When LLM-based multi-agent systems disagree, current practice treats this as noise to be resolved through consensus. We propose it can be signal. We focus on hate speech moderation, a domain where judgments depend on cultural context and individual value weightings, producing high legitimate disagreement among human annotators. We hypothesize that convergent disagreement, where agents reason similarly but conclude differently, indicates genuine value pluralism that humans also struggle to resolve. Using the Measuring Hate Speech corpus, we embed reasoning traces from five perspective-differentiated agents and classify disagreement patterns using a four-category taxonomy based on reasoning similarity and conclusion agreement. We find that raw reasoning divergence weakly predicts human annotator conflict, but the structure of agent discord carries additional signal: cases where agents agree on a verdict show markedly lower human disagreement than cases where they do not, with large effect sizes (d>0.8) surviving correction for multiple comparisons. Our taxonomy-based ordering correlates with human disagreement patterns. These preliminary findings motivate a shift from consensus-seeking to uncertainty-surfacing multi-agent design, where disagreement structure - not magnitude - guides when human judgment is needed.
Mahnuma Mahfuz Estee, Yuanyuan Wang, Stephane Heritier
et al.
ABSTRACT Background Back pain poses a significant global burden, within which individuals with more severe symptoms consume higher healthcare expenses than those with lesser back pain. Whether measures of body composition predict high‐intensity back pain and/or high‐disability in population‐based cohorts is unknown. This study aimed to examine the association between body composition at baseline and their change in the prior 5 years (between 2001–2005 and 2006–2010) and incident high‐intensity back pain and/or high‐disability in long‐term follow‐up, 10 years later (2016–2021) in a population‐based cohort of men. Method This study examined men with no or low‐intensity back pain and disability (Graded Chronic Pain Scale) at back pain study baseline (2006–2010) within the Geelong Osteoporosis Study. Those developing high‐intensity pain and/or high disability at follow‐up (2016–2021) were identified. Weight, body mass index (BMI), abdominal circumferences, fat mass and lean mass (dual energy X‐ray absorptiometry) were assessed prebaseline (2001–2005) and at baseline. The association of body composition at baseline and change in body composition from prebaseline to baseline with incident high‐intensity pain and/or high disability at follow‐up were examined using multivariable logistic regression. Result Of 695 participants with no or low‐intensity pain and disability at baseline, 441 (62.3%) completed follow‐up with a mean age of 54.3 ± 14.1 years: 37 (8.3%) developed high‐intensity pain and/or high‐disability, 33 (7.5%) developed high‐intensity pain and 14 (3.2%) high disability. No measures of body composition at baseline were associated with incident high‐intensity pain and/or high disability at follow‐up in the whole population. In subgroup analysis, among men aged over 60 years, but not younger, higher lean mass was associated with decreased likelihood of high‐intensity pain and/or high‐disability (odds ratio [OR] 0.86, 95% confidence interval [CI] 0.76, 0.97, interaction p < 0.001). In the whole population, examination of the relationship between change in measures of body composition between prebaseline and baseline, only a one unit increase in BMI, equivalent to 3.1‐kg weight gain, was associated with increased incident high disability (OR 1.63, 95% CI 1.06, 2.51). Conclusion In a population‐based sample, without severe back pain and disability, in older men aged ≥60 years, higher lean mass was protective of incident high‐intensity pain and/or high disability. An increase in BMI, over 5 years, equivalent to 3.1‐kg weight gain, was associated with incident back pain related high disability 10 years later. These results demonstrate another detrimental consequence of weight gain and highlight the importance of maintaining muscle mass in older men.
Diseases of the musculoskeletal system, Human anatomy
Abstract Alternative splicing (AS) plays a vital role in the pathogenesis of schizophrenia (SCZ). Previous studies have linked the genetic signals from genome-wide association studies (GWAS) with expression quantitative trait loci (eQTL), but the interplay with other genetic regulatory mechanisms, particularly splicing QTL (sQTL), remains unclear. Here, we constructed a comprehensive disease-specific sQTL map to provide genetic variants that could alter gene activity through RNA splicing in SCZ. We analyzed data from 539 SCZ patients, identifying a total of 24,810 significant sQTLs (FDR < 0.05) involving in AS events of 7083 unique genes. By combining this with a large-scale SCZ GWAS, we employed Mendelian randomization (MR) and colocalization analyses to pinpoint 27 significant risk genes with genetic AS regulation that may play a causal role in SCZ. Additional differential splicing analysis of these genes in 539 cases and 754 controls revealed 12 significant genes that may increase SCZ risk due to their AS dysregulation. Notably, five genes (DPYD, LACC1, CCDC122, ANAPC7, and DGKZ) showed consistent splicing regulation effects in both MR analysis and differential splicing analysis. Pathway enrichment analysis of differentially spliced genes revealed potential biologically pathways relevant to SCZ, particularly in synaptic transmission and microtubule movement. Furthermore, single-cell RNA-seq analysis revealed that several genes were preferentially expressed in specific brain cell types, including oligodendrocytes, microglia, and excitatory neurons. Overall, our findings highlight several susceptibility genes that may contribute to SCZ risk by AS regulation. Further characterization of these genes could advance mechanistic understanding and therapeutic discovery for SCZ.
