Browse Articles

Comparing descriptive and theoretical models of decision-making under uncertainty and their relation to socioeconomic factors

This study examines the selection and validation of measurement models for decision-making under uncertainty, with particular emphasis on the integration of socioeconomic contexts in these models. We critically compared four distinct models, differing in their mathematical form of risk and ambiguity, to determine which best predicts willingness-to-pay behavior in a financial decision-making task. We used Bayesian hierarchical modeling to fit each measurement model and the Leave-One-Out Information Criteria to assess how each model performed. In a sample of 74 community members, we found that the maximal descriptive model — one that accounts for the variance across participants’ sensitivity to changes under risk and ambiguity — demonstrated superior predictive accuracy in out-of-sample testing. Notably, our results revealed that adding socioeconomic factors into the model improved prediction. Further, people from higher-income households exhibited a greater aversion to ambiguity, whereas those from lower-income households showed less aversion to ambiguity. There was a lack of association between annual household income and risk. These findings not only highlight the importance of rigorously validating measurement models but also underscore the pivotal role of socioeconomic factors in shaping decision-making under uncertainty. Future research should further investigate how aspects of socioeconomic background—such as childhood economic conditions and environmental unpredictability—influence decision-making. It also is important to test whether ambiguity aversion estimates, when contextualized by socioeconomic variables, can reliably predict real-world decision-making under uncertainty.

KYLO-0603, a novel liver-targeting, thyroid hormone receptor-β agonist for the inhibition of MASH progression

Metabolic dysfunction–associated steatohepatitis (MASH) is a progressive liver disease associated with liver-related complications and death. Kylo-0603 is a novel agonist for the thyroid hormone receptor β (THR-β) that has been developed by merging the structures of three acetylgalactosamine (GalNAc)-modified ASPGR ligands with a triiodothyronine (T3) analog. This unique design enables both THR-β activation and targeted delivery to hepatocytes, which significantly reduces the risk of adverse effects related to increased systemic thyroid hormone activity. Additionally, it effectively lowers serum cholesterol levels by as much as 69.2% and low-density lipoprotein cholesterol (LDL-C) levels by up to 88.2% in the MASH mouse model. Meanwhile, Kylo-0603 demonstrated a marked improvement in histological parameters, decreasing steatosis by up to 1.3 points (P < 0.001), inflammation by 1.8 points (P < 0.0001), and ballooning by 0.8 points (P < 0.01). The non-alcoholic steatohepatitis (NASH) activity score (NAS) demonstrated a significant reduction of up to 3.7 points (P < 0.0001), while the fibrosis score decreased by 0.6 points (P < 0.05). These findings indicate that Kylo-0603 effectively ameliorates hepatic MASH pathology and attenuates fibrosis progression. In summary, Kylo-0603-a highly tissue- and target-selective, low-toxicity THR-β agonist—exhibits substantial therapeutic potential for managing MASH and represents a promising novel treatment option for affected patients.

Study on the skid resistance performance of wet and slippery asphalt pavement on curved roads in aeolian sand areas

In order to study the change law of the anti sliding performance of desert highway in the wet and slippery state of the curve, two different curves and straight roads of the same highway were selected for the field test through the method of field test. The British pencil number (BPN) and mean texture depth (MTD) were selected as the evaluation indexes, and the change law of BPN and MTD at different curves and straight roads with the amount of sand and water was measured. In order to clarify the sensitivity of the anti sliding performance under different humidity, the water volume (humidity) that has the greatest impact on BPN and MTD was obtained by using grey correlation analysis, and then the anti sliding performance at the curves and straight roads was evaluated.The performance is compared and analyzed. The results show that the anti sliding performance of the curve and the straight road decreased by 33.3% and 32.0%, respectively, and the change law of the two is basically the same. The water aggravates the decline of the anti sliding performance of the road surface, and the anti sliding performance of the curve at low speed is about 3.2% less than that of the straight road.

