Underconfidence Bias Research Articles

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.

Gender-related effects of financial knowledge and confidence on preferences for ethical intermediaries and sustainable investments

PurposeThis study analyses the role of individuals' objective financial knowledge in shaping preferences for ethical intermediaries and sustainable investments in Italy. Another goal of this study is to assess the impact of individuals' misperceptions about their own financial knowledge and to test for gender-related differences in attitudes towards socially responsible investing (SRI).Design/methodology/approachUsing nationally representative microdata from the Bank of Italy’s “Italian Literacy and Financial Competence Survey” (IACOFI), the authors use probit models, extended to account for potential endogeneity issues, to assess the causal effects of financial knowledge and confidence on stated preferences for SRI. Empirical models also allow to explicitly assess the moderating role of gender on the effects of financial knowledge and confidence on attitudes towards sustainable investing.FindingsResults indicate that individuals' preferences for sustainable finance significantly increase with financial knowledge, suggesting that inadequate financial competencies represent a barrier to participation in SRI. At the same time, lack of confidence in one’s own financial knowledge significantly hampers attitudes towards sustainable investments. Furthermore, the authors show that women have a greater preference for sustainable finance than men and point out that financial knowledge and confidence exert heterogenous effects on attitudes towards SRI.Originality/valueThis study provides several contributions to the literature on SRI. First, the authors give evidence of the causal effect of financial knowledge on preferences for both ethical financial intermediaries and sustainable investments. Moreover, this is the first study to investigate the role of financial underconfidence bias in shaping individuals' SRI attitudes. Finally, extending previous research, the authors assess differences in SRI preferences between women and men and provide novel evidence on gender-related heterogeneity in the effects of financial knowledge and underconfidence.

The Contagion of Unethical Behavior and Social Learning: An Experimental Study

Unethical behavior is discovered that is more contagious than ethical behavior. This article attempts to propose one of the possible underlying mechanisms-people may have underconfidence bias in information updating due to motivated reasoning, and such bias exhibits in a different direction compared to the overconfident bias documented in the literature on ethical environment, which generate the asymmetric pattern in contagion. This study designs an experiment which relates the unethical behavior to social learning, where a series of subjects with private information about penalty decide sequentially whether to conduct unethical behavior publicly. This study adopts a quantal response equilibrium to construct a structural model for estimation of the bias. In total, 162 university students participated in our experiment and the results confirm the asymmetric patterns that people rely more on others' precedent decisions rather than their private signal; therefore, the bias facilitates the contagion. This study also tests two punishment systems in the experiment and the results suggest a policy: slightly increasing penalties for the "followers" in the early stages would effectively suppress the contagion.

Covid-19’s Impact on Financial Planning in Short-Term and Long-Term Perspectives

This research investigates psychological factors including underconfidence bias, analytical/intuitive decision-making approach, self/significant other orientation, and time orientation that influence people's financial planning in short and long-term perspectives under the Covid-19 pandemic situation. The data were collected from the online survey by using a convenience sampling technique. A total of 331 valid responses were scaled on 7 Likert-type questions. Multiple regression analyses were performed in analyzing the hypotheses. This study provides empirical insight into the relationship between psychological factors and financial planning in the short and long-term perspectives. The results indicate both financial planning in short and long-term perspectives is positively associated with analytical decision-making style, time orientation, self, and other orientation. Underconfidence bias is negatively associated only with long-term financial planning but, in contrast, is not supported by evidence to a short-term perspective. Finally, intuitive-decision making is not correlated with financial planning in short-term perspective. This study helps individuals to design better personal financial planning and financial institution to develop more relevant financial services for targeted clients. This study uses consumer behavior's perspective and figures out the gap in consumer behavior literature, especially investigating the underconfidence bias effect on Indonesian's financial planning.