Social identity theory (SIT) and social categorization theory (SCT) are two facets of the social identity approach (SIA) to understanding social phenomena. SIT and SCT are models that describe and explain how people interact with one another socially, connecting the individual to the group through an understanding of underlying psychological mechanisms and intergroup behaviour. SIT, originally developed in the 1970s, and SCT, a later, more general offshoot, have been broadly applied to a range of social phenomena among people. The rise of increasingly social machines embedded in daily life has spurned efforts on understanding whether and how artificial agents can and do participate in SIA activities. As agents like social robots and chatbots powered by sophisticated large language models (LLMs) advance, understanding the real and potential roles of these technologies as social entities is crucial. Here, I provide a primer on SIA and extrapolate, through case studies and imagined examples, how SIT and SCT can apply to artificial social agents. I emphasize that not all human models and sub-theories will apply. I further argue that, given the emerging competence of these machines and our tendency to be taken in by them, we experts may need to don the hat of the uncanny killjoy, for our own good.
Humans inhabit a world defined by interactions -- with other humans, objects, and environments. These interactive movements not only convey our relationships with our surroundings but also demonstrate how we perceive and communicate with the real world. Therefore, replicating these interaction behaviors in digital systems has emerged as an important topic for applications in robotics, virtual reality, and animation. While recent advances in deep generative models and new datasets have accelerated progress in this field, significant challenges remain in modeling the intricate human dynamics and their interactions with entities in the external world. In this survey, we present, for the first time, a comprehensive overview of the literature in human interaction motion generation. We begin by establishing foundational concepts essential for understanding the research background. We then systematically review existing solutions and datasets across three primary interaction tasks -- human-human, human-object, and human-scene interactions -- followed by evaluation metrics. Finally, we discuss open research directions and future opportunities.
Tom Kouwenhoven, Max Peeperkorn, Roy de Kleijn
et al.
Languages are shaped by the inductive biases of their users. Using a classical referential game, we investigate how artificial languages evolve when optimised for inductive biases in humans and large language models (LLMs) via Human-Human, LLM-LLM and Human-LLM experiments. We show that referentially grounded vocabularies emerge that enable reliable communication in all conditions, even when humans \textit{and} LLMs collaborate. Comparisons between conditions reveal that languages optimised for LLMs subtly differ from those optimised for humans. Interestingly, interactions between humans and LLMs alleviate these differences and result in vocabularies more human-like than LLM-like. These findings advance our understanding of the role inductive biases in LLMs play in the dynamic nature of human language and contribute to maintaining alignment in human and machine communication. In particular, our work underscores the need to think of new LLM training methods that include human interaction and shows that using communicative success as a reward signal can be a fruitful, novel direction.
The functional anatomy of motor skill acquisition was investigated in six normal human subjects who learned to perform a pursuit rotor task with their dominant right hand during serial positron emission tomography (PET) imaging of relative cerebral blood flow (relCBF). The effect of motor execution, rather than learning, was identified by a comparison of four motor performance scans with two control scans (eye movements only). Motor execution was associated with activation of a distributed network involving cortical, striatonigral, and cerebellar sites. Second, the effect of early motor learning was examined. Performance improved from 17% to 66% mean time on target across the four PET scans obtained during pursuit rotor performance. Across the same scans, significant longitudinal increases of relCBF were located in the left primary motor cortex, the left supplementary motor area, and the left pulvinar thalamus. The results demonstrate that changes of regional cerebral activity associated with early learning of skilled movements occur in sites that are a subset of a more widely distributed network that is active during motor execution.
