Belief Update Research Articles

Unsupervised Active Visual Search With Monte Carlo Planning Under Uncertain Detections.

We propose a solution for Active Visual Search of objects in an environment, whose 2D floor map is the only known information. Our solution has three key features that make it more plausible and robust to detector failures compared to state-of-the-art methods: i) it is unsupervised as it does not need any training sessions. ii) During the exploration, a probability distribution on the 2D floor map is updated according to an intuitive mechanism, while an improved belief update increases the effectiveness of the agent's exploration. iii) We incorporate the awareness that an object detector may fail into the aforementioned probability modelling by exploiting the success statistics of a specific detector. Our solution is dubbed POMP-BE-PD (Pomcp-based Online Motion Planning with Belief by Exploration and Probabilistic Detection). It uses the current pose of an agent and an RGB-D observation to learn an optimal search policy, exploiting a POMDP solved by a Monte-Carlo planning approach. On the Active Vision Dataset Benchmark, we increase the average success rate over all the environments by a significant 35 % while decreasing the average path length by 4 % with respect to competing methods. Thus, our results are state-of-the-art, even without any training procedure.

Confidence regulates feedback processing during human probabilistic learning.

Uncertainty presents a key challenge when learning how best to act to attain a desired outcome. People can report uncertainty in the form of confidence judgments, but how such judgments contribute to learning and subsequent decisions remains unclear. In a series of three experiments employing an operant learning task, we tested the hypothesis that confidence plays a central role in learning by regulating resource allocation to the seeking and processing of feedback. We predicted that, as participants' confidence in their task knowledge grew, they would discount feedback when it was provided and be correspondingly less willing to pay for it when it was costly. Consistent with these predictions, we found that higher confidence was associated with reduced electrophysiological markers of feedback processing and decreased updating of beliefs following feedback receipt. Bayesian modeling suggests that this decrease in processing was due to a drop in the expected informative value of novel information when participants were highly confident. Thus, when choosing whether to pay a fee to receive further feedback, participants' subjective confidence, rather than the objective accuracy of their decisions, guided their choices. Overall, our results suggest that confidence regulates learning and subsequent decision making. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

People update their injunctive norm and moral beliefs after receiving descriptive norm information.

How do descriptive norms shape injunctive norm beliefs, and what does this tell us about the cognitive processes underlying social norm cognition? Across six studies (N = 2,671), we examined whether people update their injunctive norm beliefs-as well as their moral judgments and behavioral intentions-after receiving descriptive norm information about how common (or uncommon) a behavior is. Specifically, we manipulated the descriptive normativity of behaviors, describing behaviors as uncommon (20% of people were doing the behavior) or common (80% of people were doing the behavior), and the type of behavior across studies (fairness, conventional, harm, preference). To measure belief updating, we assessed beliefs prior to and after receiving information about the descriptive norm. We had three main findings: First, participants positively updated their prior injunctive norm beliefs, moral judgments, and behavioral intentions (i.e., rated behaviors more injunctively normative and moral) after receiving a common descriptive norm and negatively updated their beliefs (i.e., rated behaviors less injunctive and moral) after receiving an uncommon descriptive norm, and updated to a larger extent for the common than uncommon descriptive norm. Second, participants were more likely to update their beliefs about what is moral for others compared to what is moral for the self. Third, participants updated their beliefs to a greater extent for fairness and conventional behaviors compared to harm behaviors and preferences. Together, our findings suggest that descriptive norms shape our injunctive norm beliefs and moral judgments and help to paint a fuller picture of the social cognition of social norms. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Signaling Quality: How Refund Bonuses Can Overcome Information Asymmetries in Crowdfunding

Crowdfunding can suffer from information asymmetry, leaving some investors disappointed with low-quality projects, whereas other high-quality projects remain unfunded. We show that refund bonuses, which provide investors a payment if a fundraising campaign is unsuccessful, can signal project quality and help overcome the market failure in crowdfunding. Because strong projects have a lower risk of bonus payout, entrepreneurs with strong projects are more likely to offer bonuses. This signals high quality to investors, and due to their updated beliefs, this drives investment toward such projects. An experiment provides supporting empirical evidence for the benefits of this signaling solution to the problems of information asymmetry in crowdfunding. This paper was accepted by Joshua Gans, business strategy. Funding: This work was supported by Schmidt Sciences. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2023.03923 .

The Effect of Correcting Neuromyths on Students' and Teachers' Later Reasoning.

