The evolution and devolution of cognitive control: The costs of deliberation in a competitive world

Damon Tomlin, David G. Rand, Elliot A. Ludvig, Jonathan D. CohenScientific 2015 · DOI

TL;DR

This paper introduces an evolutionary model of cognitive control, showing that while deliberation (controlled processing) can be individually advantageous, its success can lead to population-level dynamics that undermine its prevalence. This suggests that the "costs" of deliberation, like increased competition and environmental changes, can cause controlled processing to decline or oscillate, even going extinct.

What Problem Does This Paper Address?

Dual-system theories of cognition propose that humans have both fast, automatic processes and slower, effortful, deliberative (controlled) processes. While this framework is well-supported in psychology, its implications for large-scale population dynamics haven't been fully explored. The authors aim to bridge this gap by modeling how the interplay between these two cognitive systems, within a competitive environment, shapes their evolutionary trajectory at the population level. They question why humans often exhibit seemingly irrational, short-sighted behavior despite possessing the capacity for rational deliberation, and explore how the evolutionary success of controlled processing might paradoxically lead to its own downfall.

Methods

The core of the paper is a novel agent-based evolutionary simulation. Agents are endowed with a probability, C, of employing controlled processing, otherwise defaulting to automatic processing. The simulation models a resource-consumption paradigm: agents encounter resources, decide whether to consume them immediately or store them for later, and compete with other agents for these resources.

  • Automatic Processing: Characterized by rigid, immediate consumption of any encountered resource, with a strong bias towards present rewards. This process is faster and more efficient in direct competition.
  • Controlled Processing: Characterized by flexible, future-oriented decision-making. Agents use their current energy, stored resources, and environmental estimates to optimize consumption and storage. This process is slower but more adaptive to resource scarcity.
  • Fitness: Agent fitness is directly tied to their energy level, which is influenced by resource consumption, energy expenditure for survival, and the efficiency of their processing strategy.
  • Evolutionary Dynamics: The population evolves using a Wright-Fisher process, where individuals are replaced by offspring proportional to their fitness. Mutations introduce small random changes to an agent's probability of controlled processing (C).
  • Simulation Scenarios: The authors explore two main scenarios:
  • Variable Population Size: Population size (N) co-varies with overall population fitness.
  • Variable Environmental Richness: Environmental resource availability (R) can be influenced by the population's collective energy and controlled processing levels, simulating innovation.

Datasets

Not discussed in the paper. The study relies on simulated data generated by their agent-based model.

Key Insights and Ideas

  • The Paradox of Control: The most striking insight is that the very success of controlled processing can lead to its own decline. By increasing individual fitness and resource availability, controlled processing can inadvertently create an environment where faster, automatic processing becomes more advantageous, leading to a "free-loader" effect.
  • Frequency-Dependent Dynamics: The prevalence of controlled processing is not a simple linear progression. Instead, its spread can trigger complex, non-monotonic dynamics, including oscillations and even extinction, due to interactions with population size, resource availability, and competition.
  • Environmental Sensitivity: The evolutionary outcome of controlled processing is highly sensitive to environmental conditions. In scarce environments, control thrives, but in abundant environments, automaticity can dominate.
  • Competition as a Driver of Devolution: Direct competition for resources, where automatic agents have a speed advantage, can actively push populations away from controlled processing, even if control is otherwise optimal for individual survival.
  • Innovation as a Double-Edged Sword: While controlled processing can drive innovation (leading to increased resource richness), this enrichment can then favor automatic agents, creating a cyclical dynamic.

Key Results

  • Fixed Population Size and Richness: In scenarios with fixed population size (N) and environmental richness (R), higher N and R generally favored automatic processing. However, lower R allowed for a wider range of N where controlled processing was an equilibrium. Intermediate R could lead to stable dual-processing strategies.
  • Variable Population Size:
    • Low Threshold (easy to maintain population size): Small populations initially evolved high control, but as they grew and competition increased, automatic processing re-emerged. Large starting populations with high competition saw control suppressed from the outset.
  • Intermediate Threshold: Populations reached a moderate level of controlled processing, with small populations showing an initial peak and subsequent decline in control.
  • High Threshold (difficult to maintain population size): Only populations starting with high N and sufficient control could survive and evolve. Smaller populations died out.
  • Variable Environmental Richness: Simulations revealed limit cycles where controlled processing would increase resource richness, which in turn would favor automatic processing, leading to a decline in richness and a resurgence of control. This cycle was robust across different thresholds for environmental modification.

Limitations and Future Work

  • Simplified Model: The authors acknowledge that their model simplifies aspects of cognition, such as treating automatic and controlled processing as dichotomous and assuming a fixed probability of control overriding automaticity. They also note that the optimal consumption policy for controlled agents is idealized.
  • Generalizability: While the model focuses on intertemporal choice, the authors suggest that the principles of evolutionary dynamics and the potential for self-undermining success of adaptive traits could apply to other domains of cognition and behavior.
  • Future Directions:
    • More Complex Control: Incorporating more sophisticated control strategies that can adapt to environmental changes, anticipate competition, or leverage social dynamics.
  • Pace of Adaptation: Investigating the interaction between the speed of cognitive adaptation (evolutionary and cultural) and environmental change.
  • Empirical Validation: The authors suggest that their theoretical framework can serve as a benchmark for future empirical studies on the evolutionary trajectories of cognitive processes.
  • Broader Applications: Exploring the implications for domains like diet, drug addiction, savings, and resource management, where the balance between automatic and controlled processing has significant societal consequences.

Summary updated

Abstract

Dual-system theories of human cognition, under which fast automatic processes can complement or compete with slower deliberative processes, have not typically been incorporated into larger scale population models used in evolutionary biology, macroeconomics, or sociology. However, doing so may reveal important phenomena at the population level. Here, we introduce a novel model of the evolution of dual-system agents using a resource-consumption paradigm. By simulating agents with the capacity for both automatic and controlled processing, we illustrate how controlled processing may not always be selected over rigid, but rapid, automatic processing. Furthermore, even when controlled processing is advantageous, frequency-dependent effects may exist whereby the spread of control within the population undermines this advantage. As a result, the level of controlled processing in the population can oscillate persistently, or even go extinct in the long run. Our model illustrates how dual-system psychology can be incorporated into population-level evolutionary models, and how such a framework can be used to examine the dynamics of interaction between automatic and controlled processing that transpire over an evolutionary time scale.
BibTeX
@inproceedings{tomlin2015evolution,
  author = {Damon Tomlin and David G. Rand and Elliot A. Ludvig and Jonathan D. Cohen},
  title = {{The evolution and devolution of cognitive control: The costs of deliberation in a competitive world}},
  booktitle = {Scientific},
  year = {2015},
  doi = {10.1038/srep11002},
}
Figure