The study, led by Hirono Ishikawa of Teikyo University Graduate School of Public Health together with Mio Kato of the National Institute of Infectious Diseases, Satoru Kikuchi of Shinshu University, and Alan J. Christensen of East Carolina University, set out to accomplish two related goals. The first was to validate a Japanese-language version of the Irrational Health Belief Scale, or IHBS, an instrument originally developed in English to capture belief patterns that interfere with sound health decisions. The second was to distill the full twenty-item scale into a brief five-item short form that could be deployed efficiently in large-scale public health surveys, where every additional question carries a real cost in respondent fatigue and data quality.

The Irrational Health Belief Scale works in an unusual way. Rather than asking respondents directly whether they hold certain beliefs — a format that invites defensiveness and social desirability bias — it presents vivid everyday scenarios and asks people to rate how strongly they would endorse a particular automatic thought embedded in each one. One scenario describes coworkers coming down with flu amid a news report urging vaccination, paired with the thought that immunizations never do any good because a previous shot failed to prevent illness. Another describes a doctor prescribing a full course of medication, paired with the thought that the medicine becomes unnecessary once symptoms clear up. A third involves a family history of heart disease, paired with the fatalistic thought that whether one develops the disease is simply beyond personal control.

Each of these embedded thoughts reflects a recognizable cognitive distortion. Some involve hasty generalization, in which a single personal experience — one failed flu shot, one uneventful trip abroad without immunizations — is treated as decisive evidence against a well-established medical recommendation. Others involve normalizing by majority, the assumption that if everybody eats hamburgers and fries or skips seat belts, the behavior cannot be truly dangerous. Still others reflect catastrophizing, such as the belief that a lifetime of taking a well-tolerated medication means life will never be the same, or all-or-nothing reasoning, in which a treatment that cannot cure completely is judged not worthwhile at all. The scale’s designers treat these not as isolated quirks but as expressions of a broader cognitive style that filters all health information.

To test how this construct behaves in a Japanese population, the team conducted an online survey of 408 adults aged 20 to 69 years. Participants completed the Japanese IHBS alongside established measures of communicative and critical health literacy, health locus of control, and self-reported health behaviors. The analytic strategy followed standard psychometric practice in two stages. Exploratory factor analysis was used first to examine the underlying structure of the scale and to guide selection of items for the short form, with choices informed by factor loadings, conceptual coverage across the range of belief types, and the statistical distribution of responses to each item. Confirmatory factor analysis was then applied to the full twenty-item scale to evaluate whether a single-factor model — the assumption that all twenty items tap one coherent underlying disposition — fit the observed data adequately.

The results were strikingly consistent. Exploratory analysis indicated a largely unidimensional structure, and the confirmatory analysis of the one-factor model showed generally acceptable fit. In practical terms, this means that irrational health beliefs in this Japanese sample behave as a single measurable dimension rather than a loose collection of unrelated misconceptions. Someone who endorses the thought that a failed flu shot proves vaccines are useless is also, on average, more likely to believe that a doctor’s own weight discredits advice about diet, or that not yet feeling sick means immunity to a household outbreak of influenza. Internal consistency, assessed with Cronbach’s alpha, reached 0.91 for the full twenty-item scale — a level conventionally regarded as excellent — and 0.78 for the five-item short form, which falls within the acceptable range for a brief screening instrument. Critically, the short and long versions correlated at r = .90, meaning the five selected items recover nearly all of the information contained in the full scale.

The five items chosen for the short form illustrate the conceptual breadth the authors preserved despite the drastic reduction in length. They include the flu-vaccination scenario, the medication-abandonment scenario in which symptoms disappearing renders the remaining pills seem unnecessary, the fatalistic response to family history of heart disease, the belief that a new drug cannot help after a previous medication failed, and the catastrophizing reaction to a mild condition requiring lifelong treatment. Together, these items sample hasty generalization, premature discontinuation of treatment, perceived loss of control, therapeutic pessimism, and catastrophizing — the core cognitive patterns the scale was designed to detect.

