An essential problem in artificial intelligence is whether LLMs can simulate human cognition or merely imitate surface-level behaviors, while existing datasets suffer from either synthetic reasoning traces or population-level aggregation, failing to capture authentic individual cognitive patterns. We introduce a benchmark grounded in the longitudinal research trajectories of 217 researchers across diverse domains of artificial intelligence, where each author's scientific publications serve as an externalized representation of their cognitive processes. To distinguish whether LLMs transfer cognitive patterns or merely imitate behaviors, our benchmark deliberately employs a cross-domain, temporal-shift generalization setting. A multidimensional cognitive alignment metric is further proposed to assess individual-level cognitive consistency. Through systematic evaluation of state-of-the-art LLMs and various enhancement techniques, we provide a first-stage empirical study on the questions: (1) How well do current LLMs simulate human cognition? and (2) How far can existing techniques enhance these capabilities?
Abstract In this study, we compared performance across four behavioural tasks in the same fish species (Rhinecanthus aculeatus), with one group held in long-term laboratory captivity and the other recently caught and temporarily housed at a field station laboratory. The aims of this study were twofold: first, to test whether captivity conditions influence performance in commonly used behavioural and cognitive assays; and second, to evaluate whether any of these tasks could serve as a practical tool for screening behavioural changes over time. The four tests used were a Novel Object Test, Puzzle Preference Test, Emergence Test, and Cylinder Test. We found that recently caught fish were generally more exploratory and more responsive to novel stimuli; however, their responses were object-specific, with increased neophobia towards some objects. Long-term captive fish were more variable in their responses across all tests. In the Emergence Test, long-term captive fish emerged faster but showed greater individual variability. In the Cylinder Test, all recently caught fish failed to swim around a transparent cylinder, whereas several long-term captive individuals showed possible evidence of inhibitory control. Our results demonstrate that captivity conditions can influence performance in behavioural tests at both group and individual levels. These findings have important implications for comparative cognition studies, particularly when interpreting results collected across different laboratory settings or captivity durations, even when working with the same species. Of the tasks used, the Emergence Test was identified as the most practical assay for tracking the effects of captivity on behaviour, as it was highly sensitive to individual differences and straightforward to run and analyse.
Human cognition is constrained by processing limitations, leading to cognitive overload and inefficiencies in knowledge synthesis and decision-making. Large Language Models (LLMs) present an opportunity for cognitive augmentation, but their current reactive nature limits their real-world applicability. This position paper explores the potential of context-aware cognitive augmentation, where LLMs dynamically adapt to users' cognitive states and task environments to provide appropriate support. Through a think-aloud study in an exhibition setting, we examine how individuals interact with multi-modal information and identify key cognitive challenges in structuring, retrieving, and applying knowledge. Our findings highlight the need for AI-driven cognitive support systems that integrate real-time contextual awareness, personalized reasoning assistance, and socially adaptive interactions. We propose a framework for AI augmentation that seamlessly transitions between real-time cognitive support and post-experience knowledge organization, contributing to the design of more effective human-centered AI systems.
The development of large language models (LLMs) is limited by a lack of explainability, the absence of a unifying theory, and prohibitive operational costs. We propose a neuro-theoretical framework for the emergence of intelligence in systems that is both functionally robust and biologically plausible. The model provides theoretical insights into cognitive processes such as decision-making and problem solving, and a computationally efficient approach for the creation of explainable and generalizable artificial intelligence.
Vision-language models (VLMs) aligned with general human objectives, such as being harmless and hallucination-free, have become valuable assistants of humans in managing visual tasks. However, people with diversified backgrounds have different cognition even in the same situation. Consequently, they may have personalized expectations for VLM assistants. This highlights the urgent need to align VLM assistants with personalized situated cognition for real-world assistance. To study this problem, we first simplify it by characterizing individuals based on the sociological concept of Role-Set. Then, we propose to evaluate the individuals' actions to examine whether the personalized alignment is achieved. Further, we construct a benchmark named PCogAlignBench, which includes 18k instances and 20 individuals with different Role-Sets. Finally, we present a framework called PCogAlign, which constructs a cognition-aware and action-based reward model for personalized alignment. Experimental results and human evaluations demonstrate the reliability of the PCogAlignBench and the effectiveness of our proposed PCogAlign. We will open-source the constructed benchmark and code at https://github.com/NLPGM/PCogAlign.
