PNAS nexus最新文献

The emergence of large language models (LLMs) has sparked considerable interest in their potential application in psychological research, mainly as a model of the human psyche or as a general text-analysis tool. However, the trend of using LLMs without sufficient attention to their limitations and risks, which we rhetorically refer to as "GPTology", can be detrimental given the easy access to models such as ChatGPT. Beyond existing general guidelines, we investigate the current limitations, ethical implications, and potential of LLMs specifically for psychological research, and show their concrete impact in various empirical studies. Our results highlight the importance of recognizing global psychological diversity, cautioning against treating LLMs (especially in zero-shot settings) as universal solutions for text analysis, and developing transparent, open methods to address LLMs' opaque nature for reliable, reproducible, and robust inference from AI-generated data. Acknowledging LLMs' utility for task automation, such as text annotation, or to expand our understanding of human psychology, we argue for diversifying human samples and expanding psychology's methodological toolbox to promote an inclusive, generalizable science, countering homogenization, and over-reliance on LLMs.

reasoning is a key ability for an intelligent system. Large language models (LMs) achieve above-chance performance on abstract reasoning tasks but exhibit many imperfections. However, human abstract reasoning is also imperfect. Human reasoning is affected by our real-world knowledge and beliefs, and shows notable "content effects"; humans reason more reliably when the semantic content of a problem supports the correct logical inferences. These content-entangled reasoning patterns are central to debates about the fundamental nature of human intelligence. Here, we investigate whether language models-whose prior expectations capture some aspects of human knowledge-similarly mix content into their answers to logic problems. We explored this question across three logical reasoning tasks: natural language inference, judging the logical validity of syllogisms, and the Wason selection task. We evaluate state of the art LMs, as well as humans, and find that the LMs reflect many of the same qualitative human patterns on these tasks-like humans, models answer more accurately when the semantic content of a task supports the logical inferences. These parallels are reflected in accuracy patterns, and in some lower-level features like the relationship between LM confidence over possible answers and human response times. However, in some cases the humans and models behave differently-particularly on the Wason task, where humans perform much worse than large models, and exhibit a distinct error pattern. Our findings have implications for understanding possible contributors to these human cognitive effects, as well as the factors that influence language model performance.

Efforts to bridge political divides often focus on navigating complex and divisive issues, but eight studies reveal that we should also focus on a more basic misperception: that political opponents are willing to accept basic moral wrongs. In the United States, Democrats, and Republicans overestimate the number of political outgroup members who approve of blatant immorality (e.g. child pornography, embezzlement). This "basic morality bias" is tied to political dehumanization and is revealed by multiple methods, including natural language analyses from a large social media corpus and a survey with a representative sample of Americans. Importantly, the basic morality bias can be corrected with a brief, scalable intervention. Providing information that just one political opponent condemns blatant wrongs increases willingness to work with political opponents and substantially decreases political dehumanization.

The Industrial Revolution precipitated a pivotal shift from waterpower to coal-fueled steam power in British textile mills. Although it is now widely accepted that steam was chosen to power factories despite the availability of sufficient waterpower resources across most of Britain, the location and suitability of that waterpower during the early 19th century remain underexplored. Here, we employ quantitative fluvial geomorphology alongside historical climate data, factory records, and a catalog of over 26,000 mill sites to reveal that waterpower was abundant for most of early 19th century Britain, except in the central hub of British cotton production: Greater Manchester in the Mersey Basin. Our findings show that surging factory mechanization and overcrowding on key waterways in the Mersey Basin compounded waterpower scarcity arising from a drier 19th century climate. Widespread adoption of coal-fueled steam engines in certain key industrial centers of Britain was a strategy aimed at ameliorating some of the reduced reliability of waterpower. The fact that steam engines were frequently used in water-powered factories in many industrial regions until the third quarter of the 19th century to recirculate water to provide that power, or as a power supplement when waterpower availability was restricted, adds further weight to our argument. Rapid adoption of coal-powered steam engines reshaped the social and structural landscape of industrial work, firmly established Britain's prominence as an industrial powerhouse, and had lasting global industrial and environmental impacts.

