Inference Model Research Articles

Random-effects substitution models for phylogenetics via scalable gradient approximations.

Phylogenetic and discrete-trait evolutionary inference depend heavily on an appropriate characterization of the underlying character substitution process. In this paper, we present random-effects substitution models that extend common continuous-time Markov chain models into a richer class of processes capable of capturing a wider variety of substitution dynamics. As these random-effects substitution models often require many more parameters than their usual counterparts, inference can be both statistically and computationally challenging. Thus, we also propose an efficient approach to compute an approximation to the gradient of the data likelihood with respect to all unknown substitution model parameters. We demonstrate that this approximate gradient enables scaling of sampling-based inference, namely Bayesian inference via Hamiltonian Monte Carlo, under random-effects substitution models across large trees and state-spaces. Applied to a dataset of 583 SARS-CoV-2 sequences, an HKY model with random-effects shows strong signals of nonreversibility in the substitution process, and posterior predictive model checks clearly show that it is a more adequate model than a reversible model. When analyzing the pattern of phylogeographic spread of 1441 influenza A virus (H3N2) sequences between 14 regions, a random-effects phylogeographic substitution model infers that air travel volume adequately predicts almost all dispersal rates. A random-effects state-dependent substitution model reveals no evidence for an effect of arboreality on the swimming mode in the tree frog subfamily Hylinae. Simulations reveal that random-effects substitution models can accommodate both negligible and radical departures from the underlying base substitution model. We show that our gradient-based inference approach is over an order of magnitude more time efficient than conventional approaches.

Fast Coding Unit Partitioning Algorithm for Video Coding Standard Based on Block Segmentation and Block Connection Structure and CNN

The recently introduced Video Coding Standard, VVC, presents a novel Quadtree plus Nested Multi-Type Tree (QTMTT) block structure. This structure enables a more flexible block partition and demonstrates enhanced compression performance compared to its predecessor, HEVC. However, The introduction of the new structure has led to a more complex partition search process, resulting in a considerable increase in time complexity. The QTMTT structure yields diverse Coding Unit (CU) block sizes, posing challenges for CNN model inference. In this study, we propose a representation structure termed Block Segmentation and Block Connection (BSC), rooted in texture features. This ensures that partial CU blocks are uniformly represented in size. To address different-sized CUs, various levels of CNN models are designed for prediction. Moreover, we introduce a post-processing method and a multi-thresholding scheme to further mitigate errors introduced by CNNs. This allows for flexible and adjustable acceleration, achieving a trade-off between coding time complexity and performance. Experimental results indicate that, in comparison to VTM-10.0, our “Fast” scheme reduces the average complexity by 57.14% with a 1.86% increase in BDBR. Meanwhile, the “Moderate” scheme reduces average complexity by 50.14% with only a 1.39% increase in BDBR.

Phylogenetic placement and comparative analysis of the mitochondrial genomes of Idiostoloidea (Hemiptera: Heteroptera).

The classification system and the higher level phylogenetic relationships of Pentatomomorpha, the second largest infraorder of Heteroptera (Insecta: Hemiptera), have been debated and remain controversial over decades. In particular, the placement and phylogenetic relationship of Idiostoloidea are not well resolved, which hampers a better understanding of the evolutionary history of Pentatomomorpha. In this study, for the first time, we reported the complete mitochondrial genome for two narrowly distributed families of Idiostoloidea (including Idiostolidae and Henicocoridae), respectively. The length of the mitochondrial genome of Monteithocoris hirsutus and Henicocoris sp. is 16,632 and 16,013 bp, respectively. The content of AT is ranging from 75.15% to 80.48%. The mitogenomic structure of Idiostoloidea is highly conservative and there are no gene arrangements. By using the Bayesian inference, maximum likelihood, and Bayesian site-heterogeneous mixture model, we inferred the phylogenetic relationships within Pentatomomorpha and estimated their divergence times based on concatenated mitogenomes and nuclear ribosomal genes. Our results support the classification system of six superfamilies within Pentatomomorpha and confirm the monophyletic groups of each superfamily, with the following phylogenetic relationships: (Aradoidea + (Pentatomoidea + (Idiostoloidea + (Coreoidea + (Pyrrhocoroidea + Lygaeoidea))))). Furthermore, estimated divergence times revealed that most pentatomomorphan superfamilies and families diverged during the Late Jurassic to Early Cretaceous, which coincides with the explosive radiation of angiosperms.

