Belief Update Research Articles

Updating false beliefs: The role of misplaced vs. well-placed certainty.

People can update their misconceptions or false beliefs by learning from corrective sources. However, research has shown that people vary drastically in the extent to which they learn from feedback and update their false beliefs accordingly. That past work drew attention to cognitive and motivational factors such as cognitive rigidity and closed-mindedness as inhibitors of belief updating. Here we examined a novel epistemic structure, misplaced certainty, a subjective sense of certainty while recognizing uncertainty in oneself or most people (e.g., I feel certain although I recognize X is technically uncertain or it is technically uncertain according to most people), as a unique predictor of lower belief updating. In a preregistered study, we hypothesized that those with high chronic misplaced certainty would be less likely to learn from feedback and revise their misconceptions in a feedback-learning task. In our analyses, we controlled for well-placed certainty-certainty while recognizing no doubt in oneself or most others. We also controlled for variables associated with closed-minded cognition. Consistent with our predictions, those with high misplaced certainty were less likely to revise their false beliefs in response to corrective feedback. In contrast, those with high well-placed certainty were more likely to learn from corrective feedback and revise their false beliefs. By shedding light on the nuances of different forms of subjective certainty, the present work aims to pave the way for further research on epistemic factors in the perseverance and correction of false beliefs.

Resolving ambiguity as a public good: experimental evidence from Guyana

Incomplete information is a commonly cited barrier to the adoption of new innovations. We present a decision-making experiment, conducted with farmers in the field, that explores the extent to which information which reduces ambiguity may be provided as a public good. In the experiment, participants make a series of decisions between a risky gamble and an ambiguous gamble. An initial private decision is followed by second choice in which participants know that their chosen gambles and outcomes will be publicly but anonymously revealed. Selection of the ambiguous option in this decision thus provide public good information which can be used to update beliefs about the ambiguous gamble. After the public revelation, participants make one final choice between the two options which we use to assess learning. We also test communication as a mechanism to increase provision of the public good. We find evidence that people with preference to avoid ambiguity contribute to public good information while risk averse people free ride. Communication does not affect the overall public good provision, either in a positive or negative direction. Finally, we find that people draw appropriate inference from the information that the public good provides. We relate our findings to the issue of new technology adoption in agriculture which is of particular importance for our participants.

Translating affective and cognitive mechanisms of psychotic symptoms into better therapies

For a long time, psychotic symptoms were seen as too different from normal experiences to be explicable by psychological theories, let alone modifiable by psychological therapy. In the 1990s, the epidemiological research pointing to a continuum of psychotic symptoms and the clinical success in using cognitive approaches for psychosis strongly challenged this view. This inspired the field – along with my own work – to focus on a better understanding of the psychological mechanisms of psychotic symptoms in order to develop increasingly effective interventions. Together with other researchers, I have found that negative affect and aberrant arousal play a crucial role in the formation of delusions where they appear to interact with more fundamental problems in belief updating. While our understanding of the basic mechanisms that drive psychotic symptoms is gaining precision, the challenge remains to find the most suitable ways of targeting these mechanisms.

Reinforcement Learning Based Underwater Wireless Optical Communication Alignment for Autonomous Underwater Vehicles

With the developments in underwater wireless optical communication (UWOC) technology, UWOC can be used in conjunction with autonomous underwater vehicles (AUVs) for high-speed data sharing among the vehicle formation during underwater exploration. A beam alignment problem arises during communication due to the transmission range, external disturbances and noise, and uncertainties in the AUV dynamic model. In this article, we propose an acoustic navigation method to guide the alignment process without requiring beam directors, light intensity sensors, and/or scanning algorithms as used in previous research. The AUVs need stably maintain a specific relative position and orientation for establishing an optical link. We model the alignment problem as a partially observable Markov decision process (POMDP) that takes manipulation, navigation, and energy consumption of underwater vehicles into account. However, finding an efficient policy for the POMDP under high partial observability and environmental variability is challenging. Therefore, for successful policy optimization, we utilize the soft actor–critic reinforcement learning algorithm together with AUV-specific belief updates and reward shaping based curriculum learning. Our approach outperformed baseline approaches in a simulation environment and successfully performed the beam alignment process from one AUV to another on the real AUV Tri-TON 2.