This paper examines the linguistic relativity principle (Whorf, 1956) by investigating the impact of grammatical gender on cognition in simultaneous bilinguals of three-gendered Ukrainian and Russian. It examines whether speakers of three-gendered languages show grammatical gender effects on categorisation, empirically addressing claims that such effects are insignificant due to the presence of the neuter gender (Sera et al., 2002). We conducted two experiments using a similarity judgement paradigm while manipulating the presence of neuter gender stimuli (Phillips & Boroditsky, 2003). Experiment 1, including neuter gender, revealed no significant effects, compatible with earlier studies on three-gendered languages. Conversely, Experiment 2, excluding neuter gender stimuli, showed significant language effects. Bilingual participants rated pairs as more similar when grammatical genders in both languages were congruent with the biological sex of a character. Significant effects were also found for pairs with mismatching grammatical genders in Ukrainian and Russian. Participants with higher proficiency in Ukrainian rated pairs as more similar when the grammatical gender of a noun in Ukrainian was congruent with the character’s biological sex, and incongruent in Russian. Our findings thus provide the first empirical demonstration that the exclusion of neuter gender online induces grammatical gender effects in speakers of three-gendered languages.
Cosimo Tuena, Daniele Di Lernia, Giuseppe Riva
et al.
Embodied cognition theory proposes that spatial cognition preferences facilitate the simulation of action language. Importantly, spatial cognition relies on either egocentric (body-dependent) or allocentric (body-independent) representations. Research demonstrates that spatial representation proclivity influences the simulation of non-transfer action sentences. However, the impact of individual spatial cognition preferences on transfer action sentence simulation remains unexplored. We administered an egocentric and allocentric memory task and an action sentence recognition task to 37 participants. We used an egocentric–allocentric recall strategy proclivity index to classify participants and employed this metric as a moderator between the transfer perspective (first-person perspective, 1PP vs. third-person perspective, 3PP) and the transfer type (concrete vs. abstract). We found that spatial preferences do not moderate 1PP transfer action sentence recognition. Importantly, we found that egocentric proclivity improves 3PP transfer action sentence recognition and that allocentric proclivity hampers 3PP transfer action sentence recognition. No moderation was found for the transfer type. The study suggests that recognition memory for sentences describing others’ actions is related to body-dependent spatial representations, suggesting a possible link between spatial memory proclivity and action language simulation.
Recent studies of the applications of conversational AI tools, such as chatbots powered by large language models, to complex real-world knowledge work have shown limitations related to reasoning and multi-step problem solving. Specifically, while existing chatbots simulate shallow reasoning and understanding they are prone to errors as problem complexity increases. The failure of these systems to address complex knowledge work is due to the fact that they do not perform any actual cognition. In this position paper, we present Cognitive AI, a higher-level framework for implementing programmatically defined neuro-symbolic cognition above and outside of large language models. Specifically, we propose a dual-layer functional architecture for Cognitive AI that serves as a roadmap for AI systems that can perform complex multi-step knowledge work. We propose that Cognitive AI is a necessary precursor for the evolution of higher forms of AI, such as AGI, and specifically claim that AGI cannot be achieved by probabilistic approaches on their own. We conclude with a discussion of the implications for large language models, adoption cycles in AI, and commercial Cognitive AI development.
In recent years, radar systems have advanced significantly, offering environmental adaptation and multi-task capabilities. These developments pose new challenges for electronic intelligence (Elint) and electronic support measures (ESM), which need to identify and interpret sophisticated radar behaviors. This paper introduces a Deep Multi-Intentional Inverse Reinforcement Learning (DMIIRL) method for the identification and inverse cognition of cognitive multi-function radars (CMFR). Traditional Inverse Reinforcement Learning (IRL) methods primarily target single reward functions, but the complexity of CMFRs necessitates multiple reward functions to fully encapsulate their behavior. To this end, we develop a method that integrates IRL with Expectation-Maximization (EM) to concurrently handle multiple reward functions, offering better trajectory clustering and reward function estimation. Simulation results demonstrate the superiority of the proposed method over baseline approaches.