Robust Ensemble of Two Different Multimodal Approaches to Segment 3D Ischemic Stroke Segmentation Using Brain Tumor Representation Among Multiple Center Datasets.

Ischemic stroke segmentation at an acute stage is vital in assessing the severity of patients' impairment and guiding therapeutic decision-making for reperfusion. Although many deep learning studies have shown attractive performance in medical segmentation, it is difficult to use these models trained on public data with private hospitals' datasets. Here, we demonstrate an ensemble model that employs two different multimodal approaches for generalization, a more effective way to perform on external datasets. First, after we jointly train a segmentation model on diffusion-weighted imaging (DWI) and apparent diffusion coefficient (ADC) MR modalities, the model is inferred on the DWI images. Second, a channel-wise segmentation model is trained by concatenating the DWI and ADC images as input, and then is inferred using both MR modalities. Before training with ischemic stroke data, we utilized BraTS 2021, a public brain tumor dataset, for transfer learning. An extensive ablation study evaluates which strategy learns better representations for ischemic stroke segmentation. In our study, nnU-Net well-known for robustness is selected as our baseline model. Our proposed method is evaluated on three different datasets: the Asan Medical Center (AMC) I and II, and the 2022 Ischemic Stroke Lesion Segmentation (ISLES). Our experiments are widely validated over a large, multi-center, and multi-scanner dataset with a huge amount of 846 scans. Not only stroke lesion models can benefit from transfer learning using brain tumor data, but combining the MR modalities using different training schemes also highly improves segmentation performance. The method achieved a top-1 ranking in the ongoing ISLES'22 challenge and performed particularly well on lesion-wise metrics of interest to neuroradiologists, achieving a Dice coefficient of 78.69% and a lesion-wise F1 score of 82.46%. Also, the method was relatively robust on the AMC I (Dice, 60.35%; lesion-wise F1, 68.30%) and II (Dice; 74.12%; lesion-wise F1, 67.53%) datasets in different settings. The high segmentation accuracy of our proposed method could improve radiologists' ability to detect ischemic stroke lesions in MRI images. Our model weights and inference code are available on https://github.com/MDOpx/ISLES22-model-inference .

Modeling and adaptive neuro-fuzzy inference system control of quarter electric vehicle

Electric vehicles (EVs) have gained importance in recent years, prompting the development of several control systems to improve their efficiency and performance. In this work, a quarter electric vehicle (QEV) was controlled using a conventional proportional integral derivative (PID) and fuzzy controller to examine and compare with the response of the adaptive neuro-fuzzy inference system (ANFIS) controller. The response of the ANFIS controller was evaluated using MATLAB/Simulink according to different parameters and compared with those of other controllers. In addition, the simulation was based on different driving conditions such as the acceleration and deceleration modes and the type of road: wet and dry. The simulations were carried out on a longitudinal electric vehicle model based on a brushless DC motor, including the Pacejka tire model. The results showed that the ANFIS controller outperformed the PID and fuzzy logic controllers, providing superior dynamic responsiveness and stability when the ANFIS controller smoothly followed the input speed and the longitudinal slip value reached 3%.