Does underconfidence matter in short-term and long-term investment decisions? Evidence from an emerging market

PurposeThe purpose of this article is to clarify the mechanism by which underconfidence heuristic-driven bias influences the short-term and long-term investment decisions of individual investors, actively trading on the Pakistan Stock Exchange.Design/methodology/approachInvestors' underconfidence has been measured using a questionnaire, comprising numerous items, including indicators of short-term and long-term investment decision. In order to establish the influence of underconfidence on the investment decisions in both the short and long run, a 5-point Likert scale questionnaire has been used to collect data from the sample of 203 investors. The collected data were analyzed using SPSS and AMOS graphics software. Hypotheses were tested using structural equation modeling technique.FindingsThis article provides further empirical insights into the relationship between heuristic-driven biases and investment decision-making in the short and long run. The results suggest that underconfidence bias has a markedly negative influence on the short-term and long-term decisions made by investors in developing markets. It means that heuristic-driven biases can impair the quality of both short-term and long-term investment decisions.Practical implicationsThis article encourages investors to avoid relying on cognitive heuristics, namely, underconfidence or their feelings when making short-term and long-term investment strategies. It provides awareness and understanding of heuristic-driven biases in investment management, which could be very useful for finance practitioners' such as investor who plays at the stock exchange, a portfolio manager, a financial strategist/advisor in an investment firm, a financial planner, an investment banker, a trader/broker at the stock exchange or a financial analyst. But most importantly, the term also includes all those persons who manage corporate entities and are responsible for making its financial management strategies. They can improve the quality of their decision-making by recognizing their behavioral biases and errors of judgment, to which we are all prone, resulting in more appropriate investment strategies.Originality/valueThe current study is the first to focus on links between underconfidence bias and short-term and long-term investment decision-making. This article enhanced the understanding of the role that heuristic-driven bias plays in the investment management and more importantly, it went some way toward enhancing understanding of behavioral aspects and their influence on the investment decision-making in an emerging market. It also adds to the literature in the area of behavioral finance specifically the role of heuristics in investment strategies; this field is in its initial stage, even in developed countries, while, in developing countries, little work has been done.

Temporal attention causes systematic biases in visual confidence

Temporal attention enhances the perceptual representation of a stimulus at a particular point in time. The number of possible attentional episodes in a given period is limited, but whether observers’ confidence reflects such limitations is still unclear. To investigate this issue, we adapted an “Attentional Blink” paradigm, presenting observers with a rapid visual stream of letters containing two targets cued for subsequent perceptual reports and confidence judgments. We found three main results. First, when two targets fell within the same attentional episode, the second target underwent a strong under-confidence bias. In other words, confidence neglected that a single attentional episode can benefit to both targets. Second, despite this initial bias, confidence was strongly correlated with response probability. Third, as confidence was yoked to the evidence used in perceptual reports, it remains blind to delays in response selection for the second target. Notably, the second target was often mistaken with a later item associated with higher confidence. These results suggest that confidence does not perfectly evaluate the limits of temporal attention in challenging situations.

Effects of Working Memory Capacity on Metacognitive Monitoring: A Study of Group Differences Using a Listening Span Test.

Monitoring is an executive function of working memory that serves to update novel information, focusing attention on task-relevant targets, and eliminating task-irrelevant noise. The present research used a verbal working memory task to examine how working memory capacity limits affect monitoring. Participants performed a Japanese listening span test that included maintenance of target words and listening comprehension. On each trial, participants responded to the target word and then immediately estimated confidence in recall performance for that word (metacognitive judgment). The results confirmed significant differences in monitoring accuracy between high and low capacity groups in a multi-task situation. That is, confidence judgments were superior in high vs. low capacity participants in terms of absolute accuracy and discrimination. The present research further investigated how memory load and interference affect underestimation of successful recall. The results indicated that the level of memory load that reduced word recall performance and led to an underconfidence bias varied according to participants' memory capacity. In addition, irrelevant information associated with incorrect true/ false decisions (secondary task) and word recall within the current trial impaired monitoring accuracy in both participant groups. These findings suggest that interference from unsuccessful decisions only influences low, but not high, capacity participants. Therefore, monitoring accuracy, which requires high working memory capacity, improves metacognitive abilities by inhibiting task-irrelevant noise and focusing attention on detecting task-relevant targets or useful retrieval cues, which could improve actual cognitive performance.