Water injustice in Colombia: Perceptions and realities of water quality examined through advanced machine learning

Access to safe drinking water is essential for public health. In Colombia, Resolution 2115 of 2007 mandates the use of the Water Quality Risk Index for Human Consumption (IRCA) to classify risk levels. However, a significant gap exists between objective IRCA measurements and public perception, which unfolds in the context of water injustice. This study examines the factors influencing perceptions of water quality in regions with high levels of water contamination. Based on a sample of 37,028 household heads from the 2022 Quality of Life Survey by DANE, the study applied advanced machine learning techniques, including logistic regressions with Lasso regularization and double machine learning, combining random forests and logit models. The analysis included sociodemographic factors, environmental awareness, and the environmental conditions experienced by respondents. Findings confirm the presence of water injustice in Colombia, highlighting a significant disconnect between IRCA scores and public perceptions. Additionally, perceptions of water quality are strongly influenced by visible environmental problems, such as air pollution, bad odors, and litter, suggesting that people tend to focus on more evident issues while overlooking water contamination. The results of this study highlight the need to implement policies that inform Colombians about the quality of the water they consume, in order to promote greater environmental commitment.

Enhancing epidemic forecasting with a physics-informed spatial identity neural network

Forecasting the future number of confirmed cases in each region is a critical challenge in controlling the spread of infectious diseases. Accurate predictions enable the proactive development of optimal containment strategies. Recently, deep learning-based models have increasingly leveraged graph structures to capture the spatial dynamics of epidemic spread. While intuitive, this approach often increases model complexity, and the resulting performance gains may not justify the added burden. In some cases, it may even lead to overfitting. Moreover, infectious disease data is typically noisy, making it difficult to extract infectious disease-specific dynamics from data without guidance based on epidemiological domain knowledge. To address these issues, we propose a simple yet effective hybrid model for multi-region epidemic forecasting, termed Physics-Informed Spatial IDentity neural network (PISID). This model integrates a spatio-temporal identity (STID)-based neural network module, which encodes spatio-temporal information without relying on graph structures, with an SIR module grounded in classical epidemiological dynamics. Regional characteristics are incorporated via a spatial embedding matrix, and epidemiological parameters are inferred through a fully connected neural network. These parameters are then used to govern the dynamics of the SIR model for forecasting purposes. Experiments on real-world datasets demonstrate that the proposed PISID model achieves stable and superior predictive performance compared to baseline models, with approximately 27K parameters and an average training time of 0.45 seconds per epoch. Additionally, ablation studies validate the effectiveness of the neural network’s encoding architecture, and analysis of the decoded epidemiological parameters highlights the model’s interpretability. Overall, PISID contributes to reliable epidemic forecasting by integrating data-driven learning with epidemiological domain knowledge.

Insect-habitat-plant interaction networks provide guidelines to mitigate the risk of transmission of Xylella fastidiosa to grapevine in Southern France

Xylella fastidiosa ( Xf ) is a xylem-limited bacterium that has been recorded in several European countries since its detection in 2013 in Apulia (Italy). Given the prominence of the wine industry in many southern European countries, a big threat is the development of Pierce’s disease in grapevines. Yet, the insect-habitat and insect-plant interaction networks in which xylem feeders, possible vectors of Xf , are involved around European vineyards are largely unknown. Here we describe these networks in three key wine-growing regions of southern France (Provence-Alpes-Côte d’Azur, Occitanie, Nouvelle-Aquitaine) to identify primary xylem feeder habitats, and assess their specialization degree at the habitat, plant family, and plant species levels. A total of 92 landscapes (and 700 sites) were studied over three sampling sessions in the fall 2020, spring 2021, and fall 2021. Among the habitats sampled, meadows hosted the largest xylem feeder communities, followed by alfalfa fields. Vineyard headlands and inter-rows hosted slightly smaller xylem feeder communities, indicating that potential Xf vectors thrive near crops. Grapevines hosted few xylem feeders, suggesting rare but possible transfer to vulnerable crops. Philaenus spumarius , Aphrophora alni , Lepyronia coleoptrata , and Cicadella viridis were all similarly generalists at the habitat, plant family or plant species level. The only specialist was Aphrophora grp. salicina , which was restricted to riparian forests and to Salicaceae. Neophilaenus spp. were extremely specialist at the plant family level (Poaceae), but rather generalist at the habitat and plant species levels. All 1017 insects screened for the presence of Xf tested negative, showing that Xf is not widespread in the studied regions. Our study provides new basic ecological information on potential vectors of Xf , especially on their specialization and feeding preferences, as well as practical information that may be relevant for the design of epidemiological surveillance plans.