Private-preserving language model inference based on secure multi-party computation

Private-preserving language model inference based on secure multi-party computation

Digitalization of corporate finance and firm performance: global evidence and analysis

Purpose As of now, the digitization of corporate finance presents a paradigm shift in business strategy, innovation, financing and managerial capability around the globe. However, the prevailing finance scholarly works hardly document the impact of the digitalization of corporate finance on firm performance with global evidence and analysis. Hence, the contemporary debate on whether firm performance is genuinely stimulated because of the digitalization of corporate finance or not has been a pressing issue in the relevant literature. Therefore, the purpose of this study is to identify a data-driven, concise response to an unaddressed finance issue if the performance of high-digitalized firms (HDFs) outperforms that of their counterpart peers for wealth maximization. Design/methodology/approach The first stage test models examine the firm performance of relatively high-digitalized firms as opposed to low-digitalized firms based on the system GMM. The second stage test of the probabilistic (logit) model infers that the probability of being HDFs explores because of better performance. Then, the authors execute robust checks based on the different quantile regressions and Z-score-based system GMM. In addition, the authors recheck and present the test results of the fixed effect and random effect to capture time-invariant individual heterogeneity. Finally, the supplementary test findings of firms’ credit strength by using Altman five- and four-factor Z-score models are presented. Findings By using cross-country panel analysis as 15 years’ test bed for HDFs and low digitalized firms (LDFs), the test results indicate that the overall firm performance of a digitalized firm is significantly better than that of a non-digitalized firm. The global evidence documents that HDFs are exposed to higher values and are financially more persistent as compared to their counterparts. The finding is remarkably concomitant across several possible subsample analysis, such as country–industry–size–period analysis. Practical implications This study can be remarkably effective in encouraging managers, policymakers and investors to acknowledge the need for adopting the required digitalization. Overall, this original study addresses a core research gap in the corporate finance literature and remarkably provides further direction to rethink the assumptions of firm digitalization on additive value and thereby identify optimal decisions for wealth maximization. The findings also imply that investors require an additional risk premium if they invest in relatively LDFs, which have relatively lower market value and weaker firm performance. Originality/value From an investors point of view, the academic novelty contributes to an innovative and unsettled issue on the impact of digitization of corporate finance on firm performance because there is a new question of high or low digitization of corporate finance in the global market. Hence, this academic novelty contributes to sharing global evidence of the digitalization of corporate finance and its effect on firm performances. In addition, an intensive critical review analysis is conducted based on the most recent and relevant scholarly works published in the top-tier journals of finance and business stream to fix the hypothesis. Overall, this study addresses a core research gap in the corporate finance literature; notably provides further direction to rethink firm digitalization; and thereby identifies optimal decisions for shareholders’ wealth maximization.

CCTFv2: Modeling Cyber Competitions

Cyber competitions are usually team activities, where team performance not only depends on the members’ abilities but also on team collaboration. This seems intuitive, especially given that team formation is a well-studied discipline in competitive sports and project management, but unfortunately, team performance and team formation strategies are rarely studied in the context of cybersecurity and cyber competitions. Since cyber competitions are becoming more prevalent and organized, this gap becomes an opportunity to formalize the study of team performance in the context of cyber competitions. This work follows a cross-validating two-approach methodology. The first is the computational modeling of cyber competitions using Agent-Based Modeling. Team members are modeled, in NetLogo, as collaborating agents competing over a network in a red team/blue team match. Members’ abilities, team interaction and network properties are parametrized (inputs), and the match score is reported as output. The second approach is grounded in the literature of team performance (not in the context of cyber competitions), where a theoretical framework is built in accordance with the literature. The results of the first approach are used to build a causal inference model using Structural Equation Modeling. Upon comparing the causal inference model to the theoretical model, they showed high resemblance, and this cross-validated both approaches. Two main findings are deduced: first, the body of literature studying teams remains valid and applicable in the context of cyber competitions. Second, coaches and researchers can test new team strategies computationally and achieve precise performance predictions. The targeted gap used methodology and findings which are novel to the study of cyber competitions.

Kinetics-based inference of environment-dependent microbial interactions and their dynamic variation.