Irrational Market Makers

We analyze a model where irrational and rational informed traders exchange a risky asset with irrational market makers. Irrational traders misperceive the mean of prior information (optimistic/pessimistic bias) and the variance of the noise in their private signal (overconfidence/ underconfidence bias). Irrational market makers misperceive both the mean and the variance of the prior information. We show that moderately underconfident traders can outperform rational ones and that irrational market makers can fare better than rational ones. Lastly, we find that extreme level of confidence implies high trading volume.

Modelado matemático del efecto del anclaje sobre el sesgo de subconfianza

Se define el efecto de anclaje como la tendencia del pensamiento humano a la fijación endeterminados fragmentos de información que afectan la toma de decisiones. El sesgo de subconfianza, por su parte, se entiende como la propensión de un agente a subestimar su desempeño objetivo. Si bien se conocen algunos aspectos de la interacción entre ambos efectos, no se dispone de modelos formales que describan tal relación específica. Este estudio presenta: i) un experimento que pone en evidencia el efecto que ejerce el anclaje tanto para reducir, como para amplificar la subconfianza en tareas de decisión geográfica y ii) los fundamentos de un modelo matemático novedoso que permite estudiar con mayor precisión la relación del anclaje con la subconfianza

The effects of domain knowledge on metacomprehension accuracy

In the present research, we examined the relationship between readers' domain knowledge and their ability to judge their comprehension of novel domain-related material. Participants with varying degrees of baseball knowledge read five texts on baseball-related topics and five texts on non-baseball-related topics, predicted their performance, and completed tests for each text. Baseball knowledge was positively related to absolute accuracy within the baseball domain but was unrelated to relative accuracy within the baseball domain. Also, the readers showed a general underconfidence bias, but the bias was less extreme for higher knowledge readers. The results challenge common assumptions that experts' metacognitive judgments are less accurate than novices'. Results involving topic familiarity ratings and a no-reading control group suggest that higher knowledge readers are not more likely to ignore text-specific cues in favor of a domain familiarity heuristic, but they do appear to make more effective use of domain familiarity in predicting absolute performance levels.

Efecto del entrenamiento sobre la calibración subjetiva de éxito en tareas verbales

The calibration or confidence is broadly defined as the relation of correspondence or discrepancy between the subjective estimation of success and the actual success in a series of tasks. The underconfidence bias is defined as underestimation and the overconfidence bias as overestimation. Underconfidence and overconfidence biases are presented as cognitive phenomena. The main debiasing methods are analyzed. Two debiasing strategies are identified. The methodological strategy is differentiated from the cognitive strategy. An experiment that dissolves the underconfidence bias in verbal tasks through a simple training is presented and analyzed. It is suggested that the dissolution is due to training in calibration rather than training in the experimental task. The effect of adjustment in calibration is explained by the adaptive participation of metacognitive processing.

Realism of confidence in sensory discrimination

Is there a common and general basis for confidence in human judgment? Recently, we found that the properties of confidence judgments in the sensory domain mirror those previously established in the cognitive domain; notably, we found underconfidence on easy sensory judgments and overconfidence on hard sensory judgments. In contrast, data from the Uppsala laboratory in Sweden suggest that sensory judgments are unique; they found a pervasive underconfidence bias, with overconfidence being evident only on very hard sensory judgments. Olsson and Winman (1996) attempted to resolve the debate on the basis of methodological issues related to features of the stimulus display in a visual discrimination task. A reanalysis of the data reported in Baranski and Petrusic (1994), together with the findings of a new experiment that controlled stimulus display characteristics, supports the position that the difference between the Canadian and the Swedish data is real and, thus, may reflect cross-national differences in confidence in sensory discrimination.

Underconfidence in sensory discrimination: the interaction between experimental setting and response strategies.