The comprehensive researcher development framework (CRDF): Core learning outcomes for research training

Becoming a researcher involves the iterative development of deep disciplinary knowledge, specific technical skills, and psychosocial attitudes, behaviors, and beliefs. Consequently, training researchers is resource- and time-intensive. In addition, expectations can be opaque because the traditional apprenticeship model used in research training is idiosyncratic, defined by norms and traditions that vary across disciplines. To align and make research training expectations more transparent, we developed the Comprehensive Researcher Development Framework (CRDF) by extracting and analyzing learning outcomes from 56 previously published evidence-based frameworks from across disciplines. The individual frameworks each addressed a limited range of training stages (e.g., undergraduate only), focused on a subset of learning outcomes (e.g., technical skills), and/or included a single or narrow subset of disciplines (e.g., biomedical sciences). The CRDF derived from these frameworks includes 79 core learning outcomes nested under 8 areas of researcher development that are supported by evidence of content validity collected from experts in the research community. The CRDF builds consensus across disciplines and addresses undergraduate through postdoctoral career stages to define a coherent continuum of research learning outcomes that can be used to monitor and study researcher development. The CRDF does not replace existing discipline-based or training stage specific frameworks but rather can link and coordinate their use. The CRDF can be used by research training program directors to design new or refine existing research training programs, track individual research mentee development over time, and demystify the research training process for mentors and mentees. The CRDF can also be used by scholars studying researcher development to link data on core learning outcomes across research training programs, stages, and disciplines.

ACE1 does not influence cerebral Aβ degradation or amyloid plaque accumulation in 5XFAD mice

Alzheimer’s disease is the most common form of dementia, and multiple lines of evidence support the relevance of Aβ deposition and amyloid plaque accumulation in the neurotoxicity and cognitive decline in AD. Rare mutations in angiotensin-converting-enzyme-1 have been highly associated with late onset AD patients; however, the mechanism for ACE1 mutation in AD pathogenesis is unknown. While numerous studies have shown that ACE1 indeed catabolizes Aβ, majority of these studies were performed in vitro, and conflicting results have been reported in clinical and in vivo systems. Therefore, we further investigated this in vivo by generating and examining a novel mouse model. Specifically, we analyzed 6-month-old 5XFAD mice with ACE1 knockdown restricted to excitatory neurons, achieved by driving Cre recombinase expression under the CamKIIα promoter. These mice were generated by crossing 5XFAD mice to ACE1 conditional knockout mice expressing Cre specifically in excitatory neurons. Our analyses revealed that neuronal ACE1 knockdown does not significantly affect amyloid plaque load and neuroinflammation in the hippocampus and cortex of 5XFAD mice at 6-months of age.

Enhancing security in instant messaging systems with a hybrid SM2, SM3, and SM4 encryption framework

With the rapid integration of instant messaging systems (IMS) into critical domains such as finance, public services, and enterprise operations, ensuring the confidentiality, integrity, and availability of communication data has become a pressing concern. Existing IMS security solutions commonly employ traditional public-key cryptography, centralized authentication servers, or single-layer encryption, each of which is susceptible to single-point failures and provides only limited resistance against sophisticated attacks. This study addresses the research gap regarding the complementary advantages of SM2, SM3, and SM4 algorithms, as well as hybrid collaborative security schemes in IMS security. This paper presents a hybrid encryption security framework that combines the SM2, SM3, and SM4 algorithms to address emerging threats in IMS. The proposed framework adopts a decentralized architecture with certificateless authentication and performs all encryption and decryption operations on the client side, eliminating reliance on centralized servers and mitigating single-point failure risks. It further enforces an encrypt-before-store policy to enhance data security at the storage layer. The framework integrates SM2 for key exchange and authentication, SM4 for message encryption, and SM3 for integrity verification, forming a multi-layer defense mechanism capable of countering Man-in-the-Middle (MITM) attacks, credential theft, database intrusions, and other vulnerabilities. Experimental evaluations demonstrate the system’s strong security performance and communication efficiency: SM2 achieves up to 642 times faster key generation and 2.2 times faster decryption compared to RSA-3072; SM3 improves hashing performance by up to 11.5% over SHA-256; and SM4 delivers up to 22% higher encryption efficiency than AES-256 for small data blocks. These results verify the proposed framework’s practicality and performance advantages in lightweight, real-time IMS applications.

