain more attention.330 | Petter Törnberg on the Dynamics of (Mis)Information
- In a model where agents do nothing but copy the savings rate of the neighbor currently consuming the most, the economy spontaneously starts oscillating (as long as agents don't update too often) yet the agents collectively come within a few percent of the optimal savings rate, because selection toward the highest consumer drives the result even though the agents are absolute dummies.293 | Doyne Farmer on Chaos, Crashes, and Economic Complexity
- LLM-based agents are more realistic than rule-based agents but much harder to validate, sitting awkwardly between high-validity empirical methods and parsimonious formal models; the way to use them responsibly is to ignore the text they produce and instead study robust structural outcomes that emerge from their behavior, trusting results in proportion to that robustness.330 | Petter Törnberg on the Dynamics of (Mis)Information
- Complexity economics uses agents with simple or complex decision rules (learning algorithms, heuristics like buying undervalued assets or imitating the best, or trial and error) that are not assumed optimal; information flows in, agents decide, economic consequences are computed, generating new information, and the loop repeats—so equilibrium, if it emerges at all, is an emergent property rather than an assumption, and the formalism more naturally captures endogenous dynamics like business cycles.293 | Doyne Farmer on Chaos, Crashes, and Economic Complexity
- Agent-based models can start extremely simple—even with zero-intelligence agents flipping coins or orders arriving at random like sand grains—and still make useful predictions, such as how the bid-ask spread in a financial market depends on the flow of orders into the market.293 | Doyne Farmer on Chaos, Crashes, and Economic Complexity
- Farmer's group built a disequilibrium agent model that advised the British government on lockdown economics by assuming an industry cannot produce without demand, inputs, and labor; initialized at the pre-pandemic steady state and hit with shocks (inferred from Bureau of Labor Statistics occupational data on how closely people work), it tracked shocks reverberating dynamically through upstream-downstream supply chains—effects not felt immediately—which equilibrium-from-the-start mainstream models cannot capture.293 | Doyne Farmer on Chaos, Crashes, and Economic Complexity
- Complexity economics applies the methods and epistemology of complex systems science to economics, which is fundamentally different from and much more explicitly complex-systems-based than how mainstream economists do their work.293 | Doyne Farmer on Chaos, Crashes, and Economic Complexity
- Doyne Farmer's agent-based economic modeling could force a paradigm shift in economics analogous to Kepler refining Copernicus—replacing the 'perfect circles' of equilibrium assumptions with more accurate dynamics—and if correct will eventually accrue an undeniable weight of empirical success.AMA | November 2024
- In Kristian Lindgren's prisoner's-dilemma tournament where strategies compete, replicate, mutate, and randomly 'deepen' (look further back in history), no final equilibrium strategy is ever reached: any apparent solution is disrupted when a strategy deepens to use more information, sending the system into volatility, so a solution is never quite reached because strategies can always deepen further.161 | W. Brian Arthur on Complexity Economics
- Doyne Farmer claimed classical economic models 'don't make business cycles unless they're getting kicked,' yet also described initializing his agent model in a pre-pandemic steady state and then hitting it with shocks—which appear to be the same mechanism (external perturbation driving dynamics), raising the question of what distinguishes them.AMA | November 2024