Dual Traits in Probabilistic Reasoning of Large Language Models
This research addresses cognitive biases in LLMs, which is crucial for improving their reliability in critical applications, though it is incremental in nature.
The study investigated how large language models evaluate posterior probabilities, revealing two coexisting modes: a normative mode adhering to Bayes' rule and a representative-based mode relying on similarity, similar to human cognitive systems, with models struggling to recall base rate information.
We conducted three experiments to investigate how large language models (LLMs) evaluate posterior probabilities. Our results reveal the coexistence of two modes in posterior judgment among state-of-the-art models: a normative mode, which adheres to Bayes' rule, and a representative-based mode, which relies on similarity -- paralleling human System 1 and System 2 thinking. Additionally, we observed that LLMs struggle to recall base rate information from their memory, and developing prompt engineering strategies to mitigate representative-based judgment may be challenging. We further conjecture that the dual modes of judgment may be a result of the contrastive loss function employed in reinforcement learning from human feedback. Our findings underscore the potential direction for reducing cognitive biases in LLMs and the necessity for cautious deployment of LLMs in critical areas.