The Model Thinker

1. Models Simplify Reality to Reveal Truths

It can scarcely be denied that the supreme goal of all theory is to make the irreducible basic elements as simple and as few as possible without having to surrender the adequate representation of a single datum of experience.

Simplification is key. Models are not meant to be perfect replicas of the world, but rather simplified representations that highlight key relationships and causal forces. By stripping away unnecessary details, models allow us to focus on the essential elements that drive a phenomenon. This simplification enables logical analysis and the generation of testable hypotheses.

Three classes of models. Models can be simplifications of the world, mathematical analogies, or exploratory, artificial constructs. Regardless of their form, models must be tractable, meaning they are simple enough to allow for logical analysis. This tractability allows us to derive insights, generate hypotheses, and design solutions.

All models are wrong. Because models simplify, they necessarily omit details and make assumptions that are not perfectly true. However, this does not invalidate their usefulness. By considering many models, each with its own simplifications and assumptions, we can gain a more comprehensive understanding of complex phenomena.

2. Many Models Offer Wisdom Through Diverse Lenses

To become wise you’ve got to have models in your head. And you’ve got to array your experience—both vicarious and direct—on this latticework of models.

Wisdom through multiplicity. The many-model thinking approach advocates using an ensemble of models to make sense of complex phenomena. This approach builds on the idea that wisdom is achieved through a multiplicity of lenses, allowing us to see the world from different angles and appreciate its inherent complexity.

Counter to tradition. The many-model approach runs counter to the traditional approach of relying on a single model for a given problem. While single models can be useful, they are inherently limited by their assumptions and simplifications. By engaging with many models, we can overcome these limitations and build a more nuanced understanding.

Building a lattice of models. By learning and applying a variety of models from different disciplines, we can begin to build our own lattice of understanding. This lattice allows us to see how causal processes overlap and interact, creating the possibility of making sense of the complexity that characterizes our economic, political, and social worlds.

3. Data Abundance Demands Model-Based Thinking

We’ve got facts, they say. But facts aren’t everything; at least half the battle consists in how one makes use of them!

Data is not a panacea. While the era of big data provides unprecedented access to information, it is not a substitute for critical thinking and model-based reasoning. Data alone cannot tell us why something happened or what will happen in the future.

Models organize and interpret data. We need models to make sense of the fire-hose-like streams of data that cross our computer screens. Models provide a framework for organizing and interpreting data, allowing us to identify patterns, test hypotheses, and make predictions.

Models improve decision-making. Without models, people suffer from cognitive shortcomings, such as overweighting recent events and ignoring base rates. Models clarify assumptions, promote logical thinking, and leverage big data to test causal claims, leading to better decisions.

4. Rationality, Rules, and Adaptation Guide Human Modeling

It is not possible yet to point to a single theory of human behavior that has been successfully formulated and tested in a variety of settings.

Modeling human behavior is challenging. People are diverse, socially influenced, error-prone, purposive, adaptive, and possessed of agency. These characteristics make it difficult to create simple, universal models of human behavior.

Three approaches to modeling people. We can model people as rational actors who make optimal choices, as rule-based actors who follow fixed or adaptive rules, or as a combination of both. The choice of approach depends on the context and the goals of the model.

Rationality as a benchmark. Even if people do not always act rationally, the rational-actor model provides a useful benchmark for evaluating behavior and designing policies. It allows us to identify potential inefficiencies and to understand how people might respond to incentives.

5. Distributions Quantify Variation and Diversity

Perhaps the truth depends on a walk around the lake.

Distributions capture variation. A distribution mathematically captures variation (differences within a type) and diversity (difference across types) by representing them as probability distributions defined over numerical values or categories.

Normal distributions arise from sums. The central limit theorem explains the prevalence of normal distributions, stating that the sum of many independent random variables will be approximately normally distributed. This explains why heights, weights, and test scores often follow a bell curve.

Distributions inform decisions. Knowing the shape of a distribution allows us to make better predictions, design more effective policies, and take more informed actions. For example, understanding the distribution of earthquake sizes allows us to prepare for the likelihood of large events.

6. Power Laws Highlight the Impact of Large Events

Therefore, we attempt to treat the same problem with several alternative models each with different simplifications but with a common biological assumption. Hence, our truth is the intersection of independent lies.

