Bias Confidence Research Articles

Adaptive centroid prototype-based domain adaptation for fault diagnosis of rotating machinery without source data

Domain adaptation can effectively achieve fault diagnosis tasks with unlabeled target data using similar labeled source datasets during the training stage. However, the labeled source datasets are usually not directly accessible due to data privacy concerns, which restrict the application of the domain adaptation-based fault diagnosis methods. In this study, an adaptive centroid prototype-based domain adaptation (ACPDA) method is proposed to conduct fault diagnosis tasks in the unlabeled target domain without accessing source data. In ACPDA, an entropy-based adaptive prototype memory matrix is constructed to filter reliable samples and define the initial pseudo-label in the target domain. The centroid prototype is designed using all target data to update the pseudo-label and avoid confidence bias. Furthermore, the information maximization loss function is employed to reduce the feature distribution discrepancies. Extensive experiments on real wind turbine gearbox datasets demonstrate the effectiveness and superiority of the proposed ACPDA method.

A low-dimensional approximation of optimal confidence.

Human decision making is accompanied by a sense of confidence. According to Bayesian decision theory, confidence reflects the learned probability of making a correct response, given available data (e.g., accumulated stimulus evidence and response time). Although optimal, independently learning these probabilities for all possible data combinations is computationally intractable. Here, we describe a novel model of confidence implementing a low-dimensional approximation of this optimal yet intractable solution. This model allows efficient estimation of confidence, while at the same time accounting for idiosyncrasies, different kinds of biases and deviation from the optimal probability correct. Our model dissociates confidence biases resulting from the estimate of the reliability of evidence by individuals (captured by parameter α), from confidence biases resulting from general stimulus independent under and overconfidence (captured by parameter β). We provide empirical evidence that this model accurately fits both choice data (accuracy, response time) and trial-by-trial confidence ratings simultaneously. Finally, we test and empirically validate two novel predictions of the model, namely that 1) changes in confidence can be independent of performance and 2) selectively manipulating each parameter of our model leads to distinct patterns of confidence judgments. As a tractable and flexible account of the computation of confidence, our model offers a clear framework to interpret and further resolve different forms of confidence biases.

Confidence biases in problem gambling.

Subjective confidence plays an important role in guiding behaviour, especially when objective feedback is unavailable. Systematic misjudgements in confidence can foster maladaptive behaviours and have been linked to various psychiatric disorders. In this study, we adopted a transdiagnostic approach to examine confidence biases in problem gamblers across three levels: local decision confidence, global task performance confidence, and overall self-esteem. The importance of taking a transdiagnostic perspective is increasingly recognised, as it captures the dimensional nature of psychiatric symptoms that often cut across diagnostic boundaries. Accordingly, we investigated if any observed confidence biases could be explained by transdiagnostic symptom dimensions of Anxiety-Depression and Compulsive Behaviour and Intrusive Thought. This approach allows us to gain a more comprehensive understanding of the role of metacognitive processes in problem gambling, beyond the constraints of traditional diagnostic categories. Thirty-eight problem gamblers and 38 demographically matched control participants engaged in a gamified metacognition task and completed self-report questionnaires assessing transdiagnostic symptom dimensions. Compared to controls, problem gamblers displayed significantly elevated confidence at the local decision and global task levels, independent of their actual task performance. This elevated confidence was observed even after controlling for the heightened symptom levels of Anxiety-Depression and Compulsive Behaviour and Intrusive Thought among the problem gamblers. The results reveal a notable disparity in confidence levels between problem gamblers and control participants, not fully accounted for by the symptom dimensions Anxiety-Depression and Compulsive Behaviour and Intrusive Thought. This suggests the contribution of other factors, perhaps linked to gambling-specific cognitive distortions, tothe observed confidence biases. The findings highlight the intricate link between metacognitive confidence and psychiatric symptoms in the context of problem gambling. It underscores the need for further research into metacognitive biases, which could enhance therapeutic approaches for individuals with psychiatric conditions.

Utilizing tACS to enhance memory confidence and EEG to predict individual differences in brain stimulation efficacy.

The information transfer necessary for successful memory retrieval is believed to be mediated by theta and gamma oscillations. These oscillations have been linked to memory processes in electrophysiological studies, which were correlational in nature. In the current study, we used transcranial alternating current stimulation (tACS) to externally modulate brain oscillations to examine its direct effects on memory performance. Participants received sham, theta (4 Hz), and gamma (50 Hz) tACS over frontoparietal regions while retrieving information in a source memory paradigm. Linear regression models were used to investigate the direct effects of oscillatory non-invasive brain stimulation (NIBS) on memory accuracy and confidence. Our results indicate that both theta and gamma tACS altered memory confidence. Specifically, theta tACS seemed to lower the threshold for confidence in retrieved information, while gamma tACS appeared to alter the memory confidence bias. Furthermore, the individual differences in tACS effects could be predicted from electroencephalogram (EEG) measures recorded prior to stimulation, suggesting that EEG could be a useful tool for predicting individual variability in the efficacy of NIBS.

