Mechanisms of Social Learning in Social Contagion and Cultural Evolution

Informacje na temat projektu

SOLAR

Identyfikator umowy o grant: 101042529

DOI

10.3030/101042529

Data podpisania przez KE 25 Października 2022

Data rozpoczęcia 1 Stycznia 2023

Data zakończenia 31 Grudnia 2027

Finansowanie w ramach

• European Research Council (ERC)

Koszt całkowity

€ 1 499 993,00

Wkład UE

€ 1 499 993,00

Koordynowany przez

KAROLINSKA INSTITUTET
Sweden

Periodic Reporting for period 1 - SOLAR (Mechanisms of Social Learning in Social Contagion and Cultural Evolution)

Okres sprawozdawczy: 2023-01-01 do 2025-06-30

Podsumowanie kontekstu i ogólnych celów projektu

Human societies thrive on culture, a powerful force that enables the accumulation and transmission of knowledge across generations. Culture has allowed humans to adapt to diverse environments, innovate technologically, and establish large-scale cooperation. At its core, social learning—the ability to learn from others—drives cultural evolution by shaping how information, behaviors, and beliefs spread within and across populations. However, while it is a fundamental driver of human progress, social learning also plays a central role in the propagation of misinformation, social polarization, and maladaptive behaviors.
Despite its critical role in human cognition and society, we lack a mechanistic understanding of how social learning operates. Fundamental questions remain unanswered: What are the core mechanistic principles that govern social learning? How does social learning fuel social contagion, including the spread of misinformation and harmful ideologies? How do individual learning mechanisms scale up to shape cultural evolution? Answering these questions is not only of theoretical importance but also has profound implications for addressing contemporary challenges in information dissemination, public health, and social cohesion.
The SOLAR project aims to provide a unified, neurocomputational framework that explains social learning from the level of the individual brain to the societal scale. By integrating cognitive neuroscience, reinforcement learning (RL) modeling, and cultural evolution theory, SOLAR seeks to:
• Identify the fundamental mechanisms of social learning by modeling how individuals acquire, adapt, and transmit information from others.
• Understand the role of social learning in social contagion, investigating how beliefs, behaviors, and misinformation spread through networks.
• Uncover how social learning fuels cultural evolution, determining how individual learning mechanisms give rise to cumulative knowledge and societal transformations.
By integrating cognitive neuroscience, artificial intelligence, and social network analysis, SOLAR will map the neurocomputational foundations of social learning and connect them to large-scale social dynamics. Using brain imaging (fMRI), reinforcement learning (RL) models, behavioral experiments, and real-world data from digital platforms, SOLAR will trace how individual learning decisions collectively shape societal trends. This interdisciplinary strategy will not only advance our theoretical understanding of social learning but can also offer practical insights into designing interventions that promote adaptive learning while mitigating the risks of misinformation and harmful social influence.
By bridging individual cognition with societal outcomes, SOLAR has the potential to transform our understanding of social learning and cultural transmission, offering new ways to tackle misinformation, and promote adaptive social behaviors in an increasingly complex world.

Prace wykonane od początku projektu do końca okresu sprawozdawczego oraz najważniejsze dotychczasowe rezultaty

A first series of experiments has uncovered the computational mechanisms that produce social learning (WP1). Contrary to traditional views that suggest individuals rely on fixed heuristics—such as copying the majority—our findings show that social learning is highly adaptive and dynamic. We developed the Social Feature Learning (SFL) model, a reinforcement learning framework that explains how individuals learn which social cues—such as others' success or popularity—are most predictive of rewards. Across a series of behavioural experiments, we confirmed that people adjust their reliance on different social cues based on prior experiences. Agent-based simulations further demonstrate that this simple learning mechanism can generate key social learning strategies observed across human societies. Using brain imaging, we are identifying the neural mechanisms underlying the SFL model, and test the hypothesis that reward-sensitive regions reflect the flexibly updated value of social information.
A second research program investigates how social learning fuels social contagion (WP2). Using computational modelling and behavioural experiments, we tested how individuals update their reliance on social signals in simulated and real networks. Simulations show that viral spread of ideas and misinformation emerges from individuals dynamically adjusting their social learning based on feedback, rather than passively copying others, as explained by the SFL model. Based on this model, we will design and experimentally test interventions to halt social contagion, as well as analyze real social media trace data. This work provides a mechanistic account of why certain behaviours and beliefs spread rapidly while others fade, with potential applications for public health communication, misinformation control, and digital platform design.
Finally, SOLAR examines role of language in social learning and cultural evolution (WP3). In a first large-scale study, we tested whether language allows individuals to pass down mental models of the world more effectively than observational learning. Participants engaged in a multi-generational experiment, where knowledge was transmitted either through written communication or direct observation. As predicted, language-based transmission led to increasing accuracy and complexity of shared knowledge across generations. Using machine learning analyses, we found that linguistic communication refined and optimized information transmission over time, providing strong evidence that language is a key driver of cumulative cultural evolution. In another series of experiments, we are identifying the computational mechanisms involved in innovation, a key driver of cultural evolution, and how these mechanisms are expressed in language-mediated social learning. Further studies will probe the underlying brain mechanisms of language-based social learning and innovation.

Innowacyjność oraz oczekiwany potencjalny wpływ (w tym dotychczasowe znaczenie społeczno-gospodarcze i szersze implikacje społeczne projektu)

This project has achieved significant breakthroughs across work packages WP1–WP3, advancing our understanding of social learning, cultural transmission, and social contagion. A notable achievement is the Social Feature Learning (SFL) model developed in WP1, which shifts the perspective from fixed heuristics in social learning to a dynamic, reward-based strategy. This model not only provides a mechanistic explanation for diverse learning strategies but also has implications across fields such as cultural evolution, biology, and artificial intelligence.
In WP1 and WP2, agent-based simulations using the SFL model illustrate that social contagion results from individual learning mechanisms. This understanding could have applications in misinformation research, digital communication, and policy-making for online platforms. WP3's exploration of language in cultural transmission highlights the cumulative evolution of knowledge, showing that language enhances knowledge transfer across generations more effectively than observation alone.

To ensure the successful uptake of these findings, several key needs should be addressed. First, further research is necessary to refine the SFL model and explore its applications in various contexts. Additional efforts will focus on validating computational models with real-world data, particularly in digital social networks. In the long run, collaboration with policymakers, educators, and AI researchers will be crucial for translating these findings into practical applications.