Pub Date : 2025-12-15DOI: 10.1038/s41562-025-02357-5
Haoyang Chen, Bo Liu, Shuyue Wang, Xiaosha Wang, Wenjuan Han, Xiaochun Wang, Yixin Zhu, Yanchao Bi
Comparing information structures in between deep neural networks (DNNs) and the human brain has become a key method for exploring their similarities and differences. Recent research has shown better alignment of vision–language DNN models, such as contrastive language–image pretraining (CLIP), with the activity of the human ventral occipitotemporal cortex (VOTC) than earlier vision models, supporting the idea that language modulates human visual perception. However, interpreting the results from such comparisons is inherently limited owing to the ‘black box’ nature of DNNs. Here we combine model–brain fitness analyses with human brain lesion data to examine how disrupting the communication pathway between the visual and language systems causally affects the ability of vision–language DNNs to explain the activity of the VOTC to address this. Across four diverse datasets, CLIP consistently captured unique variance in VOTC neural representations, relative to both label-supervised (ResNet) and unsupervised (MoCo) models. This advantage tended to be left-lateralized at the group level, aligning with the human language network. Analyses of 33 patients who experienced a stroke revealed that reduced white matter integrity between the VOTC and the language region in the left angular gyrus was correlated with decreased CLIP–brain correspondence and increased MoCo–brain correspondence, indicating a dynamic influence of language processing on the activity of the VOTC. These findings support the integration of language modulation in neurocognitive models of human vision, reinforcing concepts from vision–language DNN models. The sensitivity of model–brain similarity to specific brain lesions demonstrates that leveraging the manipulation of the human brain is a promising framework for evaluating and developing brain-like computer models.
{"title":"Combined evidence from artificial neural networks and human brain-lesion models reveals that language modulates vision in human perception","authors":"Haoyang Chen, Bo Liu, Shuyue Wang, Xiaosha Wang, Wenjuan Han, Xiaochun Wang, Yixin Zhu, Yanchao Bi","doi":"10.1038/s41562-025-02357-5","DOIUrl":"https://doi.org/10.1038/s41562-025-02357-5","url":null,"abstract":"Comparing information structures in between deep neural networks (DNNs) and the human brain has become a key method for exploring their similarities and differences. Recent research has shown better alignment of vision–language DNN models, such as contrastive language–image pretraining (CLIP), with the activity of the human ventral occipitotemporal cortex (VOTC) than earlier vision models, supporting the idea that language modulates human visual perception. However, interpreting the results from such comparisons is inherently limited owing to the ‘black box’ nature of DNNs. Here we combine model–brain fitness analyses with human brain lesion data to examine how disrupting the communication pathway between the visual and language systems causally affects the ability of vision–language DNNs to explain the activity of the VOTC to address this. Across four diverse datasets, CLIP consistently captured unique variance in VOTC neural representations, relative to both label-supervised (ResNet) and unsupervised (MoCo) models. This advantage tended to be left-lateralized at the group level, aligning with the human language network. Analyses of 33 patients who experienced a stroke revealed that reduced white matter integrity between the VOTC and the language region in the left angular gyrus was correlated with decreased CLIP–brain correspondence and increased MoCo–brain correspondence, indicating a dynamic influence of language processing on the activity of the VOTC. These findings support the integration of language modulation in neurocognitive models of human vision, reinforcing concepts from vision–language DNN models. The sensitivity of model–brain similarity to specific brain lesions demonstrates that leveraging the manipulation of the human brain is a promising framework for evaluating and developing brain-like computer models.","PeriodicalId":19074,"journal":{"name":"Nature Human Behaviour","volume":"158 1","pages":""},"PeriodicalIF":29.9,"publicationDate":"2025-12-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145759565","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"心理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-15DOI: 10.1038/s41562-025-02378-0
Shin Ling Wu
{"title":"Implications of Australia's under-16 social media ban.","authors":"Shin Ling Wu","doi":"10.1038/s41562-025-02378-0","DOIUrl":"https://doi.org/10.1038/s41562-025-02378-0","url":null,"abstract":"","PeriodicalId":19074,"journal":{"name":"Nature Human Behaviour","volume":"46 1","pages":""},"PeriodicalIF":29.9,"publicationDate":"2025-12-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145760087","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"心理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-15DOI: 10.1038/s41562-025-02360-w
Thomas Davidson
Multimodal large language models (MLLMs) could enhance the accuracy of automated content moderation by integrating contextual information. This study examines how MLLMs evaluate hate speech through a series of conjoint experiments. Models are provided with a hate speech policy and shown simulated social media posts that systematically vary in slur usage, user demographics and other attributes. The decisions from MLLMs are benchmarked against judgements by human participants (n = 1,854). The results demonstrate that larger, more advanced models can make context-sensitive evaluations that are closely aligned with human judgement. However, pervasive demographic and lexical biases remain, particularly among smaller models. Further analyses show that context sensitivity can be amplified via prompting but not eliminated, and that some models are especially responsive to visual identity cues. These findings highlight the benefits and risks of using MLLMs for content moderation and demonstrate the utility of conjoint experiments for auditing artificial intelligence in complex, context-dependent applications.