Moderate Message Passing Improves Calibration: A Universal Way to Mitigate Confidence Bias in Graph Neural Networks

Confidence calibration in Graph Neural Networks (GNNs) aims to align a model's predicted confidence with its actual accuracy. Recent studies have indicated that GNNs exhibit an under-confidence bias, which contrasts the over-confidence bias commonly observed in deep neural networks. However, our deeper investigation into this topic reveals that not all GNNs exhibit this behavior. Upon closer examination of message passing in GNNs, we found a clear link between message aggregation and confidence levels. Specifically, GNNs with extensive message aggregation, often seen in deep architectures or when leveraging large amounts of labeled data, tend to exhibit overconfidence. This overconfidence can be attributed to factors like over-learning and over-smoothing. Conversely, GNNs with fewer layers, known for their balanced message passing and superior node representation, may exhibit under-confidence. To counter these confidence biases, we introduce the Adaptive Unified Label Smoothing (AU-LS) technique. Our experiments show that AU-LS outperforms existing methods, addressing both over and under-confidence in various GNN scenarios.

Confidence bias, hindsight bias and perceptions of risk in motorcyclists

Confidence bias and hindsight bias were investigated as predictors of perceived levels of risk in a sample of motorcyclists. Confidence bias was predictive of both perceived absolute risk and perceived comparative risk. Hindsight bias was predictive of total absolute risk but was only predictive of comparative risk regarding car drivers and not total comparative risk or risk compared to other motorcyclists. In addition, the direction of the relationship with risk perception differed for confidence bias compared to hindsight bias. It appears that confidence and hindsight bias operate through different processes. Personal experience of accidents was not a predictor of perceived risk, but having a friend who had an accident did have an impact.

Double Confidence Calibration Focused Distillation for Task-Incremental Learning.

Task-incremental learning methods that adopt knowledge distillation face two significant challenges: confidence bias and knowledge loss. These challenges make it difficult to effectively balance the stability and plasticity of the network in the incremental learning process. In this article, we propose double confidence calibration focused distillation (DCCFD) to address these challenges. We introduce intratask and intertask confidence calibration (ECC) modules that can mitigate network overconfidence during incremental learning and reduce the degree of feature representation bias. We also propose a focused distillation (FD) module that can alleviate the problem of knowledge loss during the task increment process, improving model stability without reducing plasticity. Experimental results on the CIFAR-100, TinyImageNet, and CORE-50 datasets demonstrate the effectiveness of our method, with performance that matches or exceeds the state of the art. Furthermore, our method can be used as a plug-and-play module to consistently improve class-incremental learning methods.

Use of PD-1 blockade in refractory/relapsed natural killer T-cell lymphomas: a systematic review and synthesis of case reports

Natural killer/T-cell lymphoma (NK/T-cellL) is an aggressive non-Hodgkin’s lymphoma with limited treatment options for patients who experience disease progression or recurrence after second-line treatment. The use of new therapies, such as pembrolizumab, which involves immune checkpoint blockade mechanisms, is proposed. This systematic review followed the MOSE guidelines and searched PUBMED/MEDLINE, EMBASE, and Scopus databases. Fourteen articles were found, reporting on the use of pembrolizumab anti PD-1 in NK/T-cellL patients. The objective response rate was 84.50%, with disease-free survival ranging from two to 48 months. The complete response rate was 61.6%, and the quality of the reported studies was evaluated to be of high and moderate confidence bias levels in case reports and high bias in clinical trials. Pembrolizumab and others anti PD-1 are treatment options for refractory/recurrent NK/T-cellL, regardless of PD-L1 expression, with good short- and long-term results and low adverse events.

Natural statistics support a rational account of confidence biases

Previous work has sought to understand decision confidence as a prediction of the probability that a decision will be correct, leading to debate over whether these predictions are optimal, and whether they rely on the same decision variable as decisions themselves. This work has generally relied on idealized, low-dimensional models, necessitating strong assumptions about the representations over which confidence is computed. To address this, we used deep neural networks to develop a model of decision confidence that operates directly over high-dimensional, naturalistic stimuli. The model accounts for a number of puzzling dissociations between decisions and confidence, reveals a rational explanation of these dissociations in terms of optimization for the statistics of sensory inputs, and makes the surprising prediction that, despite these dissociations, decisions and confidence depend on a common decision variable.