Controlled bidirectional teleportation of unknown single-particle states by using an arbitrary high-dimensional entangled state

This study investigates controlled bidirectional transmission of two low-dimensional unknown single-particle quantum states through high-dimensional entangled channels. Initially, complete orthogonal non-symmetric bases in (d×f)-dimensional Hilbert space and five-particle maximally entangled states in d-dimensional Hilbert space are constructed, with (3×2)-dimensional non-symmetric bases and 3-dimensional 5-qutrit entangled states serving as specific instances. We then propose a controlled bidirectional teleportation (CBT) protocol wherein a 3-dimensional 5-qutrit maximally entangled channel enables simultaneous exchange of two 2-dimensional single-qubit states under supervisor authorization, implemented via non-symmetric basis measurements. Subsequent replacement of the maximally entangled channel with a 3-dimensional 5-qutrit non-maximally entangled state facilitates probabilistic reconstruction of the original states through auxiliary particle introduction and appropriate unitary operations, still under supervisory control. The success probabilities are analytically derived, demonstrating that the non-maximally entangled protocol generalizes the former scheme. Furthermore, these protocols admit dual extensions: (i) quantum channel dimensionality scalable from 3-dimensional to arbitrary d dimensions; (ii) transmitted state dimensionality extendable from 2-dimensional to arbitrary f-dimensional Hilbert space (f < d).

The effectiveness of a new parent education intervention program for primary school-age children and parents in Hong Kong

In Hong Kong, there is a recognised service gap and societal need for effective and evidence-based parenting education within mental health interventions. This study indtrocues the Parent Education in Mental Health (PEMH) intervention based on a theoretical framework. From 2021 to 2022, 2,246 parents from six elementary schools in Hong Kong participated in the PEMH program. An initial evaluation was conducted to asesses the effect of PEMH, with assessments at baseline and post-program. Parental stress was assessed by the Parental Stress Scale (PSS-18), parental self-efficacy was measured by the General Self-Efficacy Scale (GSE-10), the parenting style was evaluated by the Parenting Styles and Dimensions Questionnaire (PSDQ), and children’s mental health outcomes were measured using the Strengths and Difficulties Questionnaires (SDQ). Results showed that participation in the PEMH program led to significant improvements in parental stress, general self-efficacy, children’s mental health outcomes and shifts towards an authoritative parenting style. Focus groups conducted as part of the study highlighted the importance of social support in managing childcare stress and identified key components of the program for future enhancements. Parental stress was significantly associated with self-efficacy, parenting styles, and perception of children’s difficulties, while socioeconomic factors including higher educational level and income were associated with better parental and child outcomes. Our findings underscore the effect of PEMH in strengthening parental outcomes and child well-being, suggesting that a broader imputation of PEMH could be beneficial, particularly for parents of preschoolers.

Comparison of angle-closure detection between automated gonioscopy and anterior-segment optical coherence tomography

Purpose To investigate the concordance between angle-closure assessments based on GS-1 gonioscope images and those obtained with anterior-segment optical coherence tomography. Study design Retrospective clinical study. Methods We included 33 patients (53 eyes) who visited Chukyo Eye Clinic during 2020–2024, were suspected of having angle closure, and underwent anterior-segment optical coherence tomography (CASIA2 Advance STAR Analyzer) and GS-1 examinations. The 16-directional images captured with the GS-1 were divided into two halves, creating 32 directions, which were rearranged to correspond with those obtained via anterior-segment optical coherence tomography. Agreement between evaluations was analyzed using Cohen’s κ, and the area under the receiver operating characteristic curve was evaluated. Anterior-segment optical coherence tomography images were manually corrected, and eyes with areas classified as “narrow” or “closed” were categorized as angle closure. With the GS-1, two glaucoma specialists independently reviewed the images. Areas in which the posterior trabecular meshwork was obscured in more than half of the image (Scheie classification grades III–IV) were judged indicative of angle closure. Results We included 1,660 directions from 53 eyes in the agreement analysis. The proportion of directions judged as angle closure was 57.0% with anterior-segment optical coherence tomography and 46.1% with the GS-1. Cohen’s κ for inter-test agreement was 0.173 (95% confidence interval: 0.128–0.218), and the area under the receiver operating characteristic curve was 0.588 (95% confidence interval: 0.561–0.615). Conclusion Analyses using anterior-segment optical coherence tomography yielded more frequent classifications of angle closure than evaluations based on GS-1 gonioscopic images.

