Reasoning on a Spectrum: Aligning LLMs to System 1 and System 2 Thinking

Alireza Salkhordeh Ziabari, Nona Ghazizadeh, Zhivar Sourati, Farzan Karimi Malekabadi, Payam Piray, Morteza Dehghani

Keywords: Alignment, System 1 and System 2 thinking, Cognitive heuristics, LLM, NLP

Abstract: Large language models (LLMs) demonstrate remarkable reasoning capabilities, yet their reliance on step-by-step reasoning can make them brittle when tasks do not align with such structured approaches. In contrast, human cognition flexibly alternates between fast, intuitive reasoning (System 1) and slow, analytical reasoning (System 2), depending on context. To bridge this gap, we curate a dataset of 2K examples, each with valid responses from both reasoning styles, and explicitly align LLMs with System 1 and System 2 reasoning. Evaluations across diverse reasoning benchmarks reveal an accuracy-efficiency trade-off: System 2-aligned models excel in arithmetic and symbolic reasoning, while System 1-aligned models perform better in commonsense tasks. A mechanistic analysis of model responses shows that System 1 models employ more definitive answers, whereas System 2 models demonstrate greater uncertainty. Interpolating between these extremes produces a monotonic transition in reasoning accuracy, preserving coherence. This work challenges the assumption that step-by-step reasoning is always optimal and highlights the need for adapting reasoning strategies based on task demands.

Primary Area: Neuroscience and cognitive science (e.g., neural coding, brain-computer interfaces)

Submission Number: 26250