The Effect of Overconfidence Bias, Herding Behavior and Experienced Regret on Investment Decision Making on the Indonesia Stock Exchange

Indonesia will experience a surge in investors in 2021 almost twice the number of investors before. Although investment is starting to be favored by the public, not all investors are able to make the right investment decisions. This research examines the influence of overconfidence bias, herding behavior and experienced regret on investment decisions. This study used 184 respondents who live in Jakarta. The sample was obtained through a purposive sampling technique in which respondents had to meet the criteria of having invested in the capital market for at least 1 year and made at least 2 transactions. This study collected data by distributing questionnaires through the Google form and the data was measured using a Likert sca le and analyzed using SmartPLS software. The results obtained in this study are over confidence bias has no effect on investment decisions, herding behavior has no effect on investment decisions and experienced regret has a positive effect on investment.

Can Monitoring Prompts Help to Reduce a Confidence Bias When Learning With Multimedia?

Abstract: Accurate monitoring is important, as it guides subsequent control of learning. With multimedia, learners often overestimate their learning. Retrieval practice provides insight into one’s understanding; still, it might not improve monitoring. In an online experiment with two groups, participants ( N = 146) performed retrieval practice while studying multimedia instructions. We investigated whether monitoring prompts affect metacognitive accuracy, subsequent learning behavior, and performance: One group repeatedly received prompts, whereas a control group did not. After studying, both groups provided metacognitive judgments for a subsequent posttest. As expected, learners in the control group were overconfident about their learning; however, unexpectedly, prompted learners were also overconfident and did not differ from the control group. Nevertheless, prompted learners provided longer answers to retrieval tasks, which increased their posttest performance (indirect mediation). We observed no mediation for learning time or retrieval performance, although retrieval performance did relate to posttest performance. The results suggest that monitoring prompts do not improve metacognitive accuracy per se but might improve learning outcomes by controlling subsequent learning.

Financial Literacy Confidence and Retirement Planning: Evidence from China

Though ample empirical evidence demonstrates the relationship between objective financial literacy and retirement planning, we have a limited understanding of the role of individuals’ subjective financial literacy in their retirement planning. In this study, we examine how individuals’ financial literacy confidence bias affects their retirement planning behaviors using survey data in China. Based on the difference between respondents’ subjective and objective financial literacy from survey data, we construct measures of individuals’ financial literacy overconfidence and underconfidence for empirical analysis. Our results document the critical role of individuals’ assessment of financial literacy in their retirement planning. We find that individuals’ financial literacy overconfidence (underconfidence) significantly promotes (demotes) their retirement planning behaviors.

Supplemental Material for Linking Confidence Biases to Reinforcement-Learning Processes

Supplemental Material for Linking Confidence Biases to Reinforcement-Learning Processes

Is there a survival processing effect in metacognition?

Memory systems serve an adaptive function for the fitness of organisms. A good example of this is the Survival Processing Effect (SPE) which points to increased retention of information when it is processed in a survival context compared to other contexts. Survival processing may also affect metacognitive processes, by increasing confidence judgments as well as increasing metacognitive sensitivity. No previous study, however, has directly examined whether processing information for survival also has an effect on metacognitive processes. Here we ask whether SPE extends to the metacognitive system in terms of both metacognitive sensitivity and confidence bias. In Experiment 1 participants were asked to rate a list of words in terms of relevance in a survival scenario or a moving scenario. In a surprise old/new recognition test, they were given one word at a time and asked to indicate if they have rated the presented word before and state how confident they are in that choice. Surprisingly, the results did not reveal a SPE, which may have been due to high overall performance in the recognition task. In Experiment 2 we increased the level of difficulty of the memory task, which resulted in a robust SPE, but could not find this effect in metacognitive monitoring. Together, these results suggest that survival processing may not affect metacognitive processes in a reliable fashion.

Contributions of age and clinical depression to metacognitive performance

One aspect of metacognition is the ability to judge the accuracy of our own performance, even in the absence of external feedback, which is often measured using confidence ratings.Past research suggests that confidence is lower in Major Depressive Disorder (MDD). Less is known about the ability of MDD patients to discriminate correct from incorrect performance (metacognitive efficiency). The metacognitive performance of aged MDD patients (62-89y) was compared to an age-matched control group. A younger control group (21-28y) was included to also explore the relationshipbetweenageandmetacognitive performance.We found no difference in confidence bias nor metacognitive efficiency between MDD patients and age-matched controls.We found age-related differences in metacognition:normal aging was associated with higher confidencebut lower metacognitive efficiency. The overconfidence was specifically driven by overconfidence in incorrect trials. Our results point to the importance ofage while investigating the relation between MDD and metacognitive performance.

Modeling perceptual confidence and the confidence forced-choice paradigm.

