Confirmation Bias Research Articles

Max Explainability Score–A quantitative metric for explainability evaluation in knowledge graph-based recommendations

Applying machine learning (ML) models within recommender systems (RSs) has proven effective in achieving high recommendation accuracy. However, a prevalent drawback is their inherent lack of explicability. Integrating knowledge graphs (KGs) into RSs enhances interpretability by elucidating the reasoning behind specific recommendations. Explainability evaluation in RSs often relies on subjective metrics assessed using qualitative user feedback. Although they offer a robust foundation for evaluating explainability, they are susceptible to confirmation bias. Quantitative metrics for assessing the explainability of KG-driven RSs are currently unavailable. This study, thus, proposes novel metrics for the quantitative evaluation of explanation quality in KG-based RSs. The proposed metric, Max Explainability Score, is based on four evaluation parameters: the number of rules, the probability of the traversal path, the entropy value and the reward due to the chosen traversal paths. User studies employing these metrics offer enhanced validity and better integration potential with future XAI research.

Jihadist Terrorist Attacks and Far-Right Party Preferences: An “Unexpected Event During Survey Design” in Four European Countries

This article presents new empirical evidence about the impact of Jihadist terrorist attacks on far-right preferences using the “unexpected event during survey” research design. This strategy allows us to match individual-level data from the European Social Survey (ESS) to data on Jihadist terrorist attacks to compare respondents’ party preferences before and after a terrorist attack during the same survey period in the Netherlands, Sweden, France, and Germany. We theorise and test three distinct hypotheses about how different combinations of attitudinal changes including out-group prejudice and trust in institutions impact far-right preferences. We find no statistically significant effects. Analyses of the two indirect mechanisms— i.e., prejudice and trust—yield mixed results consistent with the null effect on far-right party preferences. By showing that terrorist attacks are unlikely to decisively change party support despite attracting significant public attention and affecting political attitudes, our results challenge the argument that Jihadist terrorism necessarily benefits the far-right and highlight the importance of null effects for overcoming confirmation bias in the study of voting behaviour.

Addicted to socialising and still lonely: Acomparative, corpus-driven analysis ofproblematic social networking site use.

Problematic Social Networking Site Use (PSNSU) is not a formally recognised addiction, but it is increasingly discussed as such in academic research and online. Taking a quantitative, exploratory approach, this study aims to (1) determine whether PSNSU is presented like clinically defined addictions by the affected community and (2) address how well measurements of PSNSU fit with the thematic content found within the associated discourse. Four corpora were created for this study: a corpus concerning PSNSU and three control corpora concerning established addictions, including Alcohol Use Disorder, Tobacco Use Disorder and Gaming Disorder. Keywords were identified, collocates and concordances were explored, and shared themes were compared. Findings show broad thematic similarities between PSNSU and the three control addictions as well as prominent interdiscursive references, which indicate possible confirmation bias among speakers. Scales based upon the components model of addiction are suggested as the most appropriate measure of this emerging disorder.

Cognitive Biases in High-Stakes Decision-Making: Implications for Joint Pediatric Cardiology and Cardiothoracic Surgery Conference.

Extensive research has consistently demonstrated that humans frequently diverge from rational decision-making processes due to the pervasive influence of cognitive biases. This paper conducts an examination of the impact of cognitive biases on high-stakes decision-making within the context of the joint pediatric cardiology and cardiothoracic surgery conference, offering practical recommendations for mitigating their effects. Recognized biases such as confirmation bias, availability bias, outcome bias, overconfidence bias, sunk cost fallacy, loss aversion, planning fallacy, authority bias, and illusion of agreement are analyzed concerning their specific implications within this conference setting. To counteract these biases and enhance the quality of decision-making, practical strategies are proposed, including the implementation of a no-interruption policy until all data is reviewed, leaders refraining from immediate input, requiring participants to formulate independent judgments prior to sharing recommendations, explicit probability estimations grounded in base rates, seeking external opinions, and promoting an environment that encourages dissenting perspectives.

