Model Bias Research Articles

The relationship between trait anger and reactive aggressive behavior in middle school students: the mediating role and intervention of hostile attribution bias

BackgroundThe reactive aggressive behavior in individuals typically shows a rapid growth trend as individuals enter adolescence, and peaks during middle-school period. According to the Comprehensive Cognitive Model of Trait Anger, trait anger and hostile attribution bias play important roles in the development of reactive aggressive behavior. Based on this, current study explored the relationship between trait anger and reactive aggressive behavior in middle school students, as well as the mediating role of hostile attribution bias and interventions.MethodsThe current study consisted of three sub-studies. Study 1 recruited 87 middle school students with an average age of 12.367 ± 0.889 years, investigated the relationship between trait anger and reactive aggressive behavior, as well as the mediating role of trait hostile attribution bias. Study 2 recruited 62 middle school students with an average age of 13.376 ± 0.963 years, investigated the relationship between trait anger and reactive aggressive behavior, as well as the mediating role of state hostile attribution bias. Study 3 recruited 80 middle school students with an average age of 13.392 ± 0.977 years, implemented an intervention targeting trait hostile attribution bias in middle school students with high trait anger to reduce their reactive aggressive behavior. In current study, data management was performed using SPSS 22.0. Descriptive statistics, independent samples t-test, paired samples t-test, repeated measures analysis of variance (ANOVA), and path analysis were used for statistical analysis.FindingsThe results of Study 1 showed that trait anger predicted reactive aggressive behavior through trait hostile attribution bias. The results of Study 2 indicated that trait and state hostile attribution bias played mediating role intermediary, and trait hostile attribution bias had a stronger mediating effect than state hostile attribution bias. The results of Study 3 suggested that the intervention effectively decreased trait hostile attribution bias and reactive aggressive behavior.ConclusionsTrait anger can predict the reactive aggressive behavior of junior high school students, with trait hostility attribution bias and state hostility attribution bias mediating this relationship. Intervening in the hostility attribution bias of high-anger junior high school students can effectively reduce their reactive aggressive behavior.

Fluctuation suppression for UI-driven bias accuracy enhancement based on micro-machined gyroscopes array

Unknown-input (UI) fluctuations sourcing from external environment severely deteriorate the accuracy of micro-machined gyroscope because of the unpredictable statistical characteristics. To enhance the accuracy of a four-micro-gyros array under UI fluctuations, an array-based consensus strategy (ACS) consisting of a UI-driven bias model, a local estimator, and an array fusion estimator is proposed. The UI-driven bias model is originally constructed as two behavior-contrasted independent items according to different drift characteristics of individual gyro. For the local estimator, a UI-decoupling operation is proposed to transform the variance-unknown UI model implicitly into a linear combination of variance-estimated variables, which transforms the non-stationary model to an equivalent stationary model. For the array fusion estimator, the reconstructed weight coefficient is designed based on the support theory and Markowitz mean-variance theory to evaluate the confidence level of gyros. Experiment results show that the root-mean-square error (RMSE) and Allan bias instability of the estimated angular rate under UIs are 7.9×10−3 °/s and 3.87 °/h, which are respectively reduced by 85.1 % and 35.1 % compared with the average original outputs of the gyros array.

Model Biases in Simulating Extreme Sea Ice Loss Associated With the Record January 2022 Arctic Cyclone

AbstractIn January 2022, the strongest Arctic cyclone on record resulted in a record weekly loss in sea ice cover in the Barents‐Kara‐Laptev seas. While ECMWF operational forecasts skillfully predicted the cyclone, the loss in sea ice was poorly predicted. We explore the ocean's response to the cyclone using observations from an Argo float that was profiling in the region, and investigate model biases in simulating the observed sea ice loss in a fully coupled GCM. The observations showed changes over the whole ocean column in the Barents Sea after the passage of the storm, cooling and mixing with enough implied heat release to melt roughly 1 m of sea ice. We replicate the observed cyclone in the GCM by nudging the model's winds to observations above the boundary layer. In these simulations, the associated loss of sea ice is only about 10%–15% of the observed loss, and the ocean exhibits very small changes in response to the cyclone. With the use of a simple 1‐D ice‐ocean model, we find that the overly strong ocean stratification in the GCM may be a significant source of model bias in its simulated response to the cyclone. However, even initialized with observed stratification profiles, the 1‐D model also underestimated mixing and sea ice melt relative to the observations.