Evaluation of Early Ketamine Effects on Belief-Updating Biases in Patients With Treatment-Resistant Depression

Clinical research has shown that persistent negative beliefs maintain depression and that subanesthetic ketamine infusions induce rapid antidepressant responses. To evaluate whether ketamine alters belief updating and how such cognitive effects are associated with the clinical effects of ketamine. This study used an observational case-control protocol with a mixed-effects design that nested 2 groups by 2 testing time points. Observers were not blinded. Patients with treatment-resistant depression (TRD) and healthy volunteer participants aged 34 to 68 years were included. Patients with TRD were diagnosed with major depressive disorder or bipolar depression, had a Montgomery-Åsberg Depression Rating Scale score greater than 20, a Maudsley Staging Method score greater than 7, and failed to respond to at least 2 prior antidepressant trials. Exclusion criteria were any other psychiatric, neurological, or neurosurgical comorbidities, substance use or addictive disorders, and recreational ketamine consumption. Data were collected from January to February 2019 and from May to December 2019, and data were analyzed from January 2020 to July 2021. Patients with TRD were observed 24 hours before single ketamine infusion, 4 hours after the infusion, and 4 hours after the third infusion, which was 1 week after the first infusion. Healthy control participants were observed twice 1 week apart without ketamine exposure. Montgomery-Åsberg Depression Rating Scale score and belief updating after belief updating when patients received good news and bad news measured by a cognitive belief-updating task and mathematically formalized by a computational reinforcement learning model. Of 56 included participants, 29 (52%) were male, and the mean (SEM) age was 52.3 (1.2) years. A total of 26 patients with TRD and 30 control participants were included. A significant group × testing time point × news valence interaction showed that patients with TRD updated their beliefs more after good than bad news following a single ketamine infusion (controlled for age and education: β = -0.91; 95% CI, -1.58 to -0.24; t216 = -2.67; P = .008) than controls. Computational modeling showed that this effect was associated with asymmetrical learning rates (LRs) after ketamine treatment (good news LRs after ketamine, 0.51 [SEM, 0.04]; bad news LRs after ketamine 0.36 [SEM, 0.03], t25 = 3.8; P < .001) and partially mediated early antidepressant responses (path a*b: β = -1.00 [SEM, 0.66]; t26 = -1.53; z = -1.98; P = .04). These findings provide novel insights into the cognitive mechanisms of the action of ketamine in patients with TRD, with promising perspectives for augmented psychotherapy for individuals with mood disorders.

A Pragmatic Account of the Weak Evidence Effect

Language is not only used to transmit neutral information; we often seek to persuade by arguing in favor of a particular view. Persuasion raises a number of challenges for classical accounts of belief updating, as information cannot be taken at face value. How should listeners account for a speaker's "hidden agenda" when incorporating new information? Here, we extend recent probabilistic models of recursive social reasoning to allow for persuasive goals and show that our model provides a pragmatic account for why weakly favorable arguments may backfire, a phenomenon known as the weak evidence effect. Critically, this model predicts a systematic relationship between belief updates and expectations about the information source: weak evidence should only backfire when speakers are expected to act under persuasive goals and prefer the strongest evidence. We introduce a simple experimental paradigm called the Stick Contest to measure the extent to which the weak evidence effect depends on speaker expectations, and show that a pragmatic listener model accounts for the empirical data better than alternative models. Our findings suggest further avenues for rational models of social reasoning to illuminate classical decision-making phenomena.