Microbial communities in nature are dynamically evolving as member species change their interactions subject to environmental variations. Accounting for such context-dependent dynamic variations in interspecies interactions is critical for predictive ecological modeling. In the absence of generalizable theoretical foundations, we lack a fundamental understanding of how microbial interactions are driven by environmental factors, significantly limiting our capability to predict and engineer community dynamics and function. To address this issue, we propose a novel theoretical framework that allows us to represent interspecies interactions as an explicit function of environmental variables (such as substrate concentrations) by combining growth kinetics and a generalized Lotka-Volterra model. A synergistic integration of these two complementary models leads to the prediction of alterations in interspecies interactions as the outcome of dynamic balances between positive and negative influences of microbial species in mixed relationships. The effectiveness of our method was experimentally demonstrated using a synthetic consortium of two Escherichia coli mutants that are metabolically dependent (due to an inability to synthesize essential amino acids) but competitively grow on a shared substrate. The analysis of the E. coli binary consortium using our model not only showed how interactions between the two amino acid auxotrophic mutants are controlled by the dynamic shifts in limiting substrates but also enabled quantifying previously uncharacterizable complex aspects of microbial interactions, such as asymmetry in interactions. Our approach can be extended to other ecological systems to model their environment-dependent interspecies interactions from growth kinetics.IMPORTANCEModeling environment-controlled interspecies interactions through separate identification of positive and negative influences of microbes in mixed relationships is a new capability that can significantly improve our ability to understand, predict, and engineer the complex dynamics of microbial communities. Moreover, the prediction of microbial interactions as a function of environmental variables can serve as valuable benchmark data to validate modeling and network inference tools in microbial ecology, the development of which has often been impeded due to the lack of ground truth information on interactions. While demonstrated against microbial data, the theory developed in this work is readily applicable to general community ecology to predict interactions among macroorganisms, such as plants and animals, as well as microorganisms.

Effectiveness of Using AI-Driven Hotspot Mapping for Active Case Finding of Tuberculosis in Southwestern Nigeria

Background: Nigeria is among the top five countries that have the highest gap between people reported as diagnosed and estimated to have developed tuberculosis (TB). To bridge this gap, there is a need for innovative approaches to identify geographical areas at high risk of TB transmission and targeted active case finding (ACF) interventions. Leveraging community-level data together with granular sociodemographic contextual information can unmask local hotspots that could be otherwise missed. This work evaluated whether this approach helps to reach communities with higher numbers of undiagnosed TB. Methodology: A retrospective analysis of the data generated from an ACF intervention program in four southwestern states in Nigeria was conducted. Wards (the smallest administrative level in Nigeria) were further subdivided into smaller population clusters. ACF sites and their respective TB screening outputs were mapped to these population clusters. This data were then combined with open-source high-resolution contextual data to train a Bayesian inference model. The model predicted TB positivity rates on the community level (population cluster level), and these were visualised on a customised geoportal for use by the local teams to identify communities at high risk of TB transmission and plan ACF interventions. The TB positivity yield (proportion) observed at model-predicted hotspots was compared with the yield obtained at other sites identified based on aggregated notification data. Results: The yield in population clusters that were predicted to have high TB positivity rates by the model was at least 1.75 times higher (p-value < 0.001) than the yield in other locations in all four states. Conclusions: The community-level Bayesian predictive model has the potential to guide ACF implementers to high-TB-positivity areas for finding undiagnosed TB in the communities, thus improving the efficiency of interventions.