Performance and Biases of Large Language Models in Public Opinion Simulation

Performance and Biases of Large Language Models in Public Opinion Simulation

The Temporal and Directional Relationship Between Group-Level Implicit and Explicit Gender Bias

Explicit and implicit gender-science and gender-career biases have shifted toward neutrality in the past decade. Researchers speculate that these changes result from women's increased visibility in the science field and job market, but little is known about how the changes in group-level explicit and implicit gender-science and gender-career bias relate to one another over time. Building on contemporary models of group-level bias, this study investigates the temporal and directional relationship between group-level implicit and explicit gender bias between 2007 to 2016 using multivariate multilevel modeling. We found that lower group-level implicit bias in a previous month predicts lower group-level explicit bias in the following month. We also found evidence that group-level explicit bias in a previous month was positively associated with group-level implicit bias in the following month. These findings have practical and theoretical implications for understanding the bidirectional relationship between group-level implicit and explicit biases over time.

The effects of strength and activation level of belief on belief-biased reasoning.

The belief bias effect designates the tendency to judge the validity of a conclusion based on its believability. Most studies have compared highly believable with unbelievable conclusions when examining belief-biased reasoning. In two studies, we examine a hypothesis raised by Banks (2013, Cognitive Science, 37[3], 544-577), who postulated that level of activation of belief should affect the extent to which reasoning is biased. Here, a first study found that moderately believable conclusions produced a significant biasing effect, but one that was smaller than with highly believable conclusions. The second study found that activating the category referenced by a conclusion significantly increased the biasing effect of the same conclusion. These results suggest that strength of belief is a necessary parameter of any model of belief bias. They also show that context can affect the extent to which belief affects reasoning.

Machine learning: a new era for cardiovascular pregnancy physiology and cardio-obstetrics research.

The maternal cardiovascular system undergoes functional and structural adaptations during pregnancy and postpartum to support increased metabolic demands of offspring and placental growth, labor, and delivery, as well as recovery from childbirth. Thus, pregnancy imposes physiological stress upon the maternal cardiovascular system, and in the absence of an appropriate response it imparts potential risks for cardiovascular complications and adverse outcomes. The proportion of pregnancy-related maternal deaths from cardiovascular events has been steadily increasing, contributing to high rates of maternal mortality. Despite advances in cardiovascular physiology research, there is still no comprehensive understanding of maternal cardiovascular adaptations in healthy pregnancies. Furthermore, current approaches for the prognosis of cardiovascular complications during pregnancy are limited. Machine learning (ML) offers new and effective tools for investigating mechanisms involved in pregnancy-related cardiovascular complications as well as the development of potential therapies. The main goal of this review is to summarize existing research that uses ML to understand mechanisms of cardiovascular physiology during pregnancy and develop prediction models for clinical application in pregnant patients. We also provide an overview of ML platforms that can be used to comprehensively understand cardiovascular adaptations to pregnancy and discuss the interpretability of ML outcomes, the consequences of model bias, and the importance of ethical consideration in ML use.

What Drives the Spread and Bias in the Surface Impact of Sudden Stratospheric Warmings in CMIP6 Models?