Perceptual confidence is an evaluation of the validity of our perceptual decisions. We present here a complete generative model that describes how confidence judgments result from some confidence evidence. The model that generates confidence evidence has two main parameters, confidence noise and confidence boost. Confidence noise reduces the sensitivity to the confidence evidence, and confidence boost accounts for information used for confidence judgment which was not used for the perceptual decision. The opposite effect of these two parameters creates a problem of confidence parameters indeterminacy, where the confidence in a perceptual decision is the same in spite of differences in confidence noise and confidence boost. When confidence is estimated for multiple stimulus strengths, both of these parameters can be recovered, thus allowing us to estimate whether confidence is generated using the same primary information that was used for the perceptual decision or some secondary information. We also describe a novel measure of confidence efficiency relative to the ideal confidence observer, as well as the estimate of one type of confidence bias. Finally, we apply the model to the confidence forced-choice paradigm, a paradigm that provides objective estimates of confidence, and we discuss how each parameter of the model can be recovered using this paradigm. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Are you confident enough to act? Individual differences in action control are associated with post-decisional metacognitive bias.

The art of making good choices and being consistent in executing them is essential for having a successful and fulfilling life. Individual differences in action control are believed to have a crucial impact on how we make choices and whether we put them in action. Action-oriented people are more decisive, flexible and likely to implement their intentions in the face of adversity. In contrast, state-oriented people often struggle to commit to their choices and end up second-guessing themselves. Here, we employ a model-based computational approach to study the underlying cognitive differences between action and state-oriented people in simple binary-choice decision tasks. In Experiment 1 we show that there is little-to-no evidence that the two groups differ in terms of decision-related parameters and strong evidence for differences in metacognitive bias. Action-oriented people exhibit greater confidence in the correctness of their choices as well as slightly elevated judgement sensitivity, although no differences in performance are present. In Experiment 2 we replicate this effect and show that the confidence gap generalizes to value-based decisions, widens as a function of difficulty and is independent of deliberation interval. Furthermore, allowing more time for confidence deliberation indicated that state-oriented people focus more strongly on external features of choice. We propose that a positive confidence bias, coupled with appropriate metacognitive sensitivity, might be crucial for the successful realization of intentions in many real-life situations. More generally, our study provides an example of how modelling latent cognitive processes can bring meaningful insight into the study of individual differences.

CEO confidence bias and strategic choice: a general framework

ABSTRACT An owner of a firm may choose to hire an unbiased CEO or one with confidence bias. We develop a model that demonstrates that the owner’s optimal choice depends on whether the firm and rival choice variables are strategic substitutes or strategic complements. When choice variables are strategic substitutes or strategic complements for both firms, owners optimize by hiring overconfident CEOs. When choice variables are substitutes for one firm and complements for the rival firm, each firm optimizes by hiring an underconfident CEO. We show that the model applies to price and output competition, advertising, research and development spending, and product design.

Overconfidence as a driver of entrepreneurial market entry decisions: a critical appraisal

In characterizing entrepreneurial behavior, researchers often regard nascent entrepreneurs entering risky markets as overconfident. In this paper, we challenge this prevailing view and show that a more differentiated consideration reveals the effects of overconfidence on market entry to be ambiguous if not irrelevant. In a first step, we emphasize the inconclusiveness of past empirical evidence. With a simple decision model, we show that “observed” overconfidence can also be interpreted as the outcome of fully rational behavior when acknowledging information available to and acquired by the entrepreneur. We criticize that empirical studies, which neglect individually available information, inductively propose overconfidence for observed behavior without substantiating it with appropriate data. Nevertheless, we do not generally deny overconfidence in decision making. Thus, in a second step, we explicitly postulate confidence biases of an intended rational decision maker. This enables us to analyze through which channels they may affect market entry. By considering over- and underconfidence in the entrepreneur’s forecasting ability, we find that market entry may be affected positively, negatively, or not at all, thereby revealing the overall ambiguity of confidence biases in decision behavior. Finally, we show formally that, even if overconfidence increases the probability of market entry, this does not make it a characteristic feature of market entrants. Overall, our objective is to debunk the myth of the overconfident entrepreneur and to promote the more important role of information in entrepreneurial decision making.

Pengaruh Herding Bias dan Overconfidence Bias terhadap Pengambilan Keputusan Investasi

This research study was conducted to analyze the effect of herding bias and overconfidence bias on investment decision-making. This research used quantitative methods with data collection techniques using questionnaires distributed to individual investors who have experience with a total sample of about 124 respondents. The technique used for sampling in this research was simple random sampling. The research analysis methods used the classical assumption test such as normality test, multicollinearity test, heteroscedasticity test, and multiple linear regression analysis, namely hypothesis testing. According to the analysis, the result of this research is that herding bias and confidence bias had a positive and significant effect on investment decision-making.