Interactions between the cortical midline structures and sensorimotor network track maladaptive self-beliefs in clinical high risk for psychosis

Individuals at clinical high risk for psychosis (CHR) report a maladaptive self-concept—with more negative and less positive self-beliefs—linked to clinical symptoms and functional impairment. Alterations have also been reported in brain networks associated with intrinsic (cortical midline structures, CMS) and extrinsic (sensorimotor network, SMN) self-processing. Theoretical accounts of multiple levels of self-experience in schizophrenia suggest that interactions between these networks would be relevant for self-beliefs. This study tested whether self-beliefs related to resting-state functional connectivity within and between the CMS and SMN. Participants were 56 individuals meeting CHR criteria and 59 matched healthy community participants (HC). Pearson correlations examined potential mediators and outcomes. The CHR group reported more negative and less positive self-beliefs. Greater resting-state functional connectivity between the posterior CMS (posterior cingulate cortex) and the SMN was associated with less positive self-beliefs in CHR, but more positive self-beliefs in HC. Attenuated negative symptoms and poorer social functioning were associated with CMS-SMN connectivity (trend level after FDR-correction) and self-beliefs. Reduced connectivity between the left and right PCC was associated with lower positive self-beliefs in CHR, although this effect was specific to very low levels of positive self-beliefs. Left-right PCC connectivity did not correlate with outcomes. Dynamic interactions between intrinsic and extrinsic self-processing supported positive self-beliefs in typically developing youth while undermining positive self-beliefs in CHR youth. Implications are discussed for basic self-fragmentation, narrative self-related metacognition, and global belief updating. Interventions for self-processing may be beneficial in the CHR syndrome.

Inattentive Inference

AbstractThis paper studies how people infer a state of the world from information structures that include additional, payoff-irrelevant states. For example, learning from a customer review about a product’s quality requires accounting for the reviewer’s otherwise-irrelevant taste. This creates an attribution problem common to all information structures with multiple causes. We report controlled experimental evidence for pervasive overinference about states that affect utility—a form of “omitted variable bias” in belief updating, providing an explanation for various misattribution patterns. In studying why systematic misattribution arises, we consistently find that errors are not due to deliberate effort avoidance or a lack of cognitive capacity. Instead, people behave as if they form incomplete mental models of the information structure and fail to notice the need to account for alternative causes. These mental models are not stable but context-dependent: Misattribution responds to a variety of attentional manipulations, but not to changes in the costs of inattention.

Decentralized inertial best-response with voluntary and limited communication in random communication networks

Multiple autonomous agents interact over a random communication network to maximize their individual utility functions which depend on the actions of other agents. We consider decentralized best-response with inertia type algorithms in which agents form beliefs about the future actions of other players based on local information, and take actions that maximize their expected utilities computed with respect to these beliefs or continue to take their previous actions. We show convergence of these types of algorithms to a Nash equilibrium in weakly acyclic games. The result depends on the condition that the belief update and information exchange protocols successfully learn the actions of other players with positive probability in finite time given a static environment, i.e., when other agents’ actions do not change. We design a decentralized fictitious play algorithm with voluntary and limited communication (DFP-VL) protocols that satisfy this condition. In the voluntary communication protocol, each agent decides whom to exchange information with by assessing the novelty of its information and the potential effect of its information on others’ beliefs. The limited communication protocol entails agents sending only their most frequent action to agents that they decide to communicate with. Numerical experiments on a target assignment game demonstrate that the voluntary and limited communication protocol can more than halve the number of communication attempts while retaining the same convergence rate as DFP in which agents constantly attempt to communicate.

Communication-Efficient Distributed Cooperative Learning With Compressed Beliefs

In this article, we study the problem of distributed cooperative learning, where a group of agents seeks to agree on a set of hypotheses that best describes a sequence of private observations. In the scenario where the set of hypotheses is large, we propose a belief update rule where agents share compressed (either sparse or quantized) beliefs with an arbitrary positive compression rate. Our algorithm leverages a unified communication rule that enables agents to access wide-ranging compression operators as black-box modules. We prove the almost sure asymptotic convergence of beliefs on the set of optimal hypotheses. Additionally, we show a nonasymptotic, explicit, and linear concentration rate in probability of the beliefs on the optimal hypothesis set. We provide numerical experiments to illustrate the communication benefits of our method. The simulation results show that the number of transmitted bits can be reduced to 5%–10% of the noncompressed method in the studied scenarios.

