Human-like intuitive behavior and reasoning biases emerged in large language models but disappeared in ChatGPT
By Thilo Hagendorff, Sarah Fabi, Michal Kosinski. Nature Computational Science, 2022
298 citations recorded by the source corpus.
We design a battery of semantic illusions and cognitive reflection tests, aimed to elicit intuitive yet erroneous responses. We administer these tasks, traditionally used to study reasoning and decision-making in humans, to OpenAI’s generative pre-trained transformer model family. The results show that as the models expand in size and linguistic proficiency they increasingly display human-like intuitive system 1 thinking and associated cognitive errors. This pattern shifts notably with the introduction of ChatGPT models, which tend to respond correctly, avoiding the traps embedded in the tasks. Both ChatGPT-3.5 and 4 utilize the input–output context window to engage in chain-of-thought reasoning, reminiscent of how people use notepads to support their system 2 thinking. Yet, they remain accurate even when prevented from engaging in chain-of-thought reasoning, indicating that their system-1-like next-word generation processes are more accurate than those of older models. Our findings highlight the value of applying psychological methodologies to study large language models, as this can uncover previously undetected emergent characteristics. The reasoning capabilities of OpenAI’s generative pre-trained transformer family were tested using semantic illusions and cognitive reflection tests that are typically used in human studies. While early models were prone to human-like cognitive errors, ChatGPT decisively outperformed humans, avoiding the cognitive traps embedded in the tasks.
Subfields: Artificial Intelligence, General Decision Sciences
View Paper: https://www.semanticscholar.org/paper/0bfc05adcddd4fe5d1335d96cc313c41526d4558
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
Citations in the corpus
- Multi-Agent Collaboration Mechanisms: A Survey of LLMs (2025, unknown): “While these models have shown remarkable capabilities individually, they still suffer from intrinsic limitations such as hallucination [57], auto-regressive nature (e.g., incapable of slow-thinking [49]), and scaling laws [55, 69].”
- Evaluating large language models in theory of mind tasks (2023, background): “To effectively predict the next word in a sentence generated by a human, LLMs must model not only grammar and vocabulary but also the psychological processes humans use when generating language ( 35 , 36 ).”
- Large language models empowered agent-based modeling and simulation: a survey and perspectives (2023, unknown): “To gauge LLM agents ’ susceptibility to social bias and ste-reotype, researchers use semantic illusions and cognitive re fl ection tests (Hagendorff et al., 2023), typically administered to human subjects, to quantify LLM ’ s tendency to produce intuitive yet erroneous responses.”