Outcomes after Hallux Rigidus surgery: Are we measuring what is important to patients? A qualitative analysis of patient-identified outcomes

Background Hallux rigidus (HR) is the most common form of arthritis in the foot. The joint pain and loss of motion from HR experienced during walking may lead to a significant reduction in activity and quality of life, and advanced cases may require surgery. HR surgical outcomes are often evaluated quantitatively with more generic measures not designed specifically to assess HR outcomes. The goal of these study was to determine outcomes that are important to patients with HR, possibly those that are not addressed by more general foot and ankle measures. Methods Semi-structured interviews with HR surgical patients 4–9 years after surgery were conducted. Interviews were analyzed using a team-based, iterative inductive-deductive approach to identify outcomes important to patients. Results Ten patients were interviewed; five who received motion-sparing surgery and five with fusion surgery. From interviews, seven themes were identified: pain, first MTPJ motion, walking ability, physical activity, footwear, forefoot appearance, and pain in other areas of the body. Conclusions Post-HR surgery patients indicated outcomes of importance that are not addressed by more general instruments. Specifically, forefoot appearance and pain in other areas of the body are not addressed by commonly used instruments. In addition, patient experiences of pain are more granular and HR-specific than the generic pain items used by other instruments. Patients facing HR surgery may benefit from outcome measures that are more specific to the HR surgical experience and include outcomes that patients share are important to them that are not included in commonly used outcome instruments.

Correction: Analysis of IPV success treatment from an AI approach

Enhancing mechanical and freeze-thaw performance of MICP-treated sand through palm fiber reinforcement: A sustainable approach for sandy soil stabilization

The integration of palm fiber with Microbially Induced Calcite Precipitation (MICP) technology offers a sustainable and bio-based approach to enhance the mechanical performance and durability of sandy soils, particularly under freeze-thaw conditions. In this study, a systematic experimental investigation examines the effects of varying palm fiber contents (0%−0.30%) on the bearing capacity, crust thickness, calcium carbonate deposition, and freeze-thaw resistance of MICP-treated sand. Results indicate that mechanical performance improves with increasing fiber content, peaking at 0.15%, beyond which the benefits diminish due to fiber agglomeration. At the optimal dosage, the bearing capacity increases by 24%, crust thickness by 70.5%, and calcium carbonate content reaches 16.8% compared to fiber-free MICP samples. Freeze-thaw tests demonstrate higher mass and strength retention, indicating improved durability. Microstructural analyses using SEM, XRD, EDS, and FTIR reveal enhanced microbial attachment and uniform CaCO₃ precipitation along fiber-sand interfaces, which strengthens matrix cohesion. These findings uncover a hybrid bio-mechanical reinforcement mechanism and highlight the trade-offs between fiber dosage and pore connectivity. This study provides novel insights into fiber-assisted biomineralization and offers a viable pathway for environmentally friendly soil reinforcement. Furthermore, potential directions such as predictive modeling, biodegradability assessments, and field-scale application are proposed to support long-term geotechnical and ecological engineering deployment.

“Gender dimensions of quality of life”: The determinants of life satisfaction among South Africans residing in the Gauteng province—Using a multilevel psychodemographic analysis of the GCRO’s quality of life survey (2009–2024)

This study examined gender-related differences in life satisfaction among 70,314 South Africans (33,325 males and 36,989 females) using data from the Gauteng City-Region Observatory (2009–2024). Life satisfaction levels (suffering, struggling, thriving) were analyzed in relation to demographic factors including age, education, employment, income, and population group. Using multilevel psychodemographic analyses, the study uniquely identifies how these factors operate differently for males and females over a 15-year period. Education, employment, and income were significant predictors of life satisfaction (all p < 0.01); for males, income had stronger indirect effects, while for females, education (β = 0.40, p < 0.01) and income (β = 0.93, p < 0.001) exerted direct effects. These findings reveal gender-specific pathways to life satisfaction, highlighting the importance of considering both individual and structural determinants. The results have policy relevance by underscoring the need for targeted, gender-sensitive interventions in education, employment, and income support to enhance well-being and promote equity among South Africans in Gauteng province.

Adolescence and online vulnerability: The role of fear of missing out (FoMO): A cross-sectional study during the third wave of the COVID-19 pandemic

The use of social networking sites (SNSs) has increased significantly in recent years, particularly among adolescents. These platforms have profoundly reshaped the way adolescents interact with family, peers, and strangers. However, SNSs may also expose users to specific vulnerabilities, such as victimisation, with detrimental effects on mental health and psychophysical well-being. This study examined the relationship between age and online vulnerability, with a focus on the mediating role of fear of missing out (FoMO). A cross-sectional study was conducted with 360 adolescents (meanage = 16.95 years; SDage = 1.29 years; age range = 14–19 years; 183 females). Results indicated that FoMO mediated the association between age and online vulnerability, suggesting that younger adolescents may be particularly susceptible to online vulnerability due to their heightened FoMO. These findings underscore the importance of addressing emotional and psychological factors in efforts to reduce online vulnerability. Implications for internet literacy education and preventative strategies are discussed, along with limitations and future research directions.

