System 1 vs. System 2 Thinking

Paper · Source
LLM Architecture

Abstract: This paper explores the dual-processing hypothesis of the mind, Systems 1 and 2, by examining debates between cognitive and evolutionary psychologists. I structure the discussion in a back-and-forth manner to emphasize the differences. I show that, while the majority of cognitive psychologists now embrace the dual-processing theory of the mind, Systems 1 and 2, there are still some who disagree. Most evolutionary psychologists, in contrast, dispute the existence of System 2, a domain-general mind, although some disagree. However, a consensus is growing in favor of System 2, although evolutionary psychologists’ concerns must be addressed. The uniqueness of this review is that it contrasts the perspectives of cognitive psychologists with evolutionary psychologists, which is uncommon in the cognitive psychology literature, which tends to overlook evolutionary viewpoints.

Introduction. It is widely accepted that the human mind is specialized for specific domains. But is there a domain-general mind? Cognitive psychologists concur; however, evolutionary psychologists find this notion too pricy to accept. The specialization of different cognitive processes to handle specific types of information or tasks makes the mind domain specific. This means that the mind is not a single, general-purpose processor capable of handling all types of information equally well, but rather a collection of specialized modules, each optimized for a specific type of information or task. Most cognitive psychologists now accept the dual-processing theory of the mind, which states that the mind operates in two separate but interconnected systems: the automatic system and the controlled system. The automatic system is a collection of subsystems known as System 1. It is fast, intuitive, and works unconsciously. It is responsible for processing information that is readily available in our memory and has been repeated several times.

Discussion / Conclusion. Examining the debates between cognitive and evolutionary psychologists, this article discussed the dual-processing hypothesis of the mind, Systems 1 and 2. Using a backand-forth methodology to emphasize the differences, the goal was to show that, while the majority of cognitive psychologists now support the dual-processing theory of the mind, some still disagree. Nevertheless, it appears that System 2 is garnering support. This review is unique in that it contrasts the perspectives of cognitive psychologists and evolutionary psychologists, which is uncommon in the cognitive psychology literature, which tends to put evolutionary perspectives in the background. Most cognitive psychologists agree that there are two mental processes, which Daniel Kahneman popularized as System 1 and System 2. These two systems compete for dominance over our inferences and actions. In evolutionary terms, System 1 predates the other and comprises a self-contained assembly of autonomous subsystems. System 2 enables abstract reasoning as well as the use of hypotheses. System 2 is thus a domain-general processing mechanism. Domain-specific processing mechanisms refer to System 1.

Synthesis notes from this paper's topics 8

This paper has no direct semantic matches yet — here are notes from the topic(s) it's filed under.

Lines of inquiry this paper opens 24

Research framings built by reading the notes related to this paper — the questions it feeds into.

What enables genuine semantic understanding in language models? Why do stronger reasoning capabilities create tradeoffs with instruction following? What compositional reasoning failures limit large language models despite scale? What explains language models' asymmetric difficulty with implicit versus explicit linguistic relations? Can reasoning scale in latent space without tokens? How should inference compute be allocated based on problem difficulty? What training dynamics and scale trigger emergence of reasoning capabilities? Do language models learn genuine understanding or just surface patterns? What role does sparsity play in model behavior and scaling decisions? Can prompt-based context override biases that were embedded during pretraining? How much do training data properties shape model reasoning? How do neural networks achieve compositional generalization at scale? Why does adding new knowledge through fine-tuning degrade existing capabilities? How should designers communicate what AI systems truly are and can do? Can compression size predict model complexity better than parameter count alone?