Abstract This study evaluates the representation of the composite-mean surface response to sudden stratospheric warmings (SSWs) in 28 CMIP6 models. Most models can reproduce the magnitude of the SLP response over the Arctic, although the simulated Arctic SLP response varies from model to model. Regarding the structure of the SLP response, most models exhibit a basin-symmetric negative Northern Annular Mode (NAM)-like response with a cyclonic Pacific SLP response, whereas the reanalysis shows a highly basin-asymmetric negative NAO-like response without a robust Pacific center. We then explore the drivers of these model biases and spread by applying a multiple linear regression (MLR). The results show that the polar cap temperature anomalies at 100 hPa (ΔT100) modulate the magnitude of both the Arctic SLP response and the cyclonic Pacific SLP response. Apart from ΔT100, the intensity and latitudinal location of the climatological eddy-driven jet in the troposphere also affect the magnitude of the Arctic SLP response. The compensation of model biases in these two tropospheric metrics and the good model representation of ΔT100 explain the good agreement between the ensemble mean and the reanalysis on the magnitude of the Arctic SLP response, as indicated by the fact that the ensemble mean lies well within the reanalysis uncertainty range and that the reanalysis mean sits well within the model distribution. The Niño-3.4 SST anomalies and North Pacific SST dipole anomalies together with ΔT100 modulate the cyclonic Pacific SLP response. In this case, biases in both oceanic drivers work in the same direction and lead to the cyclonic Pacific SLP response in models that are not present in the reanalysis. Significance Statement Sudden stratospheric warmings (SSWs) represent an important source of skill for forecasting winter weather on subseasonal-to-seasonal time scales. To what extent SSWs could be used to improve the prediction of surface weather depends on how well stratosphere–troposphere coupling associated with SSWs is represented in climate models. Therefore, we evaluate the representation of stratosphere–troposphere coupling associated with SSWs in 28 state-of-the-art climate models. The representation is found to diverge widely among climate models, and some are biased noticeably from the reanalysis. The models’ spread and bias are largely driven by five major factors and can be reduced substantially by making bias corrections to these factors.

ProactiV: Studying Deep Learning Model Behavior Under Input Transformations.

Deep learning (DL) models have shown performance benefits across many applications, from classification to image-to-image translation. However, low interpretability often leads to unexpected model behavior once deployed in the real world. Usually, this unexpected behavior is because the training data domain does not reflect the deployment data domain. Identifying a model's breaking points under input conditions and domain shifts, i.e., input transformations, is essential to improve models. Although visual analytics (VA) has shown promise in studying the behavior of model outputs under continually varying inputs, existing methods mainly focus on per-class or instance-level analysis. We aim to generalize beyond classification where classes do not exist and provide a global view of model behavior under co-occurring input transformations. We present a DL model-agnostic VA method (ProactiV) to help model developers proactively study output behavior under input transformations to identify and verify breaking points. ProactiV relies on a proposed input optimization method to determine the changes to a given transformed input to achieve the desired output. The data from this optimization process allows the study of global and local model behavior under input transformations at scale. Additionally, the optimization method provides insights into the input characteristics that result in desired outputs and helps recognize model biases. We highlight how ProactiV effectively supports studying model behavior with example classification and image-to-image translation tasks.

Understanding world models through multi-step pruning policy via reinforcement learning

In model-based reinforcement learning, the conventional approach to addressing world model bias is to use gradient optimization methods. However, using a singular policy from gradient optimization methods in response to a world model bias inevitably results in an inherently biased policy. This is because of constraints on the imperfect and dynamic data of state-action pairs. The gap between the world model and the real environment can never be completely eliminated. This article introduces a novel approach that explores a variety of policies instead of focusing on either world model bias or singular policy bias. Specifically, we introduce the Multi-Step Pruning Policy (MSPP), which aims to reduce redundant actions and compress the action and state spaces. This approach encourages a different perspective within the same world model. To achieve this, we use multiple pruning policies in parallel and integrate their outputs using the cross-entropy method. Additionally, we provide a convergence analysis of the pruning policy theory in tabular form and an updated parameter theoretical framework. In the experimental section, the newly proposed MSPP method demonstrates a comprehensive understanding of the world model and outperforms existing state-of-the-art model-based reinforcement learning baseline techniques.

