A unified theory of ambiguous cognition for humans and LLMs (FLCT) (zenodo.org)

Independent researcher releases FLCT (Fuzzy Latent Cognition Theory), a compact cognitive framework that unifies human and LLM cognition around two core ideas: fuzzy latent representations and biased sampling. Rather than trying to precisely define ambiguous phenomena (color qualia, subjective time, emotion, atmosphere, silence), FLCT models them as probability distributions in high-dimensional latent spaces; cognition is the result of finite, biased extraction paths through those distributions. The author presents a minimal two-equation model that maps how compressed, experience-shaped latent manifolds and context-dependent sampling produce perception, interpretation, and personality in both people and modern LLMs. For the AI/ML community, FLCT reframes ambiguity from a flaw to a computational resource and suggests concrete levers for design and analysis: how embeddings store ambiguous patterns, how sampling algorithms and context shape emergent behavior, and how bias in experience or training data yields characteristic interpretive paths. Practically, this points to new directions in model architecture, sampling strategies, interpretability, evaluation of subjective outputs, and human–AI alignment research. FLCT is positioned as both a scientific perspective on ambiguous cognition and a practical design framework for building AI systems with more human-like, context-sensitive behavior.