Data Adaptive Traceback for Vision-Language Foundation Models in Image Classification

Vision-language foundation models have been incredibly successful in a wide range of downstream computer vision tasks using adaptation methods. However, due to the high cost of obtaining pre-training datasets, pairs with weak image-text correlation in the data exist in large numbers. We call them weak-paired samples. Due to the limitations of these weak-paired samples, the pre-training model are unable to mine all the knowledge from pre-training data. The existing adaptation methods do not consider the missing knowledge, which may lead to crucial task-related knowledge for the downstream tasks being ignored. To address this issue, we propose a new adaptation framework called Data Adaptive Traceback (DAT). Specifically, we utilize a zero-shot-based method to extract the most downstream task-related subset of the pre-training data to enable the downstream tasks. Furthermore, we adopt a pseudo-label-based semi-supervised technique to reuse the pre-training images and a vision-language contrastive learning method to address the confirmation bias issue in semi-supervised learning. We conduct extensive experiments that show our proposed DAT approach meaningfully improves various benchmark datasets’ performance over traditional adaptation methods by simply.

Document-Level Relation Extraction with Uncertainty Pseudo-Label Selection and Hard-Sample Focal Loss

Relation extraction is a fundamental task in natural language processing that aims to identify structured triple relationships from unstructured text. In recent years, research on relation extraction has gradually advanced from the sentence level to the document level. Most existing document-level relation extraction (DocRE) models are fully supervised and their performance is limited by the dataset quality. However, existing DocRE datasets suffer from annotation omission, making fully supervised models unsuitable for real-world scenarios. To address this issue, we propose the DocRE method based on uncertainty pseudo-label selection. This method first trains a teacher model to annotate pseudo-labels for a dataset with incomplete annotations, trains a student model on the dataset with annotated pseudo-labels, and uses the trained student model to predict relations on the test set. To mitigate the confirmation bias problem in pseudo-label methods, we performed adversarial training on the teacher model and calculated the uncertainty of the model output to supervise the generation of pseudo-labels. In addition, to address the hard-easy sample imbalance problem, we propose an adaptive hard-sample focal loss. This loss can guide the model to reduce attention to easy-to-classify samples and outliers and to pay more attention to hard-to-classify samples. We conducted experiments on two public datasets, and the results proved the effectiveness of our method.

Does behavioral biases matter in SMEs' borrowing decisions? Insights from Morocco

Bank financing decisions by small and medium-sized enterprises (SMEs) are crucial to their growth and survival, particularly in emerging economies such as Morocco. This study aims to assess the impact of behavioral biases on these decisions, an area little explored in the existing financial literature. The main objective is to analyze how behavioral biases such as overconfidence, risk aversion, confirmation bias, anchoring, and managerial myopia biases influence bank financing decisions of Moroccan SMEs. The approach adopted is quantitative and uses robust least squares regression to analyze data collected from 167 Moroccan SMEs. The results reveal that overconfidence and anchoring have a significant positive impact on the propensity to take out bank loans, while risk aversion and confirmation bias have a negative effect. Managerial myopia had no significant influence. Control variables such as past financial performance, the length of the banking relationship, and lower risk also positively influence the financing decision.

Multiple instance learning framework can facilitate explainability in murmur detection.