Intelligent fault diagnosis for unbalanced battery data using adversarial domain expansion and enhanced stochastic configuration networks

An accurate and efficient fault diagnosis method for battery systems is crucial to ensuring the safety of battery packs. Addressing the issue of insufficient actual fault data in battery operations, this paper proposes an intelligent fault diagnosis method based on feature-enhanced stochastic configuration networks and adversarial domain expansion of imbalanced battery fault data (AFDEM-FESCN). Firstly, we designed an adversarial fault domain data expansion method (AFDEM). By learning the distribution of fault data through adversarial training, the distribution of sample domains is balanced, thereby reducing model bias. Subsequently, we adjusted the distribution of SCN iterative parameters and added a linear feature layer. This enhances the feature extraction capability of the network through distribution overlay, enabling fault diagnosis. Finally, the effectiveness and feasibility of the proposed method were validated through a practical battery system fault diagnosis case, achieving a diagnostic accuracy of 92.1%. Experimental results demonstrate that the AFDEM-FESCN method exhibits good accuracy in battery system fault diagnosis, providing an effective solution to the challenge of imbalanced data in intelligent fault diagnosis.

Explaining transfer learning models for the detection of COVID-19 on X-ray lung images

Amidst the coronavirus disease 2019 (COVID-19) pandemic, researchers are exploring innovative approaches to enhance diagnostic accuracy. One avenue is utilizing deep learning models to analyze lung X-ray images for COVID-19 diagnosis, complementing existing tests like reverse transcription polymerase chain reaction (RT-PCR). However, trusting these models, often viewed as black boxes, presents a challenge. To address this, six explainable artificial intelligence (XAI) techniques: local interpretable model agnostic explanations (LIME), Shapley additive explanations (SHAP), integrated gradients, smooth-grad, gradient-weighted class activation mapping (Grad-CAM), and Layer-CAM are applied to interpret four transfer learning models. These models: VGG16, ResNet50, InceptionV3, and DenseNet121 are analyzed to understand their workings and the rationale behind their predictions. Validating the results with medical experts poses difficulties due to time and resource constraints, alongside the scarcity of annotated X-ray datasets. To address this, a voting mechanism employing different XAI methods across various models is proposed. This approach highlights regions of lung infection, potentially reducing individual model biases stemming from their structures. If successful, this research could pave the way for an automated system for annotating infection regions, bolstering confidence in predictions and aiding in the development of more effective diagnostic tools for COVID-19.

Prevalence and management of pulmonary nodules: a systematic review and meta-analysis.

Pulmonary nodules are small, focal lesions often identified via computed tomography (CT) scans. Although the majority are benign, a small percentage of them may be malignant or potentially become malignant, underscoring the importance of early detection and effective management. This study systematically reviews the epidemiology, risk factors, and management strategies for pulmonary nodules, comparing findings across Chinese and non-Chinese populations to better inform the actuarial calculations for predicting the demand of medical services for patients with pulmonary nodules. We performed a systematic analysis of the PubMed and China Knowledge Infrastructure (CNKI) databases for studies reporting the detection rate of pulmonary nodules through CT scans. Both cross-sectional studies and the baseline data from longitudinal studies were included. A modified version of the Newcastle-Ottawa Scale was used to assess the risk of bias and random effect models were used to estimate the overall prevalence. We identified 32 studies and included 24 of them in our meta-analysis. Pooled analysis showed that the overall prevalence of pulmonary nodules was 0.27 (95% confidence interval: 0.25-0.29) after outliers removal. Subgroup analysis showed that there was no significant difference for prevalence between Chinese and non-Chinese populations. Males (0.38) were shown to have slightly higher prevalence compared to females (0.36), but not significant (P=0.88). Age and smoking are the most frequently reported risk factors by studies. Overall, 27% of participants were positive for pulmonary nodules. Advancing age and smoking were consistently identified as a key risk factor for the incidence of pulmonary nodules. Although the management strategies are different across studies, recent guidelines recommend personalized management strategies, prioritizing nodule size, characteristics, and individual risk factors to optimize outcomes.

