Confirmation Bias Research Articles

Correcting vaccine misinformation: A failure to replicate familiarity or fear-driven backfire effects.

Individuals often continue to rely on misinformation in their reasoning and decision making even after it has been corrected. This is known as the continued influence effect, and one of its presumed drivers is misinformation familiarity. As continued influence can promote misguided or unsafe behaviours, it is important to find ways to minimize the effect by designing more effective corrections. It has been argued that correction effectiveness is reduced if the correction repeats the to-be-debunked misinformation, thereby boosting its familiarity. Some have even suggested that this familiarity boost may cause a correction to inadvertently increase subsequent misinformation reliance; a phenomenon termed the familiarity backfire effect. A study by Pluviano et al. (2017) found evidence for this phenomenon using vaccine-related stimuli. The authors found that repeating vaccine "myths" and contrasting them with corresponding facts backfired relative to a control condition, ironically increasing false vaccine beliefs. The present study sought to replicate and extend this study. We included four conditions from the original Pluviano et al. study: the myths vs. facts, a visual infographic, a fear appeal, and a control condition. The present study also added a "myths-only" condition, which simply repeated false claims and labelled them as false; theoretically, this condition should be most likely to produce familiarity backfire. Participants received vaccine-myth corrections and were tested immediately post-correction, and again after a seven-day delay. We found that the myths vs. facts condition reduced vaccine misconceptions. None of the conditions increased vaccine misconceptions relative to control at either timepoint, or relative to a pre-intervention baseline; thus, no backfire effects were observed. This failure to replicate adds to the mounting evidence against familiarity backfire effects and has implications for vaccination communications and the design of debunking interventions.

Can you trust your lawyer's call? Legal advisers exhibit myside bias resistant to debiasing interventions

AbstractIn a vast majority of disputes, settlement is superior to litigation, which involves uncertainty, legal fees, and opportunity cost. Unnecessary litigation also causes judicial backlog, wastes resources, and increases societal conflict. Major contributors to the lack of settlement are intransigent litigants who harbor overoptimistic predictions of litigation outcomes, even though they are looking at identical facts and applicable law. A study (N = 166) found significant myside bias in the participants' predictions of a judicial award (claimants' advisers expected awards that were 69% higher than defendants' advisers) and in their evaluation of arguments (both sides thought the arguments supporting their side were 30% more convincing than the arguments supporting their counterparty). Debiasing interventions—alerting to the myside bias, considering the perspective of the counterparty and dialectical bootstrapping—reduced the bias but did not eliminate it. Exploratory investigation indicated that a large proportion of advisers exhibited naïve realism and bias blind spot, and that cognitive reflection provided a limited measure of resistance to myside bias.

Time Well Spent: Exploring the Role of Attitude and Topic Importance on Selective Exposure to Agricultural Messages

New digital technologies, such as Web 3.0 and algorithms, allow social media users to customize their feeds, creating their own information bubble, which tends to align with prior beliefs and/or attitude. This action of seeking information that emphasizes or confirms pre-existing beliefs is called confirmation bias, which is often expressed through selective exposure. Although previous studies have explored selective exposure in the context of political and health communications, limited research has been completed related to this phenomenon in agricultural communications. Therefore, the purpose of this study was to explore the effects of attitude and topic importance on selective exposure to different agricultural messages. Using a quasi-experimental design, this study used a Qualtrics questionnaire to collect data from undergraduate students in a laboratory setting. Participants provided their attitudes and topic important for two agricultural issues. A fictional Twitter feed was created that linked to four blog posts that served as the message stimuli. To determine selective exposure, we recorded how many blog posts they selected and how long they spent on each message. The results indicated that participants had varying attitudes of the two agricultural topics but had equal views of importance. No significant difference in time spent on the messages on was found. The results indicated that the process of selective exposure is a complex construct that involves many factors. Additional research in this area will help agricultural communicators develop more effective message strategies and understand the role of confirmation bias in information processing.

Weakening the ideological immune system: can debiasing techniques reduce confirmation bias?

ABSTRACT Psychologists have identified confirmation bias as a contributor to conflict and ideological extremism. However, the efficacy of social science’s existing arsenal of debiasing techniques remains largely unknown, and the extant positive findings have often been inconsistent and are in need of independent replication. In this registered report, we empirically tested the efficacy of three cognitive debiasing techniques in mitigating confirmation bias: consider-the-opposite, psychoeducation about bias, and social norms manipulation. In the control group, participants exhibited two manifestations of confirmation bias: selective exposure and biased assimilation. When it came to reducing these biases, we found evidence that the social norms technique reduced selective exposure relative to the control group, but little evidence that this was the case for the consider-the-opposite or psychoeducation techniques was found. None of the debiasing strategies significantly reduced biased assimilation relative to the control group. Comparing these strategies provides theoretical advancement in the processes that give rise to confirmation bias, and moreover, these results help to identify a promising technique to serve as the foundation for debiasing interventions, namely social norms interventions.