Biased belief priors versus biased belief updating: Differential correlates of depression and anxiety.

Individuals prone to anxiety and depression often report beliefs and make judgements about themselves that are more negative than those reported by others. We use computational modeling of a richly naturalistic task to disentangle the role of negative priors versus negatively biased belief updating and to investigate their association with different dimensions of Internalizing psychopathology. Undergraduate participants first provided profiles for a hypothetical tech internship. They then viewed pairs of other profiles and selected the individual they would prefer to work alongside out of each pair. In a subsequent phase of the experiment, participants made judgments about their relative popularity as hypothetical internship partners both before any feedback and after each of 20 items of feedback revealing whether or not they had been selected as the preferred teammate from a given pairing. Scores on latent factors of general negative affect, anxiety-specific affect and depression-specific affect were estimated using participants’ self-report scores on standardized measures of anxiety and depression together with factor loadings from a bifactor analysis conducted previously. Higher scores on the depression-specific factor were linked to more negative prior beliefs but were not associated with differences in belief updating. In contrast, higher scores on the anxiety-specific factor were associated with a negative bias in belief updating but no difference in prior beliefs. These findings indicate that, to at least some extent, distinct processes may impact the formation of belief priors and in-the-moment belief updating and that these processes may be differentially disrupted in depression and anxiety. Future directions for enquiry include examination of the possibility that prior beliefs biases in depression might reflect generalization from prior experiences or global schema whereas belief updating biases in anxiety might be more situationally specific.

Peering into a crystal ball: Forecasting behavior and industry foresight

AbstractResearch SummaryWhat makes some managers and entrepreneurs better at forecasting the industry context than others? We argue that, regardless of experience or expertise, a learning‐based forecasting behavior in which individuals attend to and incorporate new relevant information from the environment into an updated belief that aligns with the Bayesian belief updating process is likely to generate superior industry foresight. However, the effectiveness of such a cognitively demanding process diminishes under high levels of uncertainty. We find support for these arguments using an experimental design of forecasting tournaments in the managerially relevant context of the global automotive industry from 2016 to 2019. The study provides a novel account of individual‐level forecasting behavior and its effectiveness in an evolving industry and suggests important implications for managers and entrepreneurs.Managerial SummaryHow a focal industry will evolve is a key forecasting problem faced by managers and entrepreneurs as they seek to identify opportunities and make strategic decisions. However, developing superior industry foresight in the face of significant change, and limited and often contradictory information, can be especially challenging. We study how individuals forecast the ongoing transformation of the global automotive industry with respect to electrification and autonomy, using a novel research design of forecasting tournaments. A forecasting process in which individuals update their beliefs by neither ignoring prior information nor overacting to new information helps to generate superior industry foresight. There was a significant penalty to forecasting accuracy when individuals did not update their beliefs at all, or when they updated, but overreacted to new information.

The computational relationship between reinforcement learning, social inference, and paranoia.