This paper presents an in-depth examination of the evolution and interplay of cognitive and expressive capabilities in large language models (LLMs), with a specific focus on Baichuan-7B and Baichuan-33B, an advanced bilingual (Chinese and English) LLM series. We define and explore the model's cognitive and expressive capabilities through linear representations across three critical phases: Pretraining, Supervised Fine-Tuning (SFT), and Reinforcement Learning from Human Feedback (RLHF). Cognitive capability is defined as the quantity and quality of information conveyed by the neuron output vectors within the network, similar to the neural signal processing in human cognition. Expressive capability is defined as the model's capability to produce word-level output. Our findings unveil a sequential development pattern, where cognitive abilities are largely established during Pretraining, whereas expressive abilities predominantly advance during SFT and RLHF. Statistical analyses confirm a significant correlation between the two capabilities, suggesting that cognitive capacity may limit expressive potential. The paper also explores the theoretical underpinnings of these divergent developmental trajectories and their connection to the LLMs' architectural design. Moreover, we evaluate various optimization-independent strategies, such as few-shot learning and repeated sampling, which bridge the gap between cognitive and expressive capabilities. This research reveals the potential connection between the hidden space and the output space, contributing valuable insights into the interpretability and controllability of their training processes.
Social media challenge several established concepts of memory research. In particular, the day-to-day mundane discourse of social media blur the essential distinction between commemorative and non-commemorative memory. We address these challenges by presenting a methodological framework that explores the dynamics of social memory on various social media. Our method combines top-down data mining with a bottom-up analysis tailored to each platform. We demonstrate the application of our approach by studying how the Holocaust is remembered in different corpora, including a dataset of 5.3 million Facebook posts and comments collected between 2015 and 2017 and a 5 million Tweets and Retweets dataset collected in 2021. We first identify the mnemonic agents initiating the discussion of the memory of the Holocaust and those responding to it. Second, we compare the macro-rhythms of Holocaust discourse on the two platforms, identifying peaks and mundane discussions that extend beyond commemorative occasions. Third, we identify distinctive language and cultural norms specific to the memorialization of the Holocaust on each platform. We conceptualize these dynamics as ‘Mnemonic Markers’ and discuss them as potential pathways for memory researchers who wish to explore the unique memory dynamics afforded by social media.
Communication. Mass media, Consciousness. Cognition
Monika Lohani, Joel M. Cooper, Amy S. McDonnell
et al.
Abstract The reliability of cognitive demand measures in controlled laboratory settings is well-documented; however, limited research has directly established their stability under real-life and high-stakes conditions, such as operating automated technology on actual highways. Partially automated vehicles have advanced to become an everyday mode of transportation, and research on driving these advanced vehicles requires reliable tools for evaluating the cognitive demand on motorists to sustain optimal engagement in the driving process. This study examined the reliability of five cognitive demand measures, while participants operated partially automated vehicles on real roads across four occasions. Seventy-one participants (aged 18–64 years) drove on actual highways while their heart rate, heart rate variability, electroencephalogram (EEG) alpha power, and behavioral performance on the Detection Response Task were measured simultaneously. Findings revealed that EEG alpha power had excellent test–retest reliability, heart rate and its variability were good, and Detection Response Task reaction time and hit-rate had moderate reliabilities. Thus, the current study addresses concerns regarding the reliability of these measures in assessing cognitive demand in real-world automation research, as acceptable test–retest reliabilities were found across all measures for drivers across occasions. Despite the high reliability of each measure, low intercorrelations among measures were observed, and internal consistency was better when cognitive demand was estimated as a multi-factorial construct. This suggests that they tap into different aspects of cognitive demand while operating automation in real life. The findings highlight that a combination of psychophysiological and behavioral methods can reliably capture multi-faceted cognitive demand in real-world automation research.
Esteban Hernández-Rivera, Alessia Kalogeris, Mehrgol Tiv
et al.