Students and educators sometimes hold beliefs about intelligence and learning that lack scientific support, often called neuromyths. Neuromyths can be problematic, so it is important to find methods to correct them. Previous findings demonstrate that textual refutations are effective for correcting neuromyths. However, even after correction, erroneous information may continue to influence reasoning. In three experiments, we investigated whether feedback could be used to update students' and educators' beliefs and influence their reasoning about neuromyths. Across all experiments, the results showed that both students and educators held erroneous beliefs about learning and memory that could be updated after receiving feedback. Feedback also increased students', but not teachers', reasoning accuracy. The results demonstrate that feedback can be used to update beliefs in neuromyths, but these beliefs may influence reasoning even after correction.

Impact of risk information on health beliefs and behaviours: experimental evidence

ABSTRACT The local prevalence of infections and the severity of their consequences are among the key determinants of the adoption of preventive behaviours against infectious diseases. In Bangladesh, where local COVID-19 infection statistics were not readily accessible, I found that most people underestimated the local prevalence of COVID-19 infections while overestimated the fatality rate. In a randomized experiment, the treatment group was provided with information about the coronavirus case numbers in their districts, along with the case fatality rate in Bangladesh and globally. Immediately after receiving this information, the treatment group perceived a higher infection risk. Nine to fifteen days post-intervention, those who received information were less likely to underestimate the local prevalence and, consequently, still perceived a higher infection risk. They also updated their belief about the fatality rate downward. Potentially due to this countervailing update of risk beliefs, the information had no effect on self-reported preventive behaviours.

Individual differences in belief updating and phasic arousal are related to psychosis proneness

Many decisions entail the updating of beliefs about the state of the environment by accumulating noisy sensory evidence. This form of probabilistic reasoning may go awry in psychosis. Computational theory shows that optimal belief updating in environments subject to hidden changes in their state requires a dynamic modulation of the evidence accumulation process. Recent empirical findings implicate transient responses of pupil-linked central arousal systems to individual evidence samples in this modulation. Here, we analyzed behavior and pupil responses during evidence accumulation in a changing environment in a community sample of human participants. We also assessed their subclinical psychotic experiences (psychosis proneness). Participants most prone to psychosis showed overall less flexible belief updating profiles, with diminished behavioral impact of evidence samples occurring late during decision formation. These same individuals also exhibited overall smaller pupil responses and less reliable pupil encoding of computational variables governing the dynamic belief updating. Our findings provide insights into the cognitive and physiological bases of psychosis proneness and open paths to unraveling the pathophysiology of psychotic disorders.

Cloud security in the age of adaptive adversaries: A game theoretic approach to hypervisor-based intrusion detection

Recent advancements in cloud computing have underscored the critical need for robust security mechanisms to counter evolving cyber-threats. Traditional security solutions such as Intrusion Detection Systems (IDSs) often fall short due to their inability to anticipate the strategies of adaptive cyber adversaries. Game theory is considered a popular analytical tool for understanding the strategic interactions between defenders and adversaries, providing a more informed decision-making process. However, existing game-theoretic IDSs often employ non-comprehensive utility functions with limited parameters that fail to capture the complexity of real-world dynamics. This paper introduces a novel Game-Theoretic Hypervisor-based IDS (GHyIDS), which employs comprehensive utility functions and an innovative belief update model to enhance detection accuracy and adaptability in dynamic cloud environments. To overcome the limitations of existing models, we design comprehensive utility functions by incorporating a wider range of real-world parameters, such as trust score, risk, vulnerability, damage severity, worth of the VM, means, opportunities, and access available to the attacker, as well as success rates of attack detection and execution. We propose a Resource-Aware Static Intrusion Detection Bayesian Game (S-IDBG) and extend it into a Dynamic Multi-Stage IDBG (D-IDBG), enabling the system to dynamically adapt to changes in attack patterns and system vulnerabilities. The belief update model is pivotal in continuously refining the system’s strategies based on observed behaviors and outcomes, allowing for precise adjustments to the evolving threats. Our experimental results show a significant improvement over existing models, with our approach achieving approximately 10% increase in detection rate, 20% reduction in false positive rate and 10% reduction in false negative rate in comparative analysis against state-of-the-art models namely, the trust-based Maxmin game and the repeated Bayesian Stackelberg game.

Nucleus accumbens dopamine release reflects Bayesian inference during instrumental learning.