Theoretical accounts suggest heightened uncertainty about the state of the world underpin aberrant belief updates, which in turn increase the risk of developing a persecutory delusion. However, this raises the question as to how an agent's uncertainty may relate to the precise phenomenology of paranoia, as opposed to other qualitatively different forms of belief. We tested whether the same population (n = 693) responded similarly to non-social and social contingency changes in a probabilistic reversal learning task and a modified repeated reversal Dictator game, and the impact of paranoia on both. We fitted computational models that included closely related parameters that quantified the rigidity across contingency reversals and the uncertainty about the environment/partner. Consistent with prior work we show that paranoia was associated with uncertainty around a partner's behavioural policy and rigidity in harmful intent attributions in the social task. In the non-social task we found that pre-existing paranoia was associated with larger decision temperatures and commitment to suboptimal cards. We show relationships between decision temperature in the non-social task and priors over harmful intent attributions and uncertainty over beliefs about partners in the social task. Our results converge across both classes of model, suggesting paranoia is associated with a general uncertainty over the state of the world (and agents within it) that takes longer to resolve, although we demonstrate that this uncertainty is expressed asymmetrically in social contexts. Our model and data allow the representation of sociocognitive mechanisms that explain persecutory delusions and provide testable, phenomenologically relevant predictions for causal experiments.

Efficiently Detecting Non-Stationary Opponents: A Bayesian Policy Reuse Approach under Partial Observability

In multi-agent domains, dealing with non-stationary opponents that change behaviors (policies) consistently over time is still a challenging problem, where an agent usually requires the ability to detect the opponent’s policy accurately and adopt the optimal response policy accordingly. Previous works commonly assume that the opponent’s observations and actions during online interactions are known, which can significantly limit their applications, especially in partially observable environments. This paper focuses on efficient policy detecting and reusing techniques against non-stationary opponents without their local information. We propose an algorithm called Bayesian policy reuse with LocAl oBservations (Bayes-Lab) by incorporating variational autoencoders (VAE) into the Bayesian policy reuse (BPR) framework. Following the centralized training with decentralized execution (CTDE) paradigm, we train VAE as an opponent model during the offline phase to extract the latent relationship between the agent’s local observations and the opponent’s local observations. During online execution, the trained opponent models are used to reconstruct the opponent’s local observations, which can be combined with episodic rewards to update the belief about the opponent’s policy. Finally, the agent reuses the best response policy based on the updated belief to improve online performance. We demonstrate that Bayes-Lab outperforms existing state-of-the-art methods in terms of detection accuracy, accumulative rewards, and episodic rewards in a predator–prey scenario. In this experimental environment, Bayes-Lab can achieve about 80% detection accuracy and the highest accumulative rewards, and its performance is less affected by the opponent policy switching interval. When the switching interval is less than 10, its detection accuracy is at least 10% higher than other algorithms.

The computational roots of positivity and confirmation biases in reinforcement learning.

Humans do not integrate new information objectively: outcomes carrying a positive affective value and evidence confirming one's own prior belief are overweighed. Until recently, theoretical and empirical accounts of the positivity and confirmation biases assumed them to be specific to 'high-level' belief updates. We present evidence against this account. Learning rates in reinforcement learning (RL) tasks, estimated across different contexts and species, generally present the same characteristic asymmetry, suggesting that belief and value updating processes share key computational principles and distortions. This bias generates over-optimistic expectations about the probability of making the right choices and, consequently, generates over-optimistic reward expectations. We discuss the normative and neurobiological roots of these RL biases and their position within the greater picture of behavioral decision-making theories.

Social Prevalence Is Rationally Integrated in Belief Updating

People rely on social information to inform their beliefs. We ask whether and to what degree the perceived prevalence of a belief influences belief adoption. We present the results of two experiments that show how increases in a person's estimated prevalence of a belief led to increased endorsement of said belief. Belief endorsement rose when impressions of the belief's prevalence were increased and when initial beliefs were uncertain, as predicted by a Bayesian cue integration framework. Thus, people weigh social information rationally. An implication of these results is that social engagement metrics that prompt inflated prevalence estimates in users risk increasing the believability and adoption of viral misinformation posts.

Delaying metamemory judgments corrects the expectancy illusion in source monitoring: The role of fluency and belief.