In a recent issue of this journal, Baranski and Petrusic (1994) presented emperical data revealing overconfidence in sensory discrimination. In this paper, we propose an explanation of Baranski and Petrusic's results, based on an idiosyncrasy in the experimental setting that misleads subjects who are using an unwarranted symmetry assumption. Experiment 1 showed that when this hypothesis is controlled for, a large underconfidence bias is obtained with Baranski and Petrusic's procedure. The results of Experiment 2 confirmed that overconfidence is difficult to obtain in subject-controlled sensory discrimination tasks, even for a very low proportion of correct responses. The different results obtained in sensory and cognitive tasks suggest that one should not uncritically draw parallels between confidence in sensory and cognitive judgments.

A model for realism of confidence judgments: implications for underconfidence in sensory discrimination.

In a recent issue of this journal, Björkman, Justlin, and Winman (1993) presented a model of the calibration of subjective confidence judgments for sensory discrimination which they called "subjective distance theory." They proposed that there was a robust underconfidence bias in such judgments, that the model predicted such a bias, and that two different models were needed for the calibration of subjective confidence for cognitive judgments and for sensory ones. This paper addresses issues they raised. It points out that they have not presented a new model, but rather a portion of a more general one, the "decision-variable partition model" originally proposed in Ferrell and McGoey (1980). This paper explores properties of the model and shows, contrary to Björkman, Juslin, and Winman's hypotheses, that the model does not predict underconfidence, that the "hard-easy effect" can be observed with sensory discriminations, and that the model fits not only sensory, but also cognitive judgments.

Calibration of sensory and cognitive judgments: A single model for both

In a recent issue of this journal, Winman and Juslin (34, 135-148, 1993) present a model of the calibration of subjective probability judgments for sensory discrimination tasks. They claim that the model predicts a pervasive underconfidence bias observed in such tasks, and present evidence from a training experiment that they interpret as supporting the notion that different models are needed to describe judgment of confidence in sensory and in cognitive tasks. The model is actually part of the more comprehensive decision variable partition model of subjective probability calibration that was originally proposed in Ferrell and McGoey (Organizational Behavior and Human Performance, 26, 32-53, 1980). The characteristics of the model are described and it is demonstrated that the model does not predict underconfidence, that it is fully compatible with the overconfidence frequently found in calibration studies with cognitive tasks, and that it well represents experimental results from such studies. It is concluded that only a single model is needed for both types of task.

Calibration of sensory and cognitive judgments: Two different accounts

The purpose of this study was to test predictions of two recent theories of realism of confidence. Ecological approaches to realism of confidence in one's general knowledge (Gigerenzer et al., 1991; Juslin, in press a; Björkman, in press) predict good calibration or, in the case of poor cognitive adjustment, overconfidence, within the cognitive domain. The subjective distance theory of confidence in sensory discriminations (Björkman et al., 1992) predicts a pervasive underconfidence bias for sensory discriminations. Empirical data are reported showing that: (a) Calibration for sensory judgments is considerably poorer than calibration for well adapted cognitive judgements, a difference that can be entirely traced to underconfidence in the sensory domain. (b) While an initial overconfidence bias in the cognitive domain is removed by outcome feedback, the bias observed in sensory discriminations is unaffected even by a prolonged feedback session. It is suggested that the nature of confidence in sensory discriminations is different from the nature of confidence in cognitive judgments.

Realism of confidence in sensory discrimination: the underconfidence phenomenon.

This paper documents a very pervasive underconfidence bias in the area of sensory discrimination. In order to account for this phenomenon, a subjective distance theory of confidence in sensory discrimination is proposed. This theory, based on the law of comparative judgment and the assumption of confidence as an increasing function of the perceived distance between stimuli, predicts underconfidence--that is, that people should perform better than they express in their confidence assessments. Due to the fixed sensitivity of the sensory system, this underconfidence bias is practically impossible to avoid. The results of Experiment 1 confirmed the prediction of underconfidence with the help of present-day calibration methods and indicated a good quantitative fit of the theory. The results of Experiment 2 showed that prolonged experience of outcome feedback (160 trials) had no effect on underconfidence. It is concluded that the subjective distance theory provides a better explanation of the underconfidence phenomenon than do previous accounts in terms of subconscious processes.