Perhaps the most consequential findings concern how irrational health beliefs relate to other well-established psychological constructs. Higher scores on the IHBS were associated with lower health literacy, specifically the communicative and critical health literacy that reflects an individual’s capacity to extract, appraise, and apply health information from various sources. Higher scores also predicted a characteristic locus-of-control profile: lower internal health locus of control, meaning respondents were less likely to believe their own actions determine their health outcomes, and higher chance health locus of control, meaning they were more likely to attribute health to luck or fate. Finally, irrational health beliefs were linked to less favorable self-reported health behaviors overall, providing the criterion-related validity evidence that the scale is not merely measuring an abstract disposition but something connected to what people actually do.

The authors frame these relationships in terms of the persistent gap between health information and health action. Decades of health communication research have demonstrated that simply providing accurate, evidence-based messages does not reliably produce appropriate decisions or behaviors. The present study adds a specific cognitive mechanism to that account: individuals high in irrational health beliefs may receive the same message as everyone else but process it through filters that systematically distort its meaning. A vaccination recommendation becomes an opportunity to recall a personal exception; a prescription becomes a decision point governed by how one feels on a given day; a risk statistic becomes confirmation that the outcome was never controllable. In this view, the barrier is not ignorance but interpretation, which helps explain why information-heavy interventions so often disappoint.

What makes the new instrument potentially powerful is its brevity and its implication that the underlying beliefs may be modifiable. A five-item scale can be appended to national health surveys, workplace wellness assessments, or clinical intake questionnaires at negligible cost, allowing researchers and practitioners to identify populations or individuals who are most likely to misinterpret health information before designing interventions for them. If irrational health beliefs are a measurable cognitive determinant of behavior, they may also be a target: communication strategies could be tailored to preempt the specific distortions — the hasty generalization, the fatalism, the all-or-nothing judgment — rather than simply repeating the facts more loudly. The study, supported by JSPS KAKENHI grant 25K02857 and approved by the Ethics Review Committee of Teikyo University, was conducted entirely with Japanese adults, and the authors caution that cross-cultural validation in other populations remains an open task. But the core message travels well beyond Japan: the reason evidence so often fails to change minds about health may lie not in what people know, but in the irrational beliefs that quietly decide what their knowledge is allowed to mean.

Subject of Research: Development and validation of a Japanese short-form Irrational Health Belief Scale measuring cognitive barriers to health behaviors

Article Title: Irrational health beliefs as barriers to health behaviors: development and validation of a brief scale in a Japanese population

Article References: Ishikawa, H., Kato, M., Kikuchi, S., & Christensen, A. J. (2026). Irrational health beliefs as barriers to health behaviors: development and validation of a brief scale in a Japanese population. BMC Public Health. https://doi.org/10.1186/s12889-026-29822-8

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29822-8

Keywords: irrational health beliefs, health psychology, health literacy, health locus of control, psychometrics, scale validation, health communication, public health, cognitive bias, health behaviors, Japanese population, survey research

The story began when the teenager developed a progressively enlarging neck mass causing discomfort and airway compression. Magnetic resonance imaging showed a left lower neck multi-cystic lesion extending into the supraclavicular region and down into the mediastinum, findings read as consistent with a macrocystic lymphatic malformation. Local otolaryngologists attempted aspiration but obtained insufficient fluid to relieve the compression, prompting transfer out of state for specialized care at a vascular anomaly center. His past medical, family, and psychosocial history offered no red flags, with no known genetic syndromes and no prior malignancy. On the surface, everything pointed toward a benign vascular anomaly.

At the specialty center, evaluation confirmed left cervical cystic masses, the largest measuring seven by seven centimeters. The lesion was non-tender but firm, a detail that raised concern for intralesional hemorrhage. Then came the first cracks in the diagnostic narrative. Additional history revealed a forty-pound weight loss, vomiting, and early satiety over the preceding months, symptoms that are distinctly atypical for an isolated lymphatic malformation. On examination, the boy’s abdomen was markedly distended with a palpable mass spanning the left upper and lower quadrants. Full-body MRI and magnetic resonance angiography revealed a large multiloculated cystic abdominal mass, initially interpreted as additional sites of lymphatic malformation. The differential diagnosis included complex lymphatic anomaly, venous-lymphatic malformation, and germ cell tumor, with evolving findings progressively arguing against a vascular cause.