Impact of ocean heat transport on sea ice captured by a simple energy balance model

Future projections of Arctic and Antarctic sea ice suffer from uncertainties largely associated with inter-model spread. Ocean heat transport has been hypothesised as a source of this uncertainty, based on correlations with sea ice extent across climate models. However, a physical explanation of what sets the sea ice sensitivity to ocean heat transport remains to be uncovered. Here, we derive a simple equation using an idealised energy-balance model that captures the emergent relationship between ocean heat transport and sea ice in climate models. Inter-model spread of Arctic sea ice loss depends strongly on the spread in ocean heat transport, with a sensitivity set by compensation of atmospheric heat transport and radiative feedbacks. Southern Ocean heat transport exhibits a comparatively weak relationship with Antarctic sea ice and plays a passive role secondary to atmospheric heat transport. Our results suggest that addressing ocean model biases will substantially reduce uncertainty in projections of Arctic sea ice.

Addressing Demographic Bias in Age Estimation Models through Optimized Dataset Composition

Bias in facial recognition systems often results in unequal performance across demographic groups. This study addresses this by investigating how dataset composition affects the performance and bias of age estimation models across ethnicities. We fine-tuned pre-trained Convolutional Neural Networks (CNNs) like VGG19 on the diverse UTKFace dataset (23,705 samples: 10,078 White, 4526 Black, 3434 Asian) and APPA-REAL (7691 samples: 6686 White, 231 Black, 674 Asian). Our approach involved adjusting dataset compositions by oversampling minority groups or reducing samples from overrepresented groups to mitigate bias. We conducted experiments to identify the optimal dataset composition that minimizes performance disparities among ethnic groups. The primary performance metric was Mean Absolute Error (MAE), measuring the average magnitude of prediction errors. We also analyzed the standard deviation of MAE across ethnic groups to assess performance consistency and equity. Our findings reveal that simple oversampling of minority groups does not ensure equitable performance. Instead, systematic adjustments, including reducing samples from overrepresented groups, led to more balanced performance and lower MAE standard deviations across ethnicities. These insights highlight the importance of tailored dataset adjustments and suggest exploring advanced data processing methods and algorithmic tweaks to enhance fairness and accuracy in facial recognition technologies.

SMOTE-Tomek Re-sampling Based on Random Forest Method to Overcome Unbalanced Data for Multi-class Classification

Mobile app review data needs to be utilized to understand the app characteristics desired by users. App providers can improve app performance based on user preferences by using sentiment and emotion classification on app review data. However, problems that often arise in text-based analysis are data variation and data imbalance. This can lead to biased and inaccurate classification models. It is necessary to perform pre-processing to comprehend the data requirements and implement feature extraction for word weighting to overcome the variation in the data. In addition, re-sampling techniques are also needed to overcome the imbalance in sample distribution. Re-sampling techniques such as Tomek Links and SMOTE only focus on majority or minority data. This research applies the SMOTE-Tomek merging technique, aiming at not only the minority data but also the majority data. The model performance becomes better because the technique combines oversampling and under-sampling of the majority of data to eliminate the noise of data. The data was modeled using an Ensemble Learning Random Forest for classification. The model performance resulted in a Precision value of 84%, Recall of 84%, F1-Score of 84%, and Accuracy of 84%. Furthermore, the model was optimized using GridSearchCV and obtained an increase in Precision 85%, Recall 85%, F1-Score 85%, and Accuracy 85%.

Love Lost in Translation: Avoidant Individuals Inaccurately Perceive Their Partners' Positive Emotions During Love Conversations.

Empathic accuracy-the ability to decipher others' thoughts and feelings-promotes relationship satisfaction. Those high in attachment avoidance tend to be less empathically accurate; however, past research has been limited to relatively negative or neutral contexts. We extend work on attachment and empathic accuracy to the positive context of love. To do so, we combined data from three dyadic studies (N = 303 dyads) in which couple members shared a time of love and rated each other's positive emotions. Using the Truth and Bias Model of Judgment, we found that individuals higher (vs. lower) in attachment avoidance were less accurate in inferring their partners' positive emotions during the conversation, but did not systematically over- or under-perceive their partners' positive emotions. Our results suggest that avoidant individuals may be less sensitive to positive cues in their relationships, potentially reducing relational intimacy.