Executive function and the continued influence of misinformation: A latent-variable analysis.

Misinformation can continue to influence reasoning after correction; this is known as the continued influence effect (CIE). Theoretical accounts of the CIE suggest failure of two cognitive processes to be causal, namely memory updating and suppression of misinformation reliance. Both processes can also be conceptualised as subcomponents of contemporary executive function (EF) models; specifically, working-memory updating and prepotent-response inhibition. EF may thus predict susceptibility to the CIE. The current study investigated whether individual differences in EF could predict individual differences in CIE susceptibility. Participants completed several measures of EF subcomponents, including those of updating and inhibition, as well as set shifting, and a standard CIE task. The relationship between EF and CIE was then assessed using a correlation analysis of the EF and CIE measures, as well as structural equation modelling of the EF-subcomponent latent variable and CIE latent variable. Results showed that EF can predict susceptibility to the CIE, especially the factor of working-memory updating. These results further our understanding of the CIE's cognitive antecedents and provide potential directions for real-world CIE intervention.

Moralization and extremism robustly amplify myside sharing.

We explored whether moralization and attitude extremity may amplify a preference to share politically congruent ("myside") partisan news and what types of targeted interventions may reduce this tendency. Across 12 online experiments (N = 6,989), we examined decisions to share news touching on the divisive issues of gun control, abortion, gender and racial equality, and immigration. Myside sharing was systematically observed and was consistently amplified when participants (i) moralized and (ii) were attitudinally extreme on the issue. The amplification of myside sharing by moralization also frequently occurred above and beyond that of attitude extremity. These effects generalized to both true and fake partisan news. We then examined a number of interventions meant to curb myside sharing by manipulating (i) the audience to which people imagined sharing partisan news (political friends vs. foes), (ii) the anonymity of the account used (anonymous vs. personal), (iii) a message warning against the myside bias, and (iv) a message warning against the reputational costs of sharing "mysided" fake news coupled with an interactive rating task. While some of those manipulations slightly decreased sharing in general and/or the size of myside sharing, the amplification of myside sharing by moral attitudes was consistently robust to these interventions. Our findings regarding the robust exaggeration of selective communication by morality and extremism offer important insights into belief polarization and the spread of partisan and false information online.

Vulnerability to Disinformation in Relation to Political Affiliation in North Macedonia

This study aims to analyze the relationship between political affiliation and vulnerability to disinformation in North Macedonia through the role of psychological and social constraints in shaping how individuals respond to and process information. Research has shown that politically affiliated individuals may be particularly vulnerable to disinformation in part due to confirmation bias or the tendency to accept and seek out information that is consistent with one’s preexisting beliefs and ignore or refute information that is not. Using the quantitative method and cross-matched data from the empirical research, the study has shown that political affiliation affects the way individuals perceive disinformation. Correspondingly, disinformation with a negative connotation from one’s affiliated political party is perceived by a lower percentage as accurate, contrary to disinformation with a negative connotation from the opposing political party, which is perceived by a higher percentage as accurate. The study also found that politically affiliated individuals are more prone to disinformation than those who are not politically affiliated. The results suggest that political affiliation plays a significant role in an individual’s vulnerability to disinformation.

