cognitive-science

Introduces PM-Bench, a text-based benchmark for evaluating prospective memory in LLM agents, inspired by cognitive science. Experiments show that even the best method (GPT-5.4 agent) achieves only 65.1% F1 score, indicating significant challenges in reliable intention execution.

The author recounts a two-day experiment using GPT-5.6 as a cognitive crutch and translator for self-understanding, discovering a productive loop of representation, resistance, correction, and reconstruction that may represent an early prototype of human-LLM symbiosis.

Adaptive Recall is a memory system for AI assistants that uses cognitive science and machine learning to improve retrieval quality over time. It offers an API with multiple retrieval strategies, knowledge graph construction, and self-improving capabilities.

Blogger real屠龙的胭脂井, in her sharing at Fudan University, deeply analyzed the underlying logic of brand marketing, emphasizing the use of her background in neuroscience and cognitive science to support decision-making through modeling and data in the AI era, and criticized the misconception of marketing professionals relying on vanity metrics.

This paper argues that Elan Barenholtz's autogenerative theory of language, developed from LLM behavior, enriches Roy Harris's integrationist linguistics by providing structural mechanisms for prospective openness and the continuity of linguistic and non-linguistic signs, offering explanatory content for computational approaches.

Analysis of 2.6 billion sketches from 236 countries reveals hidden cultural variation in how common concepts are visually represented, showing that visual imagery preserves rich semantic and cultural structure that language models compress.

A Peking University study published in Cell Reports finds that the human brain can adapt to virtual non-human limbs (wings) through VR training, showing dynamic plasticity in the occipitotemporal cortex.

The article examines the link between working memory and consciousness, discussing how working memory's limitations and the 'doorway effect' might contribute to our conscious experience.

This paper extends Herbert Simon's bounded rationality to moral cognition, formalizing a tradeoff between moral breadth and moral depth for finite agents. It argues that ethical theories are locally optimal strategies within a constrained space, with implications for AI alignment.

NeuroCogMap is a cognitive neuroscience-inspired framework that maps the internal features of large language models into functional parcels, linking them to interpretable cognitive functions and revealing signatures of model failures like hallucination and bias.

This paper investigates the emergence of situation modeling and mentalizing abilities in transformer language models across training stages, finding that false belief task performance depends on model size and training volume, emerges late in pretraining, and shows fragility with non-factive verbs.

This paper presents an information-processing theory of consciousness and argues that instantiating conscious subsystems in AI could enable superior adaptation without extensive training, potentially leading to AGI.

auto-psych is an agent-based system that automates theory discovery and experimentation in computational cognitive science, using LLM agents to generate hypotheses, design experiments, and analyze data from crowdsourced participants. It demonstrates faster and better theory generation compared to human-derived theories in a classic psychology paradigm.

An essay exploring why thinking out loud with another person produces better understanding and insight than solitary reflection, drawing on cognitive science and philosophy.

This paper examines whether language models can independently discover the concept of zero as a form of out-of-distribution generalization, finding that GPT-2 sized models cannot at test time but improve with training on examples of zero, and that language pretraining reduces the number of required examples.

This article delves into the concept of embodied intelligence, its intellectual origins (philosophy, cognitive science, AI robotics), and historical development (the failure of symbolism and Brooks' subsumption architecture). It analyzes its differences from pure software AI and the challenges it faces.

This paper argues that recent claims that neural networks have solved Fodor and Pylyshyn's systematicity challenge are premature. The authors show that the meta-learning for compositionality model fails to generalize out-of-distribution and behaves unsystematically even on in-distribution problems, concluding the challenge remains unmet.

The paper introduces the Theory of Mind Utility (ToM-U), a formal computational-level specification for inferring others' epistemic states by constructing Local Epistemic World Models (LEWMs). It differs from Bayesian ToM and simulation theory by providing a domain-agnostic mechanism for belief inference without commitment to algorithmic implementation.

The article discusses a deficiency in executive control within transformer attention mechanisms, highlighting limitations in how transformers manage sequential dependencies.

This paper develops a geometric dynamical framework to model how predictive AI assistance alters exploratory cognition by stabilizing trajectories before self-generated exploration, leading to reduced exploratory responsiveness, hysteresis, and delayed recovery upon assistance withdrawal.