Dopamine release in the nucleus accumbens has been hypothesized to signal reward prediction error, the difference between observed and predicted reward, suggesting a biological implementation for reinforcement learning. Rigorous tests of this hypothesis require assumptions about how the brain maps sensory signals to reward predictions, yet this mapping is still poorly understood. In particular, the mapping is non-trivial when sensory signals provide ambiguous information about the hidden state of the environment. Previous work using classical conditioning tasks has suggested that reward predictions are generated conditional on probabilistic beliefs about the hidden state, such that dopamine implicitly reflects these beliefs. Here we test this hypothesis in the context of an instrumental task (a two-armed bandit), where the hidden state switches repeatedly. We measured choice behavior and recorded dLight signals reflecting dopamine release in the nucleus accumbens core. Model comparison among a wide set of cognitive models based on the behavioral data favored models that used Bayesian updating of probabilistic beliefs. These same models also quantitatively matched the dopamine measurements better than non-Bayesian alternatives. We conclude that probabilistic belief computation contributes to instrumental task performance in mice and is reflected in mesolimbic dopamine signaling.

The adaptive consensus reaching process with the ER-PSO interaction mechanism and its application in public charging stations evaluation

BackgroundPublic charging stations (PCSs) have become integral public infrastructure in China, responding to the considerable and rapidly growing demand. Evaluating the service quality of PCSs by a diverse user base has become essential for the effective management of PCSs. ObjectivesIn the group evaluation of PCSs' service quality, which encompasses broad public interests, consensus reaching process (CRP) is commonly utilized to manage group conflicts and achieve evaluation consensus. The intricate interactions among decision makers (DMs) significantly influence the evolution and resilience of collective viewpoints. Against this backdrop, this research aims to develop an adaptive CRP to enhance the discriminatory power of group consensus and improve consensus robustness in the presence of non-cooperative behavior. MethodologyWith the above objectives, this study proposes an adaptive CRP supported by a novel interaction mechanism, consisting of two stages: (1) In the interaction network generation stage, based on the Erdős-Rényi (ER) random network model, innovative interaction signals are generated to guide the generation of an ER-based interaction network, enhancing the dissemination of dominant viewpoints, and constraining the influence of minority opinions. (2) In the decision opinion interaction stage, based on the Particle Swarm Optimization (PSO) algorithm, utilizing novel dual individual fitness functions, the PSO algorithm is employed to formulate the expected updated belief, driving the internal evolution and convergence of opinions. The proposed adaptive CRP is implemented in the group evaluation of PCSs' service quality in Nanjing. FindingsComparative analysis reveals that the proposed CRP offers significant advantages in consensus efficiency, the discriminatory power of group consensus, and consensus robustness in the presence of non-cooperative behavior compared to existing CRPs. NoveltyDiffering from previous CRP research primarily focuses on minimizing adjustment scales, this study emphasizes the adaptive evolution and convergence of opinions, providing a more suitable description of group opinion dynamics. Additionally, from a methodological perspective, this research utilizes the ER network model and the PSO algorithm to support the adaptive CRP at both structural and preference levels, enhancing consensus discriminatory power and robustness.

Tracking politically motivated reasoning in the brain: the role of mentalizing, value-encoding, and error-detection networks.

Susceptibility to misinformation and belief polarization often reflect people's tendency to incorporate information in a biased way. Despite the presence of competing theoretical models, the underlying neurocognitive mechanisms of motivated reasoning remain elusive as previous empirical work did not properly track the belief formation process. To address this problem, we employed a design that identifies motivated reasoning as directional deviations from a Bayesian benchmark of unbiased belief updating. We asked members of a pro-immigration or an anti-immigration group how much they endorse factual messages on foreign criminality, a polarizing political topic. Both groups exhibited a desirability bias by over-endorsing attitude-consistent messages and under-endorsing attitude-discrepant messages and an identity bias by over-endorsing messages from ingroup members and under-endorsing messages from outgroup members. In both groups, neural responses to the messages predicted subsequent expression of desirability and identity biases suggesting a common neural basis of motivated reasoning across ideologically opposing groups. Specifically, brain regions implicated in encoding value, error detection, and mentalizing tracked the degree of desirability bias. Less extensive activation in the mentalizing network tracked the degree of identity bias. These findings illustrate the distinct neurocognitive architecture of desirability and identity biases and inform existing cognitive models of politically motivated reasoning.

Dynamic Opinion Formation in Networks: A Multi-Issue and Evidence-Based Approach

In this study, we present a computational model for simulating opinion dynamics within social networks, incorporating cognitive and social psychological principles such as homophily, confirmation bias, and selective exposure. We enhance our model using Dempster–Shafer theory to address uncertainties in belief updating. Mathematical formalism and simulations were performed to derive empirical results from showcasing how this method might be useful for modeling real-world opinion consensus and fragmentation. By constructing a scale-free network, we assign initial opinions and iteratively update them based on neighbor influences and belief masses. Lastly, we examine how the presence of “truth” nodes with high connectivity, used to simulate the influence of objective truth on the network, alters opinions. Our simulations reveal insights into the formation of opinion clusters, the role of cognitive biases, and the impact of uncertainty on belief evolution, providing a robust framework for understanding complex opinion dynamics in social systems.

