#55 Scott Page: Becoming a Model Thinker

Episode Summary

Executive Summary: Scott Page argues that complex modern problems are rarely solvable by one person or one model; progress comes from ensembles of diverse people and diverse mental models. The conversation covers how models structure reality, how they surface assumptions, why diversity improves decision-making, and how to choose, combine, and teach models across domains from finance to policy to child education.

Main Topics: Mental models as frameworks for making sense of reality (Priority: 5/5): Page defines mental models as frameworks for interpreting the world and argues that modeling is the act of mapping messy reality into clean mathematical/logical structures. Collective intelligence and cognitive diversity (Priority: 5/5): The episode emphasizes that no individual brain can contain enough complexity to solve large systemic problems; groups with diverse models can collectively do better than any one expert. Data, information, knowledge, and wisdom hierarchy (Priority: 5/5): Page explains a hierarchy where data becomes information, information becomes knowledge, and wisdom is choosing the right knowledge or model for a specific problem. Choosing and combining models for different contexts (Priority: 4/5): He discusses how to identify whether a problem is a decision, game, or social process, and how model choice changes based on rationality, feedback, and strategic interaction. Specific model examples: power laws, convexity/concavity, and local interactions (Priority: 4/5): The interview walks through key models from the book, showing how they explain phenomena like winner-take-all outcomes, diminishing returns, and cultural coordination. Education, teaching, and building model literacy (Priority: 4/5): Page argues schools overemphasize single-model instruction and should teach meta-thinking, cross-domain structure, and perspective-taking to better prepare students for complexity. Institutions, algorithms, and redesigning systems (Priority: 4/5): The discussion extends model thinking to governance and organizations, suggesting algorithms may need to be treated as a new institutional form alongside markets, hierarchies, and democracies.

Key Arguments: Complexity makes single-model thinking insufficient; ensembles of models and people produce better understanding than isolated experts. Mental models are useful because they force explicit assumptions and reveal when a theory does not fit reality. The move from data to information to knowledge is largely a modeling act; wisdom is selecting which models to apply and when. Crowds can outperform linear models when humans incorporate variables the model cannot capture (e.g., a product being "but ugly"). Model disagreement is informative: if humans and a linear model diverge sharply, the gap should be investigated rather than averaged blindly. Power laws emerge from multiple mechanisms such as preferential attachment, random walks, and self-organized criticality, and they imply large inequality or winner-take-all dynamics. Concavity and convexity matter because many real systems have diminishing returns or positive feedback, making linear extrapolation dangerous. Local interaction models explain culture and coordination: many behaviors are just settled conventions that persist because people adapt to each other. Education should teach structure, logic, and function across domains so learners can recognize recurring patterns rather than memorize isolated formulas. In complex societies, success often depends not only on individual talent but on filling a useful niche, connecting domains, or coordinating across perspectives.

Data Points: Models in the book: 30 - Page says The Model Thinker presents about 30 important models readers can understand and use. Book chapter length per model: 7 to 12 pages - He describes each model chapter as a concise accessible primer with math placed in boxes. Crowd/team vs individuals: People can beat linear models when the model misses key variables - He cites studies where groups outperform regression by capturing factors the model cannot include. Printer forecast example: 400,000 vs 200,000 units - A linear model predicted 400,000 printer sales, while a crowd predicted 200,000 based partly on subjective design judgment. Wisdom hierarchy: 4 layers - Data, information, knowledge, wisdom are presented as the hierarchy for interpreting and applying models. Class exercise grid size: 100 by 100 - In a collective intelligence experiment, students searched for maxima on 100x100 value grids. Collective intelligence teams: 3 - He compared physicists, decentralized market participants, and waggle-dancing bees in the grid experiment. Search rounds in experiment: 5 rounds of 10 points - The physicist team was allowed structured probing across five rounds to locate the optimum. Child/teen incentives example: 30 minutes - Page notes his children worked hard on Rush Hour puzzles for 30 minutes of iPad time. Stock market / company understanding: 30% to 90% missing variation - He critiques social science models that explain only a fraction of variation, leaving most of reality unexplained.

Pivotal Quotes: "You yourself are not going to sort of solve the obesity epidemic... But collections of people by creating a larger ensemble model actually have a hope of addressing these problems." — Scott Page: Central thesis on the limits of individual cognition and the value of collective intelligence. "What wisdom is, is wisdom is understanding which knowledges to bring to bear on a particular problem." — Scott Page: Definition of wisdom in the data-information-knowledge-wisdom hierarchy. "Imagine how difficult physics would be if electrons could think." — Scott Page: Used to show how systems get harder when the actors inside them can adapt, strategize, and change the rules.

Implications: Listeners should think less like lone experts and more like model curators: build a diverse toolkit, know when to defer to others, and use perspective-taking to improve decisions. For organizations, the future favors interdisciplinary teams and better institutional design, including how algorithms shape choices.