In schema-based source monitoring, people mistakenly predict better source memory for expected sources (e.g., oven in the kitchen; expectancy effect), whereas actual source memory is better for unexpected sources (e.g., hairdryer in the kitchen; inconsistency effect; Schaper et al., 2019b). In three source-monitoring experiments, the authors tested whether a delay between study and metamemory judgments remedied this metamemory expectancy illusion. Further, the authors tested whether delayed judgments were based on in-the-moment experiences of retrieval fluency or updating of belief due to experiences with one's source memory. Participants studied source-item pairs and provided metamemory judgments either at study or after delay. After delay, they made judgments either on the complete source-item pair (eliciting no source retrieval, Experiment 1) or on the item only (eliciting covert, Experiment 1, or overt source retrieval, Experiments 2 and 3). Metamemory judgments at study showed the established illusory expectancy effect, as did delayed judgments when no source retrieval was elicited. However, when participants retrieved the source prior to delayed judgments, they predicted an inconsistency effect on source memory, which concurred with actual memory. Thus, delaying judgments remedied the metamemory expectancy illusion. Results further indicate that in-the-moment experiences of retrieval fluency and updated general belief about the effect of expectancy on source memory jointly contributed to this remedial effect. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

COVID-19 conspiracy ideation is associated with the delusion proneness trait and resistance to update of beliefs

The rapid spread of conspiracy ideas associated with the recent COVID-19 pandemic represents a major threat to the ongoing and coming vaccination programs. Yet, the cognitive factors underlying the pandemic-related conspiracy beliefs are not well described. We hypothesized that such cognitive style is driven by delusion proneness, a trait phenotype associated with formation of delusion-like beliefs that exists on a continuum in the normal population. To probe this hypothesis, we developed a COVID-19 conspiracy questionnaire (CCQ) and assessed 577 subjects online. Their responses clustered into three factors that included Conspiracy, Distrust and Fear/Action as identified using principal component analysis. We then showed that CCQ (in particular the Conspiracy and Distrust factors) related both to general delusion proneness assessed with Peter’s Delusion Inventory (PDI) as well as resistance to belief update using a Bias Against Disconfirmatory Evidence (BADE) task. Further, linear regression and pathway analyses suggested a specific contribution of BADE to CCQ not directly explained by PDI. Importantly, the main results remained significant when using a truncated version of the PDI where questions on paranoia were removed (in order to avoid circular evidence), and when adjusting for ADHD- and autistic traits (that are known to be substantially related to delusion proneness). Altogether, our results strongly suggest that pandemic-related conspiracy ideation is associated with delusion proneness trait phenotype.

Dynamic task-belief is an integral part of decision-making

Natural decisions involve two seemingly separable processes: inferring the relevant task (task-belief) and performing the believed-relevant task. The assumed separability has led to the traditional practice of studying task-switching and perceptual decision-making individually. Here, we used a novel paradigm to manipulate and measure macaque monkeys' task-belief and demonstrated inextricable neuronal links between flexible task-belief and perceptual decision-making. We showed that in animals, but not in artificial networks that performed as well or better than the animals, stronger task-belief is associated with better perception. Correspondingly, recordings from neuronal populations in cortical areas 7a and V1 revealed that stronger task-belief is associated with better discriminability of the believed-relevant, but not the believed-irrelevant, feature. Perception also impacts belief updating; noise fluctuations in V1 help explain how task-belief is updated. Our results demonstrate that complex tasks and multi-area recordings can reveal fundamentally new principles of how biology affects behavior in health and disease.

Biased learning under ambiguous information

This paper proposes a model of how biased individuals update beliefs in the presence of informational ambiguity. Individuals are ambiguous about the actual signal-generating process and interpret signals according to the model that can best support their biases. This paper provides a complete characterization of the limit beliefs under this rule. The presence of model ambiguity has the following effects. First, it destroys correct learning even if infinitely many informative signals can be observed. When the ambiguity is sufficiently high, individuals can justify their biases, leading to belief extremism and polarization. Second, an ambiguous individual can exhibit greater confidence than a Bayesian individual with any feasible model perception. This phenomenon comes from a novel complementary effect of different models in the belief set.