Implications of zero party data on user decision-making in digital advertising

Today’s digital advertising is predominantly founded on third party data such as cookies, which track the browsing behavior history. However, the data collected by cookies is incomplete, it merely reflects past browsing behavior instead of providing information on current interests and specific consumption desires.
 Due to an increased need of privacy, Google announced the disappearance of third party cookies. This means that websites will not be allowed to collect user data through cookies anymore in order to share it with third parties. As a result, companies will not be able to rely on user data gathered through external websites.
 On the other hand, zero party data allows companies to collect direct information from its users. Through the means of intentionally provided data by users (such as user profiles, forms and surveys), digital advertisers could create more pertinent advertising content without relying on third party cookies.
 When subject to digital advertising, several biases play a role in the decision making process. This work intends to assess the impact of digital advertising using zero party data, by analyzing the potential decrease in discrepancy between rational consumption intent and actual consumption considering behavioral biases. 
 Browsing is defined as the “act of looking through a set of information quickly, without a specific sense of purpose”[1]. In Thinking Fast and Slow, Kahneman, D & Tvesky, A (2011) explain how the thinking system 1 uses past decision-making experience to generate a fast response, which is often influenced by a variety of behaviour biases. Therefore, browsing information collected by third party cookies results from system 1. 
 Zero party data primarily relies on system 2. This thinking system works in a slower manner as it requires effort to lead to logical conclusions. It is used for instance to resolve hard problems and evaluate pros and cons. When answering a complex survey on personal goals, or completing a form requiring conscious attention, then, system 2 is used. As zero party data consists of more meaningful and intentionally provided information compared to third party data, digital advertisement based on zero party data has a higher likelihood that rational consumption intent and actual consumption are in line.
 The methodology conducted consists of research through two surveys (which served as A/B test) and interviews to suggest an alternative approach to third party cookies. Through the surveys and interviews, eight biases and heuristics present in the user journey were identified; the paradox of choice, the availability bias, the confirmation bias, the consistency bias, the bandwagon effect, the anchoring bias, the Ikea effect, and the affect heuristic. Each section aims to analyze how these biases affect the online consumption process of a user in both scenarios of third party data and zero party data. Each outcome revolves around how to use the identified bias in the benefit of the user using zero party data.
 Digital advertisement based on zero party data is a win-win solution for the user and the advertisement company.

Protecting democracy from disinformation: Implications for a model of communication

This article analyses the consequences that disinformation phenomena have for a model of communication, focusing on the dangers that disinformation poses to democratic societies, especially when it is disseminated by the media. Disinformation is examined here from the perspective of social cognitive psychology, with special attention to the role played by motivated reasoning and confirmation bias in human cognition. From this perspective, disinformation phenomena should be studied not only through an analysis of how the media operate, but also through an understanding of how we process information and what we use it for from a social cognitive point of view. This article emphasizes the role that intuition and affective persuasion play in communication processes, as key elements of motivated reasoning, and argues that once this cognitive dimension is integrated into communication theory, preventive strategies can be designed to protect democracies from the dangers caused by disinformation. Ideological polarization and a lack of consensus are highlighted here as being among the biggest dangers, preventing agreement on issues that affect the proper functioning of democracy. While a certain conception of communication posits reasoning, the media and education as the tools for resolving conflicts and preventing disagreements, this article concludes that the success of disinformation phenomena points to the need for a model that includes the cognitive elements mentioned above.

조세불복사건 판단에 있어 인지 편향과 극복방안

Humans acquired heuristics for survival. Heuristics are mental shortcuts that save time by simplifying complex decisions into simple rules of thumb. Heuristics are useful in everyday life because they enable us to automatically make quick decisions with little effort. However, when making important decisions and legal judgments, heuristics can be dangerous. Before filing a lawsuit in an appeals proceeding against a taxation, it must be needed to file an administrative appeal. It is the procedure enforced by law. Administrative hearings is similar to the judicial proceedings. The evaluating of the weight of proof depends on the committee members just as the judge. The problem is that the judge could make wrong decision due to his or her cognitive bias. In the United States, many experimental studies have been conducted on jurors and judges to see if various heuristics actually occur. It is important for parties to accept the legal decision, so that the legal decision must not be biased.
 The purpose of this research is to analyze the cases that had been shown heuristics in administrative trial decisions and to research on debiasing. This paper analyzes decisions in which representativeness heuristics, hindsight bias, and confirmation bias appeared. In order to debiase, there are several trials like the cognitive correction(thinking the opposite etc.), environmental correction(reducing cognitive losses etc.), and systemic correction(decision support systems etc.).

Mechanisms in continued influence: The impact of misinformation corrections on source perceptions.

Research on the continued influence effect (CIE) of misinformation has demonstrated that misinformation continues to influence people's beliefs and judgments even after it has been corrected. Although most theorizing about the CIE attempts to explain why corrections do not eliminate belief in and influences of the misinformation, the present research takes a different approach and focuses instead on why corrections do reduce belief in misinformation (even if not entirely). We examined how a correction can change perceptions of the original source of the misinformation and how these changes in perceptions can mediate continued influence effects. We also examined causal evidence linking manipulations of misinformation source perceptions to continued belief and misinformation-relevant inferential reasoning. Study 1 demonstrated that an external correction (i.e., a new source labeling misinformation as false) influences perceptions of the misinformation source, and these perceptions of the misinformation source then correlated with belief in the misinformation. Study 2 replicated the results of Study 1 and used source derogation to manipulate misinformation source perceptions and further lessen continued belief. Study 3 was a preregistered replication of previous results using new methodology. These studies suggest that perceptions of the misinformation source is one mechanism that can cause changes in belief in misinformation, and changes in the perception of a source can be achieved simply by correcting the source or through other means. This approach can be used to find other mechanisms responsible for reducing belief in misinformation.