Power laws include large events. Power-law distributions, also known as long-tailed distributions, are characterized by a large number of small events and a few very large events. These distributions are common in social and natural phenomena, such as city sizes, book sales, and earthquake magnitudes.

Feedbacks create power laws. Power laws arise from non-independence, often in the form of positive feedbacks. The Matthew effect, where those who have more also receive more, is a key driver of power-law distributions.

Power laws impact equity and risk. Long-tailed distributions can lead to increased inequality, as a few individuals or entities capture a disproportionate share of the rewards. They also increase the risk of catastrophic events, requiring us to prepare for the possibility of extreme outcomes.

7. Linear Models Offer a Foundation for Understanding Relationships

Logic takes care of itself; all we have to do is to look and see how it does it.

Linearity simplifies analysis. Linear models assume a constant relationship between variables, making them easy to estimate and interpret. They provide a starting point for understanding complex phenomena.

Regression reveals relationships. Linear regression is a statistical technique for fitting a line to data, revealing the sign, magnitude, and significance of the relationship between variables.

Linearity has limits. While linear models can be useful, they are often limited by their simplicity. Many real-world relationships are nonlinear, requiring more sophisticated models to capture their complexity.

8. Nonlinear Models Capture Complex Dynamics

Knowing reality means constructing systems of transformations that correspond, more or less adequately, to reality.

Nonlinearity reflects reality. Nonlinear models capture relationships where the effect of one variable on another is not constant. These models can exhibit diminishing returns, increasing returns, thresholds, and other complex behaviors.

Concavity and convexity shape outcomes. Concave functions exhibit diminishing returns and lead to risk aversion, while convex functions exhibit increasing returns and lead to risk-seeking behavior.

Models reveal conditions. By exploring nonlinear models, we can uncover the conditions under which certain outcomes are likely to occur. For example, growth models reveal the importance of innovation for sustaining long-term economic growth.

9. Value and Power Arise from Position and Contribution

The great end of life is not knowledge but action.

Value and power are measurable. Models can help us quantify the value and power of individual actors within a system. These measures can be used to understand how resources are allocated and how decisions are made.

Shapley value measures contribution. The Shapley value is a measure of an actor's average marginal contribution to a group, providing a way to assess their importance to the collective.

Models guide action. By understanding the sources of value and power, we can design institutions and policies that promote efficiency, fairness, and other desired outcomes.

10. Networks Connect and Constrain Systems

Plurality must never be posited without necessity.

Networks are ubiquitous. Networks are a fundamental feature of many systems, connecting people, organizations, and entities in complex webs of relationships.

Network structure matters. The structure of a network, as measured by degree, path length, clustering coefficient, and community structure, can have a profound impact on its behavior.

Models reveal network effects. Network models can help us understand phenomena such as the friendship paradox, the six degrees of separation, and the spread of information and influence.

11. Contagion Models Explain Spread and Influence

The supreme goal of all theory is to make the irreducible basic elements as simple and as few as possible without having to surrender the adequate representation of a single datum of experience.

Contagion models capture diffusion. Models of broadcast, diffusion, and contagion help us understand how information, technologies, behaviors, beliefs, and diseases spread throughout a population.

Transmission shapes adoption. The shape of the adoption curve, whether r-shaped or S-shaped, depends on the mode of transmission, with broadcast models producing r-shapes and diffusion models producing S-shapes.

Models guide interventions. By understanding the dynamics of contagion, we can design more effective interventions to promote desirable outcomes, such as vaccination strategies or the adoption of new technologies.

12. Mechanism Design Shapes Institutional Outcomes

Logic takes care of itself; all we have to do is to look and see how it does it.

Institutions structure interactions. Institutions, such as markets, voting systems, and hierarchies, provide a framework for communication, decision-making, and resource allocation.

Mechanism design optimizes institutions. Mechanism design is a framework for designing institutions that achieve desired outcomes, taking into account the incentives and information of the participants.

Models reveal trade-offs. By applying mechanism design principles, we can identify the trade-offs inherent in different institutional designs and choose the mechanisms that best achieve our goals.

Last updated: March 1, 2025