Cardiovascular diseases (CVDs) account for a high fatality rate worldwide. Heart murmurs can be detected from phonocardiograms (PCGs) and may indicate CVDs. Still, they are often overlooked as their detection and correct clinical interpretation require expert skills. In this work, we aim to predict the presence of murmurs and clinical outcomes from multiple PCG recordings employing an explainable multitask model. Our approach consists of a two-stage multitask model. In the first stage, we predict the murmur presence in single PCGs using a multiple instance learning (MIL) framework. MIL also allows us to derive sample-wise classifications (i.e. murmur locations) while only needing one annotation per recording ("weak label") during training. In the second stage, we fuse explainable hand-crafted features with features from a pooling-based artificial neural network (PANN) derived from the MIL framework. Finally, we predict the presence of murmurs and the clinical outcome for a single patient based on multiple recordings using a simple feed-forward neural network. We show qualitatively and quantitatively that the MIL approach yields useful features and can be used to detect murmurs on multiple time instances and may thus guide a practitioner through PCGs. We analyze the second stage of the model in terms of murmur classification and clinical outcome. We achieved a weighted accuracy of 0.714 and an outcome cost of 13612 when using the PANN model and demographic features on the CirCor dataset (hidden test set of the George B. Moody PhysioNet challenge 2022, team "Heart2Beat", rank 12 / 40). To the best of our knowledge, we are the first to demonstrate the usefulness of MIL in PCG classification. Also, we showcase how the explainability of the model can be analyzed quantitatively, thus avoiding confirmation bias inherent to many post-hoc methods. Finally, our overall results demonstrate the merit of employing MIL combined with handcrafted features for the generation of explainable features as well as for a competitive classification performance.

Challenges in Millennials Investment Decision: A Study of Behavioural Biases

Aims: This research aimed to analyze the effect of Cognitive Dissonance Bias, Overconfidence Bias, Herding Bias, Endowment Bias, and Confirmation Bias on the investment decisions of the millennial generation in the capital market. Study Design: The sampling method used in this study was purposive sampling, which obtained 128 respondents. Place and Duration of Study: The research was conducted with investors in Banyumas Regency. Methodology: This research method uses the SEM (Structural Equation Modeling) analysis method with the Partial Least Square (PLS) approach. Each hypothesis is tested to understand the relationship between variables. To test the validity and reliability of research using an outer model. Hypothesis testing uses inner models. Results: The results of this study showed that cognitive dissonance bias, overconfidence bias, and endowment bias have an effect on investment decisions. However, herding bias and confirmation bias do not affect investment decisions. Conclusion: To increase investment opportunities, investors must pay attention to cognitive dissonance bias, overconfidence bias, and endowment bias because they can cause investment failure. Investors must also pay attention to information circulating in the media because analysis needs the latest information on investment transaction targets to avoid investment failure.

Integrating Pre-test Probability and Point-of-Care Ultrasound (POCUS) in the Emergency Department (ED) Diagnosis of Small Bowel Obstruction (SBO).

Introduction A computed tomography (CT) scan and point-of-care ultrasound (POCUS) are commonly employed for diagnosing small bowel obstructions (SBOs). Prior studies demonstrated that POCUS has 90-95% sensitivity and specificity compared with CT scanning, which is the gold standard. Unlike other imaging modalities (in which the ordering and performing clinician are not the same), POCUS-performing/interpreting sonologists must recognize the risk of confirmation bias in the POCUS application. Per Bayesian analysis, the likelihood of a diagnosis being true following a diagnostic test is based on the ordering clinician's pre-test probability and the test characteristics (sensitivity and specificity, from which positive and negative likelihood ratios can be calculated). Consequently, establishing pre-test probability is important in informing downstream diagnostic or therapeutic interventions, as pre-test probability influences post-test odds. Little research has been done on the role of POCUS sonologist's pre-test probability and actual POCUS results regarding SBO. This study assessed the role of POCUS, integrating pre-test probability and POCUS results to determine post-test odds. Methods One hundred six patients were recruited on a convenience basis and underwent POCUS and CT between April 2017 and December 2022. All sonographers were credentialed in POCUS. POCUS sonologists' pre-test probabilities and POCUS and CT results were captured, which were compared. Sensitivity, specificity, LR+, and LR- were calculated, and correlations were made between pre-test probability and POCUS and CT results. Results POCUS exhibited a sensitivity of 92% and specificity of 90%, with a corresponding positive likelihood ratio (LR+) of 9.3 and a negative likelihood ratio (LR-) of 0.09 for diagnosing SBO. Among patients with a high pre-test probability of SBO, a negative ultrasound yielded post-test odds of 0.4%, whereas a positive POCUS yielded post-test odds of 39.6%. Among patients with a low pre-test probability, a negative POCUS resulted in post-test odds of 0%, while a positive POCUS led to post-test odds of 2.1%, yielding a number needed to scan (NNS) of ~50 to identify a patient with an SBO on CT. Conclusion This study confirmed POCUS's sensitivity and specificity of ~90-95% and a corresponding LR+ of 9.2 and LR- of 0.9. Pre-test probability substantially affected post-test odds. Patients with a high pre-test probability and a positive POCUS had post-test odds of 39.6 and should have a confirmatory CT, while those with a negative POCUS have very low post-test odds and very likely will not benefit from CT. Patients with low pre-test probability and a positive POCUS have post-test odds of 2.1%, similar to the Wells Score and HEART score; such patients may not benefit from a CT, though clinicians should use their judgment/discretion. Patients with a low pre-test probability and a negative POCUS have post-test odds of 0% and should not have a CT. Among low pre-test probability patients, the NNS was ~50 to identify patients with an SBO on CT.