Abstract People are often asked to self-evaluate their abilities, and these evaluations may not always reflect objective reality. Here, we investigated this issue for bilingual adults’ self-evaluations of language proficiency and usage. We specifically examined how people’s self-reported language solidarity impacted their first- (L1) and second-language (L2) self-evaluations, while statistically controlling for their objective language performance (i.e. LexTALE). We also investigated whether this impact varied for value-laden evaluations (e.g. how “good” am I at my L2) vs. usage-based evaluations (e.g. how often do I use my L2) for two sociolinguistically distinct groups (i.e. English-L1 speakers vs. French-L1 speakers in Montreal). Starting with value-laden self-evaluations, we found that French-L1 speakers with more favourable L2-English solidarity tended to underestimate their objective L2 ability, whereas French-L1 speakers with less favourable L2-English solidarity more accurately estimated their objective L2 ability. In contrast, English-L1 speakers with more favourable L2-French solidarity more accurately estimated their objective L2 ability than those with less favourable L2-French solidarity who underestimated their L2-French abilities. Turning to usage-based self-evaluations, we found that participants' self-evaluations were generally more accurate reflections of their performance, in a manner that was less affected by individual differences in self-reported language solidarity. This implies that language solidarity (or perhaps language attitudes more generally) can implicitly or explicitly impact bilingual adults’ language self-evaluations when these evaluations are value-laden. These data suggest that people’s language attitudes can bias how they perceive their abilities, although self-evaluations based on language use may be less susceptible to bias than those that are value-laden. These data have implications for the study of language and cognition that depend on self-assessments of individual differences and are relevant to work on how people self-assess their abilities generally.
Timely implementation of interventions to slow cognitive decline among older adults requires accurate monitoring to detect changes in cognitive function. Data gathered using wearable devices that can continuously monitor factors known to be associated with cognition could be used to train machine learning models and develop wearable-based cognitive monitoring systems. Using data from over 2,400 older adults in the National Health and Nutrition Examination Survey (NHANES) we developed prediction models to differentiate older adults with normal cognition from those with poor cognition based on outcomes from three cognitive tests measuring different domains of cognitive function. During repeated cross-validation, CatBoost, XGBoost, and Random Forest models performed best when predicting cognition based on processing speed, working memory, and attention (median AUCs >0.82) compared to immediate and delayed recall (median AUCs >0.72) and categorical verbal fluency (median AUC >0.68). Activity and sleep parameters were also more strongly associated with processing speed, working memory, and attention compared to other cognitive subdomains. Our work provides proof of concept that wearable-based cognitive monitoring systems may be a viable alternative to traditional methods for monitoring processing speeds, working memory, and attention. We further identified novel metrics that could be targets in future causal studies seeking to better understand how sleep and activity parameters influence cognitive function among older adults.
Effective connectivity estimation plays a crucial role in understanding the interactions and information flow between different brain regions. However, the functional time series used for estimating effective connectivity is derived from certain software, which may lead to large computing errors because of different parameter settings and degrade the ability to model complex causal relationships between brain regions. In this paper, a brain diffuser with hierarchical transformer (BDHT) is proposed to estimate effective connectivity for mild cognitive impairment (MCI) analysis. To our best knowledge, the proposed brain diffuser is the first generative model to apply diffusion models to the application of generating and analyzing multimodal brain networks. Specifically, the BDHT leverages structural connectivity to guide the reverse processes in an efficient way. It makes the denoising process more reliable and guarantees effective connectivity estimation accuracy. To improve denoising quality, the hierarchical denoising transformer is designed to learn multi-scale features in topological space. By stacking the multi-head attention and graph convolutional network, the graph convolutional transformer (GraphConformer) module is devised to enhance structure-function complementarity and improve the ability in noise estimation. Experimental evaluations of the denoising diffusion model demonstrate its effectiveness in estimating effective connectivity. The proposed model achieves superior performance in terms of accuracy and robustness compared to existing approaches. Moreover, the proposed model can identify altered directional connections and provide a comprehensive understanding of parthenogenesis for MCI treatment.
Whenever humans use tools human performance is enhanced. Cognitive systems are a new kind of tool continually increasing in cognitive capability and are now performing high level cognitive tasks previously thought to be explicitly human. Usage of such tools, known as cogs, are expected to result in ever increasing levels of human cognitive augmentation. In a human cog ensemble, a cooperative, peer to peer, and collaborative dialog between a human and a cognitive system, human cognitive capability is augmented as a result of the interaction. The human cog ensemble is therefore able to achieve more than just the human or the cog working alone. This article presents results from two studies designed to measure the effect information supplied by a cog has on cognitive accuracy, the ability to produce the correct result, and cognitive precision, the propensity to produce only the correct result. Both cognitive accuracy and cognitive precision are shown to be increased by information of different types (policies and rules, examples, and suggestions) and with different kinds of problems (inventive problem solving and puzzles). Similar effects shown in other studies are compared.