Early Season Forecasting of Corn Yield at Field Level from Multi-Source Satellite Time Series Data

Crop yield forecasting during an ongoing season is crucial to ensure food security and commodity markets. For this reason, here, a scalable approach to forecast corn yields at the field-level using machine learning and satellite imagery from Sentinel-2 and Landsat missions is proposed. The model, evaluated on 1319 corn fields in the U.S. Corn Belt from 2017 to 2022, integrates biophysical parameters from Sentinel-2, Land Surface Temperature (LST) from Landsat, and agroclimatic data from ERA5 reanalysis dataset. Resampling the time series over thermal time significantly enhances predictive performance. The addition of LST to our model further improves in-season yield forecasting, through its capacity to detect early drought, which is not immediately visible to optical sensors such as the Sentinel-2. Finally, we propose a new two-stage machine learning strategy to mitigate early season partially available data. It consists in extending the current time series on the basis of complete historical data and adapting the model inference according to the crop progress.

Montelukast as a repurposable additive drug for standard-efficacy multiple sclerosis treatment: Emulating clinical trials with retrospective administrative health claims data

Background: Effective and safe treatment options for multiple sclerosis (MS) are still needed. Montelukast, a leukotriene receptor antagonist (LTRA) currently indicated for asthma or allergic rhinitis, may provide an additional therapeutic approach. Objective: The study aimed to evaluate the effects of montelukast on the relapses of people with MS (pwMS). Methods: In this retrospective case–control study, two independent longitudinal claims datasets were used to emulate randomized clinical trials (RCTs). We identified pwMS aged 18–65 years, on MS disease-modifying therapies concomitantly, in de-identified claims from Optum’s Clinformatics® Data Mart (CDM) and IQVIA PharMetrics® Plus for Academics. Cases included 483 pwMS on montelukast and with medication adherence in CDM and 208 in PharMetrics Plus for Academics. We randomly sampled controls from 35,330 pwMS without montelukast prescriptions in CDM and 10,128 in PharMetrics Plus for Academics. Relapses were measured over a 2-year period through inpatient hospitalization and corticosteroid claims. A doubly robust causal inference model estimated the effects of montelukast, adjusting for confounders and censored patients. Results: pwMS treated with montelukast demonstrated a statistically significant 23.6% reduction in relapses compared to non-users in 67.3% of emulated RCTs. Conclusion: Real-world evidence suggested that montelukast reduces MS relapses, warranting future clinical trials and further research on LTRAs’ potential mechanism in MS.

A knowledge-based intelligent recognition method for rock discontinuities with point cloud data

ABSTRACT Rock discontinuity information plays a key role in engineering activities. In severely fragmented and discontinuous samples, the identification accuracy of rock discontinuities can be improved with the help of geologic knowledge with identified spatial features and geoscience interaction mechanisms. Therefore, we proposed a new method – a knowledge-based intelligent recognition method for rock discontinuities that used point cloud data – to correct inaccurate identification by considering the spatial relationships of rock discontinuities. This paper integrated the geologic knowledge bases in rock discontinuity information extraction rules and combined the generative rule inference model with the region growing method for intelligent extraction. In the case study, we developed four typical rule sets as examples based on the constructed rule system to group rock discontinuities and subsume the fractured rock discontinuities. Results show that this new method is effective in identifying and classifying rock discontinuities of complex areas with a greater reduction in manual intervention.

Design and Evaluation of CPU-, GPU-, and FPGA-Based Deployment of a CNN for Motor Imagery Classification in Brain-Computer Interfaces

Brain–computer interfaces (BCIs) have gained popularity in recent years. Among noninvasive BCIs, EEG-based systems stand out as the primary approach, utilizing the motor imagery (MI) paradigm to discern movement intentions. Initially, BCIs were predominantly focused on nonembedded systems. However, there is now a growing momentum towards shifting computation to the edge, offering advantages such as enhanced privacy, reduced transmission bandwidth, and real-time responsiveness. Despite this trend, achieving the desired target remains a work in progress. To illustrate the feasibility of this shift and quantify the potential benefits, this paper presents a comparison of deploying a CNN for MI classification across different computing platforms, namely, CPU-, embedded GPU-, and FPGA-based. For our case study, we utilized data from 29 participants included in a dataset acquired using an EEG cap for training the models. The FPGA solution emerged as the most efficient in terms of the power consumption–inference time product. Specifically, it delivers an impressive reduction of up to 89% in power consumption compared to the CPU and 71% compared to the GPU and up to a 98% reduction in memory footprint for model inference, albeit at the cost of a 39% increase in inference time compared to the GPU. Both the embedded GPU and FPGA outperform the CPU in terms of inference time.