Interventional radiology was consulted for aspiration and possible sclerotherapy, the standard treatment for macrocystic lymphatic malformations. What the procedures yielded should have been a turning point. The abdominal fluid was brown-tinged and the cervical fluid was clear, both appearances atypical for lymph or chyle, the milky lymphatic fluid expected from a true lymphatic lesion. Sclerotherapy was attempted on the cervical lesion, but fluid analysis told a different story: a triglyceride level of just 13 milligrams per deciliter and a lymphocyte fraction of only 2 percent, values wholly inconsistent with lymphatic fluid. Further sclerotherapy attempts were discontinued. The treatment was failing not because the technique was flawed, but because the underlying diagnosis was wrong.

Multidisciplinary consensus led to surgical resection of the abdominal mass, a twelve-hour operation that removed 95 percent of a retroperitoneal tumor measuring 25 by 25 by 20 centimeters. The mass had both solid and cystic components and encased critical structures, including the aorta, inferior vena cava, kidneys, ureters, and iliac vessels. Pathology revealed a mixed germ cell tumor, approximately 98 percent immature teratoma, with additional mature teratomatous elements including squamous, neural, and gastrointestinal-type epithelium. Foci of embryonal carcinoma and areas consistent with yolk sac tumor were also identified. Tumor markers confirmed the diagnosis: alpha-fetoprotein was 1,240 nanograms per milliliter against a reference range of zero to eight, and human chorionic gonadotropin was 268 milli-international units per milliliter against a reference of zero to five. Following pathologic confirmation, the patient transitioned to oncology care and began standard bleomycin, etoposide, and cisplatin chemotherapy without dose modification or delay. Follow-up imaging showed regression of residual disease, tumor markers declined appropriately, and treatment was well tolerated without significant toxicity or interruptions.

The authors of the report argue that this case is a vivid illustration of diagnostic anchoring, the cognitive tendency to rely too heavily on initial information and interpret everything that follows in relation to that first reference point. Once the lymphatic malformation label was attached, subsequent findings were filtered through that lens. Confirmation bias compounded the problem: clinicians sought and favored information that supported the existing hypothesis while downplaying evidence that contradicted it. The firm texture of the mass, the systemic symptoms, the multifocal disease across non-contiguous anatomic compartments, the atypical aspirate characteristics, and the failure to respond to sclerotherapy were all signals that something did not fit. Each was initially absorbed into the vascular anomaly framework rather than triggering a fundamental reconsideration.

The report lays out explicit diagnostic checkpoints that should prompt clinicians to question a lymphatic malformation diagnosis. These include systemic symptoms such as weight loss or early satiety, which are atypical for isolated lymphatic disease; multifocal involvement across separate anatomic compartments; aspirate characteristics inconsistent with chyle or lymph; and lack of biochemical or clinical response to sclerotherapy. When an initial diagnosis fails to account for the full clinical picture, the authors contend, structured re-evaluation is essential to avoid diagnostic delay. In this case, the presence of gastrointestinal symptoms and dramatic weight loss deviated from the typical lymphatic malformation course, yet an atypical complex vascular anomaly with possible intestinal lymphangiectasia was initially considered before the team pivoted toward malignancy.

The stakes of such cognitive errors are far from abstract. Diagnostic errors account for approximately 75 percent of malpractice lawsuits and 38 percent of total payouts, and pediatricians report diagnostic errors affecting patient outcomes at least annually. Misdiagnosis remains a leading cause of avoidable harm and cost in medicine. Lymphatic malformations are typically diagnosed through history, examination, and imaging, but their radiologic features can be nonspecific and can be confused with teratomas, particularly when cystic morphology is interpreted without integrating systemic symptoms or treatment response. A limited number of prior reports describe teratomas mimicking vascular anomalies, but few address the diagnostic delay caused specifically by cognitive bias. Unlike cases diagnosed early through surgery, this case shows how diagnostic mimicry and specialty silos can delay recognition when no one actively re-evaluates the working diagnosis.