Associations between emotions and psychophysiological states and confirmation bias in question formulation in ongoing simulated investigative interviews of child sexual abuse.

In forensic settings interviewers are advised to ask as many open-ended questions as possible. However, even experts may have difficulty following this advice potentially negatively impacting an investigation. Here, we sought to investigate how emotions and psychophysiological parameters are associated with question formulation in real time in an ongoing (simulated) child sexual abuse (CSA) interview. In a experimental study, psychology students (N = 60, Mage = 22.75) conducted two interviews with child avatars, while their emotions (anger, sadness, disgust, surprise and relief), GSR and heart rate (HR) were registered. First, we found that general emotionality related to CSA and perceived realness of the avatars was associated with stronger overall emotional reactions. Second, we found that closed (vs. open) questions were preceded by more facially observable anger, but not disgust, sadness, surprise or relief. Third, closed (vs. open) questions were preceded by higher GSR resistance and lower heart rate. Results suggest for the first time that emotions and psychophysiological states can drive confirmation bias in question formulation in real time in CSA.

The Complexity Paper: A Writing Assignment that Targets Cognitive Bias

Cognitive bias, especially confirmation bias and motivated reasoning, poses a significant challenge to argumentative writing genres, including the persuasive essay. To address this challenge, I introduce the complexity paper. Rather than attempting to convince the reader of a particular position on an issue as with a persuasive essay, the goal of a complexity paper is to convince the reader that the issue under discussion is difficult to resolve. This altered motivational structure encourages the writer to engage in perspective taking, thereby addressing confirmation bias and motivated reasoning. I outline a three-part structure of the complexity paper using the example of physician-assisted suicide with the goal to help university instructors implement this genre into their courses. I recommend incorporating a complexity paper into a writing scaffold structure that precedes an argumentative paper.

Rigour and reproducibility in perinatal and paediatric epidemiologic research using big data.

Rigour and reproducibility in perinatal and paediatric epidemiologic research using big data.

Crisis alert: (Dis)information selection and sharing in the COVID-19 pandemic

Abstract High levels of threat and uncertainty characterize the onset of societal crises. Here, people are exposed to conflicting information in the media, including disinformation. Because individuals often base their news selection on pre-existing attitudes, the present study aims to examine selective exposure effects in the face of a crisis, and identify right-wing ideological, trust-, and science-related beliefs that might influence the selection and sharing of disinformation. A representative survey of German internet users (N = 1101) at the time of the COVID-19 pandemic outbreak showed a confirmation bias in crisis-related news selection and sharing. It revealed right-wing authoritarianism and political mistrust as significant predictors of disinformation selection. The influence of social dominance orientation, mistrust in politics, and perceived certainty of knowledge were significant for sharing disinformation. The present results extend previous knowledge about people’s (dis)information behavior in times of crisis, and shed light on groups particularly vulnerable to disinformation.

Observations of Player (de)Selection Within a Professional UK Soccer Academy.

The present study engaged in an ethnographical observation of the processes used to determine player (de)selections within a professional academy. English category-2 youth academy players (n = 96) from U10-U16 age groups undertook anthropometric profiling (height, mass and somatic maturation) and fitness assessments (10m, 20m & 30m linear sprints, 505-agility test, countermovement and squat jumps). Each players lead coach (n = 4) subjectively graded players utilising a red, amber and green (RAG) rating system on a weekly (current performance) and quarterly (perceived potential) basis, across 25weeks. A MANCOVA, controlling for maturation, was applied to determine differences in (de)selection by physical performance. Mann Whitney-U tests were used to distinguish difference in (de)selection by subjective grading (weekly and quarterly). The key finding was that quarterly subjective gradings established a higher cumulative score of green ratings in selected players and a low cumulative score of red ratings, and vice versa for deselected players (P ≤ 0.001 to 0.03). However, whilst these findings suggest that quarterly subjective grades of potential were able to provide the best predictors for player (de)selection, the findings should be viewed with caution due to high potential for confirmatory bias.