A Unique Conditions Model for Landslide Susceptibility Mapping

Several methods and approaches have been proposed to assess landslide susceptibility. The likelihood of landslides occurring can be determined by applying statistical models to historical landslides, taking into account controlling factors. Popular methods for predicting the probability of landslides are weights-of-evidence and logistic regression. We discuss the assumptions and interpretations of these methods, the relationships between them, and their strengths and weaknesses in case of categorical factors. Of particular interest is the conditional independence of the controlling factors and its effect on model bias. To avoid lack of conditional independence of factors and model bias, we present a unique conditions model that is always unbiased. To illustrate the theoretical developments, a practical application is given using observed landslides and geo-environmental factors from a previous study. The unique conditions model appears superior to the other models.

Revising VOC emissions speciation improves the simulation of global background ethane and propane.

Non-Methane Volatile Organic Compounds (NMVOCs) generate ozone (O3) when they are oxidized in the presence of oxides of nitrogen, modulate the oxidative capacity of the atmosphere and can lead to the formation of aerosol. Here, we assess the capability of a chemical transport model (GEOS-Chem) to simulate NMVOC concentrations by comparing ethane, propane and higher alkane observations in remote regions from the NOAA Flask Network and the World Meteorological Organization's Global Atmosphere Watch (GAW) network. Using the Community Emissions Data System (CEDS) inventory we find a significant underestimate in the simulated concentration of both ethane (35%) and propane (64%), consistent with previous studies. We run a new simulation where the total mass of anthropogenic NMVOC emitted in a grid box is the same as that used in CEDS, but with the NMVOC speciation derived from regional inventories. For US emissions we use the National Emissions Inventory (NEI), for Europe we use the UK National Atmospheric Emissions Inventory (NAEI), and for China, the Multi-resolution Emission Inventory for China (MEIC). These changes lead to a large increase in the modelled concentrations of ethane, improving the mean model bias from -35% to -4%. Simulated propane also improves (from -64% to -48% mean model bias), but there remains a substantial model underestimate. There were relatively minor changes to other NMVOCs. The low bias in simulated global ethane concentration is essentially removed, resolving one long-term issue in global simulations. Propane concentrations are improved but remain significantly underestimated, suggesting the potential for a missing global propane source. The change in the NMVOC emission speciation results in only minor changes in tropospheric O3 and OH concentrations.

Deep Learning-Based Multifidelity Surrogate Modeling for High-Dimensional Reliability Prediction

Abstract Multifidelity surrogate modeling offers a cost-effective approach to reducing extensive evaluations of expensive physics-based simulations for reliability prediction. However, considering spatial uncertainties in multifidelity surrogate modeling remains extremely challenging due to the curse of dimensionality. To address this challenge, this paper introduces a deep learning-based multifidelity surrogate modeling approach that fuses multifidelity datasets for high-dimensional reliability analysis of complex structures. It first involves a heterogeneous dimension transformation approach to bridge the gap in terms of input format between the low-fidelity and high-fidelity domains. Then, an explainable deep convolutional dimension-reduction network (ConvDR) is proposed to effectively reduce the dimensionality of the structural reliability problems. To obtain a meaningful low-dimensional space, a new knowledge reasoning-based loss regularization mechanism is integrated with the covariance matrix adaptation evolution strategy (CMA-ES) to encourage an unbiased linear pattern in the latent space for reliability prediction. Then, the high-fidelity data can be utilized for bias modeling using Gaussian process (GP) regression. Finally, Monte Carlo simulation (MCS) is employed for the propagation of high-dimensional spatial uncertainties. Two structural examples are utilized to validate the effectiveness of the proposed method.