{"title":"Multimodal large language models can make context-sensitive hate speech evaluations aligned with human judgement","authors":"Thomas Davidson","doi":"10.1038/s41562-025-02360-w","DOIUrl":"https://doi.org/10.1038/s41562-025-02360-w","url":null,"abstract":"Multimodal large language models (MLLMs) could enhance the accuracy of automated content moderation by integrating contextual information. This study examines how MLLMs evaluate hate speech through a series of conjoint experiments. Models are provided with a hate speech policy and shown simulated social media posts that systematically vary in slur usage, user demographics and other attributes. The decisions from MLLMs are benchmarked against judgements by human participants (n = 1,854). The results demonstrate that larger, more advanced models can make context-sensitive evaluations that are closely aligned with human judgement. However, pervasive demographic and lexical biases remain, particularly among smaller models. Further analyses show that context sensitivity can be amplified via prompting but not eliminated, and that some models are especially responsive to visual identity cues. These findings highlight the benefits and risks of using MLLMs for content moderation and demonstrate the utility of conjoint experiments for auditing artificial intelligence in complex, context-dependent applications.","PeriodicalId":19074,"journal":{"name":"Nature Human Behaviour","volume":"1 1","pages":""},"PeriodicalIF":29.9,"publicationDate":"2025-12-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145759566","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"心理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-08DOI: 10.1038/s41562-025-02377-1
Cason D Schmit,Gogoal Falia,Philip Sanusi
{"title":"Threats to democracy are threats to health.","authors":"Cason D Schmit,Gogoal Falia,Philip Sanusi","doi":"10.1038/s41562-025-02377-1","DOIUrl":"https://doi.org/10.1038/s41562-025-02377-1","url":null,"abstract":"","PeriodicalId":19074,"journal":{"name":"Nature Human Behaviour","volume":"30 1","pages":""},"PeriodicalIF":29.9,"publicationDate":"2025-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145704513","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"心理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-08DOI: 10.1038/s41562-025-02352-w
Nurit Nobel,Michael Hiscox
Climate-related disasters such as wildfires and floods pose escalating risks to communities worldwide, yet motivating individuals to adopt protective measures remains a persistent challenge. In a pre-registered field experiment with 12,985 Australian homeowners in wildfire-prone areas, we demonstrate that a simple behavioural intervention-integrating proximal cues, such as participants' suburbs, into climate risk communications-significantly increases engagement. Participants who received localized messages were twice as likely to seek further information about wildfire preparedness compared with those who received generic communications (odds ratio of 2.03, 95% confidence interval of 1.33 to 3.16). This effect highlights the power of behavioural interventions in addressing barriers to climate adaptation, particularly by reducing psychological distance and fostering place attachment. By making abstract climate risks tangible and personally relevant, the intervention nudges individuals towards action. These findings suggest a scalable, low-cost approach for enhancing disaster preparedness, offering insights for leveraging behavioural science to mitigate the impact of climate-related disasters.