Historically, the study of human anatomy has had a very complex relationship with race and racism in the United States. Today, BIPOC students are disproportionately excluded from the health sciences, in part because anatomy courses play the role of “gatekeepers” for the health professions. Anatomy instructors–including biological anthropologists teaching anatomy-may passively support white supremacy in science and medicine by ignoring anatomy’s problematic history and by teaching in outdated, exclusionary ways, rather than using anatomy courses as opportunities to provide insight into structural racism and support the success of students who identify as Black, Indigenous, and/or a Person of Color (BIPOC). The objectives of this work were to 1) uncover how latent racism in anatomy and anatomy education may be contributing to marginalized students’ exclusion from health care careers, and 2) offer recommendations which will promote the success of BIPOC health sciences students and produce antiracist healthcare practitioners of all identities. Historical, anthropological, and critical pedagogical analysis of anatomy education was conducted. Paolo Freire’s Pedagogy of the Oppressed (2018) was used as a theoretical framework for dissecting the ways in which the traditional pedagogy of anatomy may be particularly exclusionary for BIPOC students in the US. Pedagogical recommendations and recent case studies were collected from the academic literature. Anatomy instructors and medical schools are encouraged to develop a new, humanistic way of teaching anatomy, which requires extensive changes to the anatomy curriculum. Five categories of reform are recommended: improving pedagogical training for anatomy instructors, reconsidering course organization and modalities, emphasizing variation, implementing culturally-responsive teaching and improving culture, and including history in the anatomy curriculum.
Abstract Background Undergraduate medical education is currently undergoing a remarkable period of transformation. The exponential growth of medical knowledge, accompanied by societal changes and expectations for the upcoming generation of physicians, is placing immense pressure on academic institutions to reform their curricula, particularly foundational courses such as human anatomy. Consequently, instructors are grappling with the challenge of striking a balance between a new curriculum and maintaining the time-honored benchmarks of medical education. Methods This study proposes a 9,5-week medical gross anatomy course containing collaborative quizzes to improve the efficacy in a condensed pre-clerkship curriculum. It was hypothesized that the implementation of collaborative quizzes facilitates group learning while ultimately helping students to be better prepared for the NBME (National Board of Medical Examiners) unit examinations used as summative assessment within the course. During the four collaborative quizzes, medical and medical masters student groups rotated through twenty dissected donor stations, each containing a short clinical or anatomy-related multiple-choice question formatted in NBME style. Students individually answered the question first and then collaboratively discussed the question in their group. Success was measured by individual student outcomes of four collaborative quizzes, two NBME anatomy-focused unit examinations, and overall course evaluations. Results Overall, all students (n = 203) passed the course and performed well on collaborative quizzes and NBME anatomy-focused unit examinations. Analysis of the quizzes revealed that questions involving tagged structures, blood supply, and nerves resulted in the most answer changes after group discussion. Course evaluations proved the collaborative quizzes to be enjoyable and beneficial to identify lack of knowledge areas. Conclusions To address time constraints within the pre-clerkship curricula effectively, it is crucial for educators to explore smaller adjustments to core learning principles. Collaborative quizzes in the gross anatomy laboratory provided students with opportunities to assess their knowledge in a low-stakes setting while simultaneously benefiting from peer learning. The collection and analysis of quiz grades offered a method of identifying struggling students before their first summative examination, enabling timely academic support. In the future, the research team hopes to continue using this assessment and integrate content from various other pre-clerkship courses.
Jordyn Young, Laala M Jawara, Diep N Nguyen
et al.
Generative Artificial Intelligence (AI) is integrated into everyday technology, including news, education, and social media. AI has further pervaded private conversations as conversational partners, auto-completion, and response suggestions. As social media becomes young people's main method of peer support exchange, we need to understand when and how AI can facilitate and assist in such exchanges in a beneficial, safe, and socially appropriate way. We asked 622 young people to complete an online survey and evaluate blinded human- and AI-generated responses to help-seeking messages. We found that participants preferred the AI-generated response to situations about relationships, self-expression, and physical health. However, when addressing a sensitive topic, like suicidal thoughts, young people preferred the human response. We also discuss the role of training in online peer support exchange and its implications for supporting young people's well-being. Disclaimer: This paper includes sensitive topics, including suicide ideation. Reader discretion is advised.