Nudges to Mitigate Confirmation Bias during Web Search on Debated Topics: Support vs. Manipulation

When people use web search engines to find information on debated topics, the search results they encounter can influence opinion formation and practical decision-making with potentially far-reaching consequences for the individual and society. However, current web search engines lack support for information-seeking strategies that enable responsible opinion formation, e.g., by mitigating confirmation bias and motivating engagement with diverse viewpoints. We conducted two preregistered user studies to test the benefits and risks of an intervention aimed at confirmation bias mitigation. In the first study, we tested the effect of warning labels, warning of the risk of confirmation bias, combined with obfuscations, hiding selected search results per default. We observed that obfuscations with warning labels effectively reduce engagement with search results. These initial findings did not allow conclusions about the extent to which the reduced engagement was caused by the warning label (reflective nudging element) versus the obfuscation (automatic nudging element). If obfuscation was the primary cause, this would raise concerns about harming user autonomy. We thus conducted a follow-up study to test the effect of warning labels and obfuscations separately. According to our findings, obfuscations run the risk of manipulating behavior instead of guiding it, while warning labels without obfuscations (purely reflective) do not exhaust processing capacities but encourage users to actively choose to decrease engagement with attitude-confirming search results. Therefore, given the risks and unclear benefits of obfuscations and potentially other automatic nudging elements to guide engagement with information, we call for prioritizing interventions that aim to enhance human cognitive skills and agency instead.

The continued influence of implied and explicitly stated misinformation with different emotional valence

ABSTRACT Misinformation often affects people’s cognition and judgment even when they are aware of a retraction; this is known as the continued influence effect of misinformation (CIE). The aim of the present study was to verify if the expressive style affects the CIE with different emotional valence. We designed a 2 (emotional valence: positive or negative) * 2(expressive style: explicit or implied) * 2 (correction condition: uncorrected or corrected) within-subject experiment. Participants were asked to finish memory questions and inference questions. The results found an interaction between expressive style and emotional valence. For positive misinformation, the expressive style had no effect; for negative misinformation, the CIE was stronger in the explicitly stated condition compared with the implied condition. The explicitly stated misinformation was more difficult to correct than the implied misinformation. The results provided evidence to explain the role of emotional valence and expressive style on CIE.