This study explores the efficacy of our novel and personalized brain-computer interface (BCI) therapy, in enhancing hand movement recovery among stroke survivors. Stroke often results in impaired motor function, posing significant challenges in daily activities and leading to considerable societal and economic burdens. Traditional physical and occupational therapies have shown limitations in facilitating satisfactory recovery for many patients. In response, our study investigates the potential of motor imagery-based BCIs (MI-BCIs) as an alternative intervention. In this study, MI-BCIs translate imagined hand movements into actions using a combination of scalp-recorded electrical brain activity and signal processing algorithms. Our prior research on MI-BCIs, which emphasizes the benefits of proprioceptive feedback over traditional visual feedback and the importance of customizing the delay between brain activation and passive hand movement, led to the development of RehabSwift therapy. In this study, we recruited 12 chronic-stage stroke survivors to assess the effectiveness of our solution. The primary outcome measure was the Fugl-Meyer upper extremity (FMA-UE) assessment, complemented by secondary measures including the action research arm test, reaction time, unilateral neglect, spasticity, grip and pinch strength, goal attainment scale, and FMA-UE sensation. Our findings indicate a remarkable improvement in hand movement and a clinically significant reduction in poststroke arm and hand impairment following 18 sessions of neurofeedback training. The effects persisted for at least 4 weeks posttreatment. These results underscore the potential of MI-BCIs, particularly our solution, as a prospective tool in stroke rehabilitation, offering a personalized and adaptable approach to neurofeedback training.

In this paper, we examine how different governance types impact prosocial behaviors in a heterogenous society. We construct a general theoretical framework to examine a game-theoretic model to assess the ease of achieving a cooperative outcome. We then build a dynamic agent-based model to examine three distinct governance types in a heterogenous population: monitoring one's neighbors, despotic leadership, and influencing one's neighbors to adapt strategies that lead to better fitness. In our research, we find that while despotic leadership may lead towards high prosociality and high returns it does not exceed the effects of a local individual who can exert positive influence in the community. This may suggest that greater individual gains can be had by cooperating and that global hierarchical leadership may not be essential as long as influential individuals exert their influence for public good and not for public ill.

Mass behavior is the rapid adoption of similar conduct by all group members, with potentially catastrophic outcomes such as mass panic. Yet, these negative consequences are rare in integrated social systems such as social insect colonies, thanks to mechanisms of social regulation. Here, we test the hypothesis that behavioral deactivation between active individuals is a powerful social regulator that reduces energetic spending in groups. Borrowing from scaling theories for human settlements and using behavioral data on harvester ants, we derive ties between the hypermetric scaling of the interaction network and the hypometric scaling of activity levels, both relative to the colony size. We use elements of economics theory and metabolic measurements collected with the behavioral data to link activity and metabolic scalings with group size. Our results support the idea that metabolic scaling across social systems is the product of different balances between their social regulation mechanisms.

Fireflies were believed to originally evolve their novel bioluminescence as warning signals to advertise their toxicity to predators, which was later adopted in adult mating. Although the evolution of bioluminescence has been investigated extensively, the warning signal hypothesis of its origin has not been tested. In this study, we test this hypothesis by systematically determining the presence or absence of firefly toxin lucibufagins (LBGs) across firefly species and inferring the time of origin of LBGs. We confirm the presence of LBGs in the subfamily Lampyrinae, but more importantly, we reveal the absence of LBGs in other lineages, including the subfamilies of Luciolinae, Ototretinae, and Psilocladinae, two incertae sedis lineages, and the Rhagophthalmidae family. Ancestral state reconstructions for LBGs based on firefly phylogeny constructed using genomic data suggest that the presence of LBGs in the common ancestor of the Lampyrinae subfamily is highly supported but unsupported in more ancient nodes, including firefly common ancestors. Our results suggest that firefly LBGs probably evolved much later than the evolution of bioluminescence. We thus conclude that firefly bioluminescence did not originally evolve as direct warning signals for toxic LBGs and advise that future studies should focus on other hypotheses. Moreover, LBG toxins are known to directly target and inhibit the α subunit of Na⁺, K⁺-ATPase (ATPα). We further examine the effects of amino acid substitutions in firefly ATPα on its interactions with LBGs. We find that ATPα in LBG-containing fireflies is relatively insensitive to LBGs, which suggests that target-site insensitivity contributes to LBG-containing fireflies' ability to deal with their own toxins.

When people make choices, the items they consider are often embedded in a context (of other items). How this context affects the valuation of the specific item is an important question. High-value context might make items appear less attractive because of contrast-the tendency to normalize perception of an object relative to its background-or more attractive because of assimilation-the tendency to group objects together. Alternatively, a high-value context might increase prior expectations about the item's value. Here, we investigated these possibilities. We examined how unavailable context items affect choices between two target items, as well as the willingness-to-pay for single targets. Participants viewed sets of three items for several seconds before the target(s) were highlighted. In both tasks, we found a significant assimilation-like effect where participants were more likely to choose or place a higher value on a target when it was surrounded by higher-value context. However, these context effects were only significant for participants' fastest choices. Using variants of a drift-diffusion model, we established that the unavailable context shifted participants' prior expectations towards the average values of the sets but had an inconclusive effect on their evaluations of the targets during the decision (i.e. drift rates). In summary, we find that people use context to inform their initial valuations. This can improve efficiency by allowing people to get a head start on their decision. However, it also means that the valuation of an item can change depending on the context.