A decision support system for upper limb rehabilitation robot based on hybrid reasoning with RBR and CBR

The rehabilitation robot can assist hemiplegic patients to complete the training program effectively, but it only focuses on helping the patient’s training process and requires the rehabilitation therapists to manually adjust the training parameters according to the patient’s condition. Therefore, there is an urgent need for intelligent training prescription research of rehabilitation robots to promote the clinical applications. This study proposed a decision support system for the training of upper limb rehabilitation robot based on hybrid reasoning with rule-based reasoning (RBR) and case-based reasoning (CBR). The expert knowledge base of this system is established base on 10 professional rehabilitation therapists from three different rehabilitation departments in Shanghai who are enriched with experiences in using desktop-based upper limb rehabilitation robot. The rule-based reasoning is chosen to construct the cycle plan inference model, which develops a 21-day training plan for the patients. The case base consists of historical case data from 54 stroke patients who underwent rehabilitation training with a desktop-based upper limb rehabilitation robot. The case-based reasoning, combined with a Random Forest optimized algorithm, was constructed to adjust the training parameters for the patients in real-time. The system recommended a rehabilitation training program with an average accuracy of 91.5%, an average AUC value of 0.924, an average recall rate of 88.7%, and an average F1 score of 90.1%. The application of this system in rehabilitation robot would be useful for therapists.

EAD-Net: Efficiently Asymmetric Network for Semantic Labeling of High-Resolution Remote Sensing Images with Dynamic Routing Mechanism

Semantic labeling of high-resolution remote sensing images (HRRSIs) holds a significant position in the remote sensing domain. Although numerous deep-learning-based segmentation models have enhanced segmentation precision, their complexity leads to a significant increase in parameters and computational requirements. While ensuring segmentation accuracy, it is also crucial to improve segmentation speed. To address this issue, we propose an efficient asymmetric deep learning network for HRRSIs, referred to as EAD-Net. First, EAD-Net employs ResNet50 as the backbone without pooling, instead of the RepVGG block, to extract rich semantic features while reducing model complexity. Second, a dynamic routing module is proposed in EAD-Net to adjust routing based on the pixel occupancy of small-scale objects. Concurrently, a channel attention mechanism is used to preserve their features even with minimal occupancy. Third, a novel asymmetric decoder is introduced, which uses convolutional operations while discarding skip connections. This not only effectively reduces redundant features but also allows using low-level image features to enhance EAD-Net’s performance. Extensive experimental results on the ISPRS 2D semantic labeling challenge benchmark demonstrate that EAD-Net achieves state-of-the-art (SOTA) accuracy performance while reducing model complexity and inference time, while the mean Intersection over Union (mIoU) score reaching 87.38% and 93.10% in the Vaihingen and Potsdam datasets, respectively.

GAT TransPruning: progressive channel pruning strategy combining graph attention network and transformer.

Recently, large-scale artificial intelligence models with billions of parameters have achieved good results in experiments, but their practical deployment on edge computing platforms is often subject to many constraints because of their resource requirements. These models require powerful computing platforms with a high memory capacity to store and process the numerous parameters and activations, which makes it challenging to deploy these large-scale models directly. Therefore, model compression techniques are crucial role in making these models more practical and accessible. In this article, a progressive channel pruning strategy combining graph attention network and transformer, namely GAT TransPruning, is proposed, which uses the graph attention networks (GAT) and the attention of transformer mechanism to determine the channel-to-channel relationship in large networks. This approach ensures that the network maintains its critical functional connections and optimizes the trade-off between model size and performance. In this study, VGG-16, VGG-19, ResNet-18, ResNet-34, and ResNet-50 are used as large-scale network models with the CIFAR-10 and CIFAR-100 datasets for verification and quantitative analysis of the proposed progressive channel pruning strategy. The experimental results reveal that the accuracy rate only drops by 6.58% when the channel pruning rate is 89% for VGG-19/CIFAR-100. In addition, the lightweight model inference speed is 9.10 times faster than that of the original large model. In comparison with the traditional channel pruning schemes, the proposed progressive channel pruning strategy based on the GAT and Transformer cannot only cut out the insignificant weight channels and effectively reduce the model size, but also ensure that the performance drop rate of its lightweight model is still the smallest even under high pruning ratio.