Plinko: Eliciting beliefs to build better models of statistical learning and mental model updating.

Prior beliefs are central to Bayesian accounts of cognition, but many of these accounts do not directly measure priors. More specifically, initial states of belief heavily influence how new information is assumed to be utilized when updating a particular model. Despite this, prior and posterior beliefs are either inferred from sequential participant actions or elicited through impoverished means. We had participants to play a version of the game 'Plinko', to first elicit individual participant priors in a theoretically agnostic manner. Subsequent learning and updating of participant beliefs was then directly measured. We show that participants hold various priors that cluster around prototypical probability distributions that in turn influence learning. In follow-up studies, we show that participant priors are stable over time and that the ability to update beliefs is influenced by a simple environmental manipulation (i.e., a short break). These data reveal the importance of directly measuring participant beliefs rather than assuming or inferring them as has been widely done in the literature to date. The Plinko game provides a flexible and fecund means for examining statistical learning and mental model updating.

The Bright Side of Price Volatility in Global Commodity Procurement

This paper studies two competing firms’ choices between the contingent-price contract (CPC) and fixed-price contract (FPC) in global commodity procurement. The FPC price is determined when signing the contract, whereas the CPC price is pegged to an underlying index and remains open until the delivery date. Under both contracts, each firm determines its order quantity based on the updated belief about the market demand. The unrealized CPC price correlates with the market demand, allowing a firm to update its belief about the CPC price using demand information, thereby generating a price-learning effect. We find that, contrary to conventional wisdom, a larger price volatility could benefit the firms, and, under differentiated contracts, a firm might benefit from the improvement of forecast accuracy at its rival. We further show that the price-learning effect plays a critical role in the firms’ contract choices. First, significant price volatility forces the firms to pursue the responsiveness of the CPC. Second, the firms may adopt differentiated contracts to enhance their responses to market changes and dampen competition, and a higher competition intensity more likely leads to contract differentiation. Third, the firms in a small market seek responsiveness and contract differentiation rather than cost efficiency. This study reveals the bright side of price volatility and takes a step toward understanding the effect of two-dimensional information updating. This paper was accepted by Jeannette Song, operations management. Funding: This work was supported in part by the National Natural Science Foundation of China [W. Xing was supported by Grant 72071210, L. Liu was supported by Grant 72371122, and F. Zhang was supported by Grants 71929201 and 72131004]. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2023.00304 .

Disaster learning and aggregate investment

We extend a production-based asset pricing model by introducing learning about disaster risk. The information is not perfect, and Bayesian learning is adopted to update beliefs about the likelihood of rare disasters. We show that disaster learning reconciles key stylized facts about macroeconomic quantities and financial markets. For macroeconomic quantities, during the crisis, the decline in aggregate investments is much worse than that of output, whereas the decline in aggregate consumption is moderate relative to that of output. Additionally, the model endogenously features lower consumption volatility and higher investment volatility than that of output. For financial markets, belief updating over a rare disaster produces a higher equity premium, lower risk-free rate, and more volatile stock returns. Finally, we show that jump intensity uncertainty accounts for a substantial fraction of the total welfare cost of rare disasters.

Updating Prospective Self-Efficacy Beliefs About Cardiac Interoception in Anorexia Nervosa: An Experimental and Computational Study.

Patients with anorexia nervosa (AN) typically hold altered beliefs about their body that they struggle to update, including global, prospective beliefs about their ability to know and regulate their body and particularly their interoceptive states. While clinical questionnaire studies have provided ample evidence on the role of such beliefs in the onset, maintenance, and treatment of AN, psychophysical studies have typically focused on perceptual and 'local' beliefs. Across two experiments, we examined how women at the acute AN (N = 86) and post-acute AN state (N = 87), compared to matched healthy controls (N = 180) formed and updated their self-efficacy beliefs retrospectively (Experiment 1) and prospectively (Experiment 2) about their heartbeat counting abilities in an adapted heartbeat counting task. As preregistered, while AN patients did not differ from controls in interoceptive accuracy per se, they hold and maintain 'pessimistic' interoceptive, metacognitive self-efficacy beliefs after performance. Modelling using a simplified computational Bayesian learning framework showed that neither local evidence from performance, nor retrospective beliefs following that performance (that themselves were suboptimally updated) seem to be sufficient to counter and update pessimistic, self-efficacy beliefs in AN. AN patients showed lower learning rates than controls, revealing a tendency to base their posterior beliefs more on prior beliefs rather than prediction errors in both retrospective and prospective belief updating. Further explorations showed that while these differences in both explicit beliefs, and the latent mechanisms of belief updating, were not explained by general cognitive flexibility differences, they were explained by negative mood comorbidity, even after the acute stage of illness.