Incidence and risk factors of neonatal hypoglycemia at Hawassa University Comprehensive Specialized Hospital

Background Neonatal hypoglycemia is the most common metabolic emergency in neonates, with a reported incidence of 15% among neonates overall and 50% among high-risk newborns. If neonatal hypoglycemia is not diagnosed and managed promptly and properly, it can result in brain damage, neurological problems, and death. Over one-third of hypoglycemic neonates die in resource-limited settings. This study aimed to assess the incidence and risk factors of neonatal hypoglycemia in the neonatal intensive care unit at Hawassa University Comprehensive Specialized Hospital, Sidama Region, Ethiopia. Methods A retrospective cohort study was conducted among 308 neonates admitted to Hawassa University Comprehensive Specialized Hospital from July 2023 to July 2024. Data were extracted using a standard abstraction format from medical records. Descriptive statistics were summarized using tables and graphs. The Kaplan-Meier survival function and log-rank were used to show a hypoglycemia-free survival rate. An adjusted hazard ratio (aHR) with 95% Confidence interval was used to measure the strength of the association, and the statistical significance was declared at a p-value of less than or equal to 0.05. The Cox proportional hazard regression model assumption was checked by using the Schoenfeld residual test. Results The study found that the incidence rate of neonatal hypoglycemia was 3.1 (95% CI: 2.3–4.2) per 100 neonatal days of observation. Female gender (aHR: 3.4; 95% CI: 1.68, 6.83), neonatal sepsis (aHR: 2.1; 95% CI: 1.10–4.00), cesarean section (aHR: 2.1; 95% CI: 1.10, 4.21), preterm (aHR: 5.1; 95% CI: 2.41–10.9), and gestational diabetes mellitus (aHR: 3.4; 95% CI: 1.45, 8.11) were predictors of neonatal hypoglycemia. Conclusion The study found that the incidence rate of neonatal hypoglycemia was 3.1 per 100 days. Female sex, preterm birth, neonatal sepsis, cesarean delivery, and maternal gestational diabetes were independently associated with a higher incidence of hypoglycemia. These findings highlight the importance of close monitoring of blood glucose levels in at-risk neonates during neonatal intensive care unit admission.

Preparation of long afterglow luminescent road marking coatings and its applicability simulation

Traffic markings have inadequate nighttime visibility for safe driving. To address this and reduce nighttime accidents, we researched and developed long afterglow luminescent coatings, including performance and simulation studies. By varying the dosage of luminescent powder, glass powder, and titanium dioxide, nine groups of coatings were fabricated via blending. Their luminescence and chromaticity properties were investigated. Using CW-VIKOR comprehensive evaluation, the optimal formulation was selected. This optimal coating was further analyzed using Twinmotion and DIALux simulations, assessing marking visibility and tunnel illumination effects with the coatings.Results revealed that increased luminescent powder dosage enhanced luminescence performance. The optimal formulation (30% luminescent powder, 8% glass powder, 4% titanium dioxide) achieved the best luminescence performance and colorimetric characteristics, sustaining light emission for 9 hours with good road performance. Twinmotion simulations demonstrated better nighttime visibility at different brightness levels. DIALux simulations showed a 22.3% increase in average tunnel brightness after coating deployment, significantly improving lighting effects.

Media choice and audience perceptions: Evidence from visual framing of immigration in news stories

Where does visual media bias come from, and how is it reinforced? This study investigates the often overlooked interplay between the visual frames chosen by media outlets for politically charged news stories and how these frames are perceived by their audiences. Using computer vision tools and qualitative content analysis, we analyzed over 2,000 images from 393 media outlets on X. Our findings reveal that U.S. media outlets across the political spectrum consistently emphasize visual narratives that align with their ideological stances while minimizing opposing viewpoints. Their partisan audiences assign identity-driven interpretations to identical visuals, turning them into instruments of antagonistic narratives even without any textual or source cues. This reveals a critical implication: the perceived bias is not merely a product of the media’s framing choices, but also a reflection of how audiences project their ideological filters onto these frames. This study helps us understand how the interplay between media frame curation and partisan interpretations reinforces and perpetuates existing divides.