Dynamic image super-resolution via progressive contrastive self-distillation

Convolutional neural networks (CNNs) are highly successful for image super-resolution (SR). However, they often require sophisticated architectures with high memory cost and computational overhead, significantly restricting their practical deployments on resource-limited devices. In this paper, we propose a novel dynamic contrastive self-distillation (Dynamic-CSD) framework to simultaneously compress and accelerate various off-the-shelf SR models, and explore using the trained model for dynamic inference. In particular, to build a compact student network, a channel-splitting super-resolution network (CSSR-Net) can first be constructed from a target teacher network. Then, we propose a novel contrastive loss to improve the quality of SR images via explicit knowledge transfer. Furthermore, progressive CSD (Pro-CSD) is developed to extend the two-branch CSSR-Net into multi-branch, leading to a switchable model at runtime. Finally, a difficulty-aware branch selection strategy for dynamic inference is given. Extensive experiments demonstrate that the proposed Dynamic-CSD scheme effectively compresses and accelerates several standard SR models such as EDSR, RCAN and CARN.

Imprecise probabilistic inference from sequential data.

Although the Bayesian paradigm is an important benchmark in studies of human inference, the extent to which it provides a useful framework to account for human behavior remains debated. We document systematic departures from Bayesian inference under correct beliefs, even on average, in the estimates by experimental subjects of the probability of a binary event following observations of successive realizations of the event. In particular, we find underreaction of subjects' estimates to the evidence ("conservatism") after only a few observations and at the same time overreaction after longer sequences of observations. This is not explained by an incorrect prior nor by many common models of Bayesian inference. We uncover the autocorrelation in estimates, which suggests that subjects carry imprecise representations of the decision situations, with noise in beliefs propagating over successive trials. But even taking into account these internal imprecisions and assuming various incorrect beliefs, we find that subjects' updates are inconsistent with the rules of Bayesian inference. We show how subjects instead considerably economize on the attention that they pay to the information relevant to the decision, and on the degree of control that they exert over their precise response, while giving responses fairly adapted to the task. A "noisy-counting" model of probability estimation reproduces the several patterns we exhibit in subjects' behavior. In sum, human subjects in our task perform reasonably well while greatly minimizing the amount of information that they pay attention to. Our results emphasize that investigating this economy of attention is crucial in understanding human decisions. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Conspiracy beliefs and perceptual inference in times of political uncertainty.

Sociopolitical crises causing uncertainty have accumulated in recent years, providing fertile ground for the emergence of conspiracy ideations. Computational models constitute valuable tools for understanding the mechanisms at play in the formation and rigidification of these unshakeable beliefs. Here, the Circular Inference model was used to capture associations between changes in perceptual inference and the dynamics of conspiracy ideations in times of uncertainty. A bistable perception task and conspiracy belief assessment focused on major sociopolitical events were administered to large populations from three polarized countries. We show that when uncertainty peaks, an overweighting of sensory information is associated with conspiracy ideations. Progressively, this exploration strategy gives way to an exploitation strategy in which increased adherence to conspiracy theories is associated with the amplification of prior information. Overall, the Circular Inference model sheds new light on the possible mechanisms underlying the progressive strengthening of conspiracy theories when individuals face highly uncertain situations.