Frequentist belief update under ambiguous evidence in social networks

In this paper, we study a frequentist approach to belief updating in the framework of Dempster-Shafer Theory (DST). We propose several mechanisms that allow the gathering of possibly ambiguous pieces of evidence over time to obtain a belief mass assignment. We then use our approach to study the impact of ambiguous evidence on the belief distribution of agents in social networks. We illustrate our approach by taking three representative situations. In the first one, we suppose that there is an unknown state of nature, and agents form belief in the set of possible states. Nature constantly sends a signal which reflects the true state with some probability but which can also be ambiguous. In the second situation, there is no ground truth, and agents are against or in favor of some ethical or societal issues. In the third situation, there is no ground state either, but agents have opinions on left, center, and right political parties. We show that our approach can model various phenomena often observed in social networks, such as polarization or bounded confidence effects.

Multi-sensor multi-target tracking with generalized labeled multi-Bernoulli filter based on track-before-detect observation model using Gaussian belief propagation

Tracking an unknown number of low-observable targets is notoriously challenging problem. This paper extends the multi-sensor multi-target tracking with generalized labeled multi-Bernoulli (GLMB) filter into the track-before-detect (TBD) observation model and presents its efficient implementation. The Gaussian belief propagation (BP) is applied into multi-sensor GLMB-TBD filter to address the computational bottleneck of computing GLMB filtering density for tracking multiple targets in challenging scenario. The Gaussian BP algorithm is used to directly compute marginal densities from the joint posterior density through the BP belief update wherein the GLMB filtering density at given time serves as prior information for next recursion. Compared to the Gibbs sampling-version implementation, the Gaussian BP-version implementation avoids the pruning of GLMB components without association information loss and the relevant association is preserved for BP belief update. Hence, the proposed Gaussian BP-version implementation can improve the tracking accuracy. Moreover, the computational complexity of our proposed Gaussian BP-version implementation scales linear complexity in the total number of measurements from all sensors. More importantly, the linear scaling in number of Bernoulli components for our proposed Gaussian BP-version implementation improves on the quadratic complexity exhibited by the Gibbs-version one. The experiment results show that the computational cost of the Gaussian BP-version implementation is significantly less than that of the Gibbs sampling-version implementation for the multi-sensor GLMB-TBD filter.

Information Transparency With Targeting Technology for Online Service Operations Platform

Social technologies have enabled the emergence of online platforms that provide offline service consultations and recommendations. In this environment, economic inefficiency arises when customers are not fully aware of their horizontally differentiated preferences. With its expertise or data dominance, a platform can be more informed about customers’ hidden preferences. We focus on an instrumental social technology, that is, targeting, which is a type of data-driven personalized information provision to manipulate customers’ beliefs about service quality. We propose a Hotelling model wherein customers are sensitive to the delays for service while making Bayesian belief updates based on a platform’s recommendations. When customers self-select their favorite service, their choices impose negative externalities through congestion and welfare loss. Our results indicate that service recommendations allow customers to navigate toward the more appropriate service, thus improving matching efficiency, reducing congestion costs, and enhancing aggregate customer welfare. We further identify the role of “information transparency” and study how the platform should strategically release information by making personalized service recommendations to customers. Interestingly, when a customer-centric platform maximizes aggregate customer welfare, we identify the “value of opaqueness” by strategically withholding service recommendations from a subset of customers and notice that this effect is more pronounced for a profit-seeking platform. Our results offer a better understanding of information transparency policies in the joint design of service recommendation systems and pricing mechanisms.

Distributed multi-robot source term estimation with coverage control and information theoretic based coordination

In this paper, we introduce a novel coordination strategy for a group of autonomous robots tasked with estimating the source term of an airborne chemical release. This strategy integrates distributed Bayesian filtering, coverage control, information-theoretic sampling, and proximity constraint handling, forming an efficient and fully distributed coordination protocol. In the proposed framework, each robot employs a consensus-based belief update rule, allowing it to adaptively incorporate information from neighbouring robots to ensure a unified belief across the network. The overall control action is designed to maximise information gain while maintaining network connectivity and minimising communication requirements during movement between sampling points. Extensive numerical simulations are conducted to analyse the performance of the proposed strategy, which demonstrate significant performance improvements compared to popular filtering practices and advanced path-planning strategies. The simulation study is also designed to substantiate the design choices of the proposed coordination strategy and to emphasise its advantages.