A Formal Model for Polarization under Confirmation Bias in Social Networks

We describe a model for polarization in multi-agent systems based on Esteban and Ray's standard family of polarization measures from economics. Agents evolve by updating their beliefs (opinions) based on an underlying influence graph, as in the standard DeGroot model for social learning, but under a confirmation bias; i.e., a discounting of opinions of agents with dissimilar views. We show that even under this bias polarization eventually vanishes (converges to zero) if the influence graph is strongly-connected. If the influence graph is a regular symmetric circulation, we determine the unique belief value to which all agents converge. Our more insightful result establishes that, under some natural assumptions, if polarization does not eventually vanish then either there is a disconnected subgroup of agents, or some agent influences others more than she is influenced. We also prove that polarization does not necessarily vanish in weakly-connected graphs under confirmation bias. Furthermore, we show how our model relates to the classic DeGroot model for social learning. We illustrate our model with several simulations of a running example about polarization over vaccines and of other case studies. The theoretical results and simulations will provide insight into the phenomenon of polarization.

Adversarial co-training for semantic segmentation over medical images

Background and objective:Abundant labeled data drives the model training for better performance, but collecting sufficient labels is still challenging. To alleviate the pressure of label collection, semi-supervised learning merges unlabeled data into training process. However, the joining of unlabeled data (e.g., data from different hospitals with different acquisition parameters) will change the original distribution. Such a distribution shift leads to a perturbation in the training process, potentially leading to a confirmation bias. In this paper, we investigate distribution shift and develop methods to increase the robustness of our models, with the goal of improving performance in semi-supervised semantic segmentation of medical images. We study distribution shift and increase model robustness to it, for improving practical performance in semi-supervised segmentation over medical images. Methods:To alleviate the issue of distribution shift, we introduce adversarial training into the co-training process. We simulate perturbations caused by the distribution shift via adversarial perturbations and introduce the adversarial perturbation to attack the supervised training to improve the robustness against the distribution shift. Benefiting from label guidance, supervised training does not collapse under adversarial attacks. For co-training, two sub-models are trained from two views (over two disjoint subsets of the dataset) to extract different kinds of knowledge independently. Co-training outperforms single-model by integrating both views of knowledge to avoid confirmation bias. Results:For practicality, we conduct extensive experiments on challenging medical datasets. Experimental results show desirable improvements to state-of-the-art counterparts (Yu and Wang, 2019; Peng et al., 2020; Perone et al., 2019). We achieve a DSC score of 87.37% with only 20% of labels on the ACDC dataset, almost same to using 100% of labels. On the SCGM dataset with more distribution shift, we achieve a DSC score of 78.65% with 6.5% of labels, surpassing 10.30% over Peng et al. (2020). Our evaluative results show superior robustness against distribution shifts in medical scenarios. Conclusion:Empirical results show the effectiveness of our work for handling distribution shift in medical scenarios.

Expl(AI)ned: The Impact of Explainable Artificial Intelligence on Users’ Information Processing

Although future regulations increasingly advocate that AI applications must be interpretable by users, we know little about how such explainability can affect human information processing. By conducting two experimental studies, we help to fill this gap. We show that explanations pave the way for AI systems to reshape users' understanding of the world around them. Specifically, state-of-the-art explainability methods evoke mental model adjustments that are subject to confirmation bias, allowing misconceptions and mental errors to persist and even accumulate. Moreover, mental model adjustments create spillover effects that alter users' behavior in related but distinct domains where they do not have access to an AI system. These spillover effects of mental model adjustments risk manipulating user behavior, promoting discriminatory biases, and biasing decision making. The reported findings serve as a warning that the indiscriminate use of modern explainability methods as an isolated measure to address AI systems' black-box problems can lead to unintended, unforeseen problems because it creates a new channel through which AI systems can influence human behavior in various domains.

Confirmation Bias in Seeking Climate Information: Employing Relative Search Volume to Predict Partisan Climate Opinions

In an increasingly digitized world, online information-seeking (OIS) behaviors have reflected people’s intentions and constituted a critical component in synthesizing public opinion. Climate change is among the gravest threats facing the world today, and previous studies have adopted OIS data to gauge public interest in climate change. However, such studies have ignored the psychological attributes of search keywords and the role of social identities in influencing OIS. This study explores whether search strategies align with the expected confirmation biases of regions with different partisan beliefs. We use spatial web search trends to show the significant differences in the search keywords adopted by the Democrat-majority (“climate change”) versus the Republican-majority (“global warming”) regions of the United States. Furthermore, using the region-level search and survey data (2008–2018), we demonstrate that the preferential use of search keywords can predict climate opinions. This study concludes by discussing the significant findings and the open questions for future work.