A Comprehensive Review of Impacts of Social Media and Information Technology on Decision Making of General People

Social media platforms and information technology have revolutionized the way individuals obtain, use, and share information in the digital age. This has affected people's decision-making processes in a variety of fields. This thorough analysis investigates the complex effects of information technology and social media on the way the public makes decisions. The impact of social media and information technology on decision-making in domains like politics, healthcare, education, consumer behavior, and interpersonal relationships is examined in this review, which draws on an extensive array of scholarly literature, empirical investigations, and theoretical frameworks. According to the research, information technology and social media can influence decisions in both favorable and unfavorable ways. On the one hand, these platforms give users access to a multitude of information, empowering them to take part in social movements, activism, and civic engagement with greater knowledge and understanding. Social media platforms also enable peer-to-peer communication, cooperation, and knowledge sharing, giving people the confidence to ask their social networks for guidance, affirmation, and support. In the digital age, the review emphasizes how psychological variables, cognitive biases, and sociocultural influences shape decision-making processes. In online environments, people's information processing, judgment-making, and decision-making processes are greatly influenced by social comparison, confirmation bias, FOMO, and social validation, among other factors. This thorough analysis advances our knowledge of the intricate interactions among decision-making, information technology, and social media. Informed decision making, digital literacy, and the moral application of social media and information technology in society are all goals of this review, which synthesizes the literature and highlights new trends to guide future research, legislative efforts, and hands-on interventions.

Scrolling through fake news: The effect of presentation order on misinformation retention.

Sharing information in real time leaves little room for double-checking. This leads to an abundance of low-quality information that might later need to be corrected and provides a foundation on which false beliefs can arise. Today, the general population often consults digital media platforms for news content. Because of the sheer amount of news articles and the various ways digital media platforms organize material, readers may encounter news articles with faulty content and their subsequent corrections in various orders. They might read the misinformation before the corrected version or vice versa. We conducted two studies in which participants were presented with two reports of a news event: one report that included a piece of misinformation and one report in which that misinformation was retracted. The order in which the two reports were encountered was manipulated. In Study 1, the retraction contained an explicit reminder of the misinformation; in Study 2, it did not. Neither Study 1 nor Study 2 found an effect of presentation order on misinformation reliance. These findings run counter to predictions by those accounts of the continued influence effect that suggest a better encoding of retractions and subsequent lesser reliance on misinformation when retractions are encountered first. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

RAgE: Robust Age Estimation Through Subject Anchoring With Consistency Regularisation.

Modern facial age estimation systems can achieve high accuracy when training and test datasets are identically distributed and captured under similar conditions. However, domain shifts in data, encountered in practice, lead to a sharp drop in accuracy of most existing age estimation algorithms. In this article, we propose a novel method, namely RAgE, to improve the robustness and reduce the uncertainty of age estimates by leveraging unlabelled data through a subject anchoring strategy and a novel consistency regularisation term. First, we propose an similarity-preserving pseudo-labelling algorithm by which the model generates pseudo-labels for a cohort of unlabelled images belonging to the same subject, while taking into account the similarity among age labels. In order to improve the robustness of the system, a consistency regularisation term is then used to simultaneously encourage the model to produce invariant outputs for the images in the cohort with respect to an anchor image. We propose a novel consistency regularisation term the noise-tolerant property of which effectively mitigates the so-called confirmation bias caused by incorrect pseudo-labels. Experiments on multiple benchmark ageing datasets demonstrate substantial improvements over the state-of-the-art methods and robustness to confounding external factors, including subject's head pose, illumination variation and appearance of expression in the face image.

품질이 관리된 스트레스 측정용 데이터셋 구축을 위한 제언

The construction of a stress measurement dataset plays a crucial role in various modern applications. In particular, for the efficient training of artificial intelligence models for stress measurement, it is essential to compare various biases and construct a quality-controlled dataset. In this paper, we propose the construction of a stress measurement dataset with quality management through the comparison of various biases. To achieve this, we introduce stress definitions and measurement tools, the process of building an artificial intelligence stress dataset, strategies to overcome biases for quality improvement, and considerations for stress data collection. Specifically, to manage dataset quality, we discuss various biases such as selection bias, measurement bias, causal bias, confirmation bias, and artificial intelligence bias that may arise during stress data collection. Through this paper, we aim to systematically understand considerations for stress data collection and various biases that may occur during the construction of a stress dataset, contributing to the construction of a dataset with guaranteed quality by overcoming these biases.