{"title":"Enhancing climate resilience with proximal cues in personalized climate disaster preparedness messaging.","authors":"Nurit Nobel,Michael Hiscox","doi":"10.1038/s41562-025-02352-w","DOIUrl":"https://doi.org/10.1038/s41562-025-02352-w","url":null,"abstract":"Climate-related disasters such as wildfires and floods pose escalating risks to communities worldwide, yet motivating individuals to adopt protective measures remains a persistent challenge. In a pre-registered field experiment with 12,985 Australian homeowners in wildfire-prone areas, we demonstrate that a simple behavioural intervention-integrating proximal cues, such as participants' suburbs, into climate risk communications-significantly increases engagement. Participants who received localized messages were twice as likely to seek further information about wildfire preparedness compared with those who received generic communications (odds ratio of 2.03, 95% confidence interval of 1.33 to 3.16). This effect highlights the power of behavioural interventions in addressing barriers to climate adaptation, particularly by reducing psychological distance and fostering place attachment. By making abstract climate risks tangible and personally relevant, the intervention nudges individuals towards action. These findings suggest a scalable, low-cost approach for enhancing disaster preparedness, offering insights for leveraging behavioural science to mitigate the impact of climate-related disasters.","PeriodicalId":19074,"journal":{"name":"Nature Human Behaviour","volume":"3 1","pages":""},"PeriodicalIF":29.9,"publicationDate":"2025-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145704512","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"心理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-08DOI: 10.1038/s41562-025-02358-4
James N Druckman,Katherine Ognyanova,Alauna Safarpour,Jonathan Schulman,Kristin Lunz Trujillo,Ata Aydin Uslu,Jon Green,Matthew A Baum,Alexi Quintana-Mathé,Hong Qu,Roy H Perlis,David M J Lazer
Scientists provide important information to the public. Whether that information influences decision-making depends on trust. In the USA, gaps in trust in scientists have been stable for 50 years: women, Black people, rural residents, religious people, less educated people and people with lower economic status express less trust than their counterparts (who are more represented among scientists). Here we probe the factors that influence trust. We find that members of the less trusting groups exhibit greater trust in scientists who share their characteristics (for example, women trust women scientists more than men scientists). They view such scientists as having more benevolence and, in most cases, more integrity. In contrast, those from high-trusting groups appear mostly indifferent about scientists' characteristics. Our results highlight how increasing the presence of underrepresented groups among scientists can increase trust. This means expanding representation across several divides-not just gender and race/ethnicity but also rurality and economic status.
{"title":"Representation in science and trust in scientists in the USA.","authors":"James N Druckman,Katherine Ognyanova,Alauna Safarpour,Jonathan Schulman,Kristin Lunz Trujillo,Ata Aydin Uslu,Jon Green,Matthew A Baum,Alexi Quintana-Mathé,Hong Qu,Roy H Perlis,David M J Lazer","doi":"10.1038/s41562-025-02358-4","DOIUrl":"https://doi.org/10.1038/s41562-025-02358-4","url":null,"abstract":"Scientists provide important information to the public. Whether that information influences decision-making depends on trust. In the USA, gaps in trust in scientists have been stable for 50 years: women, Black people, rural residents, religious people, less educated people and people with lower economic status express less trust than their counterparts (who are more represented among scientists). Here we probe the factors that influence trust. We find that members of the less trusting groups exhibit greater trust in scientists who share their characteristics (for example, women trust women scientists more than men scientists). They view such scientists as having more benevolence and, in most cases, more integrity. In contrast, those from high-trusting groups appear mostly indifferent about scientists' characteristics. Our results highlight how increasing the presence of underrepresented groups among scientists can increase trust. This means expanding representation across several divides-not just gender and race/ethnicity but also rurality and economic status.","PeriodicalId":19074,"journal":{"name":"Nature Human Behaviour","volume":"29 1","pages":""},"PeriodicalIF":29.9,"publicationDate":"2025-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145704514","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"心理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-05DOI: 10.1038/s41562-025-02345-9
David Rose, Siying Zhang, Shaun Nichols, Ellen M. Markman, Tobias Gerstenberg
{"title":"How children map causal verbs to different causes across development","authors":"David Rose, Siying Zhang, Shaun Nichols, Ellen M. Markman, Tobias Gerstenberg","doi":"10.1038/s41562-025-02345-9","DOIUrl":"https://doi.org/10.1038/s41562-025-02345-9","url":null,"abstract":"","PeriodicalId":19074,"journal":{"name":"Nature Human Behaviour","volume":"130 1","pages":""},"PeriodicalIF":29.9,"publicationDate":"2025-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145680108","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"心理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-05DOI: 10.1038/s41562-025-02376-2
Laurel Raffington
{"title":"Academia is just a job","authors":"Laurel Raffington","doi":"10.1038/s41562-025-02376-2","DOIUrl":"https://doi.org/10.1038/s41562-025-02376-2","url":null,"abstract":"","PeriodicalId":19074,"journal":{"name":"Nature Human Behaviour","volume":"25 1","pages":""},"PeriodicalIF":29.9,"publicationDate":"2025-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145680223","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"心理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-05DOI: 10.1038/s41562-025-02375-3
Mindy Nunez Duffourc, Falk Gerrik Verhees, Stephen Gilbert
{"title":"Artificial intelligence characters are dangerous without legal guardrails","authors":"Mindy Nunez Duffourc, Falk Gerrik Verhees, Stephen Gilbert","doi":"10.1038/s41562-025-02375-3","DOIUrl":"https://doi.org/10.1038/s41562-025-02375-3","url":null,"abstract":"","PeriodicalId":19074,"journal":{"name":"Nature Human Behaviour","volume":"88 8 1","pages":""},"PeriodicalIF":29.9,"publicationDate":"2025-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145680222","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"心理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：