BACKGROUND: The aim of the present study was to investigate the protective effect and mechanism of liriodendrin (LDN) is a lignan diglucoside in hepatic ischaemia/ /reperfusion (I/R) injury. Materials and methods: The liver I/R was established in male C57BL/6 mice. The effect of LDN is initially investigated on hepatic I/R injury via estimating histopathology of liver. The level of metabolic enzymes alanine aminotransferase (ALT), aspartate aminotransferase (AST) and alkaline phosphatase (ALP) was studied along with apoptosis of mouse hepatocytes via TUNEL and flow cytometry analysis. The effect of LDN was investigated on oxidative stress biomarkers (glutathione [GSH] content, malondialdehyde [MDA] and superoxide dismutase [SOD] activities) and pro-inflammatory cytokines (tumour necrosis factor alpha [TNF-α], interleukin [IL]-1β and IL-6). Western blot study was also conducted to elucidate the effect of LDN on toll-like receptor 4/nuclear factor kappa B (TLR4/NF-kB). RESULTS: Liriodendrin alleviates liver I/R injury, as manifested by decreased plasma ALT, AST and ALP with improvement in liver necrotic area. LDN also reduces apoptosis of mouse hepatocytes with reduction of oxidative stress and generation of pro-inflammatory cytokines. It significantly reduces the expression of TLR4 and NF-kB. CONCLUSIONS: The study demonstrated that LDN reduces liver injury and prevented apoptosis of hepatocytes following I/R injury. In addition, LDN also reduces oxidative stress, inflammation, and TLR4/NF-kB in I/R injured mice.
Human brain atlases are tools to gather, present, use, and discover knowledge about the human brain. The developments in brain atlases parallel the advances in neuroanatomy. The brain atlas evolution has been from hand-drawn cortical maps to print atlases to digital platforms which, thanks to tremendous advancements in acquisition techniques and computing, has enabled progress in neuroanatomy from gross (macro) to meso-, micro-, and nano-neuroanatomy. Advances in neuroanatomy have been feasible because of introducing new modalities, from the initial cadaveric dissections, morphology, light microscopy imaging and neuroelectrophysiology to non-invasive in vivo imaging, connectivity, electron microscopy imaging, genomics, proteomics, transcriptomics, and epigenomics. Presently, large and long-term brain projects along with big data drive the development in micro- and nano-neuroanatomy. The goal of this work is to address the relationship between neuroanatomy and human brain atlases and, particularly, the impact of these atlases on the understanding, presentation, and advancement of neuroanatomy. To better illustrate this relationship, a brief outline on the evolution of the human brain atlas concept, creation of brain atlases, atlas-based applications, and future brain-related developments is also presented. In conclusion, human brain atlases are excellent means to represent, present, disseminate, and support neuroanatomy.
Humans have internal models of robots (like their physical capabilities), the world (like what will happen next), and their tasks (like a preferred goal). However, human internal models are not always perfect: for example, it is easy to underestimate a robot's inertia. Nevertheless, these models change and improve over time as humans gather more experience. Interestingly, robot actions influence what this experience is, and therefore influence how people's internal models change. In this work we take a step towards enabling robots to understand the influence they have, leverage it to better assist people, and help human models more quickly align with reality. Our key idea is to model the human's learning as a nonlinear dynamical system which evolves the human's internal model given new observations. We formulate a novel optimization problem to infer the human's learning dynamics from demonstrations that naturally exhibit human learning. We then formalize how robots can influence human learning by embedding the human's learning dynamics model into the robot planning problem. Although our formulations provide concrete problem statements, they are intractable to solve in full generality. We contribute an approximation that sacrifices the complexity of the human internal models we can represent, but enables robots to learn the nonlinear dynamics of these internal models. We evaluate our inference and planning methods in a suite of simulated environments and an in-person user study, where a 7DOF robotic arm teaches participants to be better teleoperators. While influencing human learning remains an open problem, our results demonstrate that this influence is possible and can be helpful in real human-robot interaction.
Viktor Kewenig, Andrew Lampinen, Samuel A. Nastase
et al.
The potential of multimodal generative artificial intelligence (mAI) to replicate human grounded language understanding, including the pragmatic, context-rich aspects of communication, remains to be clarified. Humans are known to use salient multimodal features, such as visual cues, to facilitate the processing of upcoming words. Correspondingly, multimodal computational models can integrate visual and linguistic data using a visual attention mechanism to assign next-word probabilities. To test whether these processes align, we tasked both human participants (N = 200) as well as several state-of-the-art computational models with evaluating the predictability of forthcoming words after viewing short audio-only or audio-visual clips with speech. During the task, the model's attention weights were recorded and human attention was indexed via eye tracking. Results show that predictability estimates from humans aligned more closely with scores generated from multimodal models vs. their unimodal counterparts. Furthermore, including an attention mechanism doubled alignment with human judgments when visual and linguistic context facilitated predictions. In these cases, the model's attention patches and human eye tracking significantly overlapped. Our results indicate that improved modeling of naturalistic language processing in mAI does not merely depend on training diet but can be driven by multimodality in combination with attention-based architectures. Humans and computational models alike can leverage the predictive constraints of multimodal information by attending to relevant features in the input.