Bringing Sustainable Development to the Corporate Level: Boards’ Cognitive Biases towards ESG and Relevant Debiasing Interventions

The proposed research is an attempt to view how internationally recognised principles of sustainable development can be implemented at the corporate level, in particular, how the company’s board of directors should respond to climate and environmental challenges. Specifically, the paper examines the problem of reducing board members' repeated cognitive biases towards such a mainstream governance concept as ESG (Environmental, Social and Corporate Governance). The research focuses on the five most well-known cognitive biases of a board towards ESG, often encountered and described conceptually in behavioural literature: over-optimism, status quo effect, confirmation bias, hyperbolic discounting, and groupthink. Subsequently, the paper draws recommendations for overcoming or mitigating these biases by introducing and describing several debiasing interventions and quality control tools. These tools are conditionally divided into three groups: changing incentives, optimising choice architecture, and debiasing training. The author believes that incorporating behavioural insights into an analysis of the board’s ESG commitment would bring a fresh perspective to this research area, which now is typically dominated by a rational choice framework. The paper combines theoretical analysis of the existing literature in law, behavioural economics and psychology with qualitative research application based on data obtained from first-hand observations, board empirical studies, questionnaires, and focus groups.

THE INFLUENCE OF FINANCIAL CONFIDENCE, HERDING BIAS, AND CONFIRMATION BIAS ON INVESTMENT DECISIONS IN GENERATION Z IN JAKARTA

Interest in investing in the capital market for young investors is increasing. This is reflected in the increasing number of young investors in the capital market. Internet and digital technology have become part of today's society in all aspects. Technological developments in the financial services industry are referred to as Fintech (Financial Technology). Fintech is present starting from as a means of payment, E-Wallet, digital bank, to investment provider brokers. The purpose of this study is to examine and analyse the factors that influence Financial Confidence, Herding Bias, and Confirmation Bias on Generation Z's Investment Decision in Jakarta. The sample in the study were investors born between 1995-2010 called Generation Z and residing in Jakarta. The purposive sampling method was used to collect 205 respondents through Google Forms shared online on social media. The research data was analysed with a structural equation system using SmartPLS.3.2.9 software. The results showed, Financial confidence has a significant effect on investment decision making, Herding Bias has a significant effect on investment decision making, Confirmation bias has no effect on investment decision making.

The management of bias and noise in public sector decision-making: experimental evidence from healthcare

ABSTRACT In six randomized online experiments with 2,647 medical doctors we test whether – depending on the choice architecture – physicians engaged in prescribing decisions in public organizations fall prey to systematic error (i.e. bias) and make significantly different choices when faced with the same clinical case (i.e. level noise). Results show that experts tend to make irrational choices that are influenced by outgroup bias, social comparison, past behaviour, confirmation bias, loss aversion, equivalence framing, and asymmetric dominance. We also find evidence of significant level noise, that is, between-prescriber variability, with the distribution of responses differing remarkably across experimental arms.

The roots of polarization in the individual reward system.

Polarization raises concerns for democracy and society, which have expanded in the internet era where (mis)information has become ubiquitous, its transmission faster than ever, and the freedom and means of opinion expressions are expanding. The origin of polarization however remains unclear, with multiple social and emotional factors and individual reasoning biases likely to explain its current forms. In the present work, we adopt a principled approach and show that polarization tendencies can take root in biased reward processing of new information in favour of choice confirmatory evidence. Through agent-based simulations, we show that confirmation bias in individual learning is an independent mechanism and could be sufficient for creating polarization at group level independently of any additional assumptions about the opinions themselves, a priori beliefs about them, information transmission mechanisms or the structure of social relationship between individuals. This generative process can interact with polarization mechanisms described elsewhere, but constitutes an entrenched biological tendency that helps explain the extraordinary resilience of polarization against mitigating efforts such as dramatic informational change in the environment.