Project
Social learning in a structured world

Research Background and Objectives

Social learning—the process of acquiring knowledge, skills, and behaviors by observing others—is fundamental to human development and cultural evolution. Theoretical models predict that payoff-biased social learning (focusing on traits with the highest visible rewards) should be the most effective individual strategy (Mesoudi, 2011; Kendal et al., 2018). However, empirical research shows that payoff bias is often underutilized in real-world contexts (Molleman et al., 2014).

This research project addresses this discrepancy by introducing trait interdependencies as a critical factor. Most cultural evolution models implicitly assume that traits can be learned independently, when in reality, knowledge and skills typically follow hierarchical structures where foundational elements must be mastered before learning advanced ones. This principle is well-established in educational science, where concepts like “learning trajectories,” “scaffolding,” and “zone of proximal development” (Vygotsky, 1978) recognize that effective learning depends on building upon appropriate prerequisites.

Educational institutions themselves can be viewed as cultural adaptations that organize knowledge into structured curricula precisely because traits are interdependent. While educational scientists have long recognized the importance of knowledge structure, formal modeling approaches that connect this insight to cultural evolution theory have been lacking.

Our preliminary analyses suggest that when traits are interdependent, the effectiveness of payoff-biased learning is substantially reduced as individuals attempt to learn advanced traits without mastering prerequisites. This challenges core assumptions in cultural evolution theory and suggests the need for more sophisticated models that incorporate insights from educational science.

The central research question is: How do trait interdependencies shape social learning strategies at individual levels and cultural adaptations at population levels?

Methodological Approach

This project proposes several methodological innovations:

• Markov Chain Modeling Framework: Developing probabilistic models with closed-form solutions to precisely track individuals’ transitions between knowledge states. This analytical approach allows us to derive exact solutions for expected learning outcomes under various social learning strategies across different knowledge structures, providing mathematical precision that complements empirical approaches.
• Graph-Theoretical Representation: Representing trait interdependencies as Directed Acyclic Graphs (DAGs) to analyze how topological features of knowledge structures influence learning outcomes. This approach enables systematic analysis of how different network properties (connectivity, path length, clustering) affect the efficiency of various learning strategies.
• Agent-Based Simulations: Implementing computational models to explore dynamics that resist closed-form solutions, such as adaptive strategy selection, variable population structures, and co-evolutionary dynamics between learning strategies and knowledge structures. These simulations complement the analytical models by addressing questions about emergent behaviors across different timescales.
• Parameter Estimation Methods: Developing techniques to estimate strategy selection parameters, stimulus weights, and strategy persistence from experimental data, providing tools to measure individual differences in social learning and connect theoretical predictions with observed behavior.
• Hierarchical Extensions: Extending the base markov model to include additional dimensions for educational factors such as motivation, feedback effects, and learning difficulties, allowing analysis of how these factors interact with knowledge structure to affect learning outcomes.
• Planned Research Activities
• The project consists of three interconnected components:
• Formal Modeling: Development of mathematical models to analyze how trait interdependencies affect the performance of different social learning strategies and predict population-level outcomes.
• Agent-Based Simulations: Implementing computational simulations to investigate how learning strategies emerge and evolve when individuals can adjust their strategies based on experience.
• Experimental Validation: Designing controlled experiments that manipulate the degree of constraints in learning tasks to measure how structural factors affect strategy selection and application in human participants.

Supervisors
Dr. Lucas Molleman
Dr. Annemie Ploeger
Dr. Ingmar Visser

Financed by
Starting grant

Period
1 September 2024 – 1 March 2029