The authors emphasize that overcoming bias requires deliberate tools rather than good intentions alone. Metacognition, the practice of actively reflecting on one’s own clinical reasoning, allows clinicians to recognize when their thought processes may be drifting toward premature closure. Cognitive forcing strategies offer structured ways to break out of entrenched thinking: case-based discussions with peers, red team reviews inviting external expert critique, prospective hindsight that imagines future outcomes to anticipate pitfalls, the five whys technique for root-cause analysis, decision trees and flowcharts, deliberate consideration of contrary evidence, and simulation-based reviews. Clinicians should also avoid the pseudo-diagnostic effect, in which irrelevant data are used to shore up a diagnosis. The report suggests two questions worth asking whenever a diagnosis fails to explain the clinical course: What else could this be? And what doesn’t fit?

Ultimately, the case is a story of both failure and rescue. The initial reliance on typical referral pathways without questioning atypical features delayed definitive treatment, a limitation the authors acknowledge candidly. But clinical intuition, combined with a structured multidisciplinary approach drawing on radiology, interventional radiology, pediatric surgery, pathology, and oncology, broke through the anchoring. The authors argue that balancing intuitive and analytical thinking, while remaining vigilant for bias, is central to diagnostic excellence, and that medical education should emphasize debiasing strategies, guided reflection, and structured re-evaluation. Time pressures, sleep deprivation, and cognitive overload all work against consistent analytical reasoning, which is why they also point to technology such as clinical decision support systems and artificial intelligence tools as potential safeguards. Improving the diagnostic process, they conclude, is both a clinical and a moral imperative, and recognizing diagnostic mimicry while applying critical thinking throughout the clinical course can improve care in complex pediatric conditions such as mixed germ cell tumors and lymphatic malformations.

Subject of Research: Cognitive bias and diagnostic error in the diagnosis of a pediatric mixed germ cell tumor mimicking a lymphatic malformation

Article Title: Case Report: When bias clouds judgment in diagnosing a pediatric mixed germ cell tumor

Article References: Elsabbagh, E. M., Maloney, L., Kulungowski, A., Annam, A., Siegele, B., & Nakano, T. A. (2026). Case Report: When bias clouds judgment in diagnosing a pediatric mixed germ cell tumor. Frontiers in Cancer Control and Society, 4, Article 1638457. https://doi.org/10.3389/fcacs.2026.1638457

Image Credits: AI Generated

DOI: 10.3389/fcacs.2026.1638457

Keywords: cognitive bias, diagnostic error, anchoring bias, confirmation bias, pediatric germ cell tumor, immature teratoma, lymphatic malformation, diagnostic mimicry, interventional radiology, metacognition, pediatric oncology, multidisciplinary diagnosis

The research team, led by Twain Dai and Jennifer Rodger, set out to address a persistent gap in translational neuroscience. Human neuroimaging studies have linked sex differences in negative bias to aberrant functional connectivity in depression, but the causal chain connecting adolescent stress exposure to sex-specific cognitive distortion and resting-state network reorganization in adulthood has remained poorly understood. Animal models allow investigators to control the timing and duration of stress with a precision impossible in clinical populations, making rats an ideal system for disentangling cause from correlation.

Twenty-eight Wistar rats, balanced across sex and arriving at the laboratory at five weeks of age—a developmental window corresponding to adolescence—were first trained on a judgment bias task, a behavioral paradigm widely regarded as one of the best available measures of emotional valence in animals. In this task, distinct tactile cues signaled different reward outcomes, teaching the animals that one texture predicted a large reward while another predicted a smaller one. Once the rats had learned the distinction, the experimenters introduced ambiguous probe cues falling between the trained extremes. How optimistically or pessimistically an animal responded to those ambiguous signals provided a quantifiable index of its cognitive bias.

After training was complete, the animals were randomly and equally assigned to either a chronic restraint stress group or a control group, again balanced by sex. Chronic restraint stress is a standard and well-characterized rodent model of prolonged psychological stress, in which animals are repeatedly placed in restraining tubes for defined periods. Importantly, the study employed a longitudinal design: each animal served as its own baseline, undergoing resting-state functional magnetic resonance imaging before the stress protocol began, and again in young adulthood after it ended. Probe trials followed the second scan, allowing the researchers to relate neural changes to behavioral changes within the same individuals.

Resting-state fMRI measures spontaneous, low-frequency fluctuations in blood-oxygenation signals while an animal lies still, revealing which brain regions tend to activate together and therefore how distributed networks are functionally organized. Because the scans were collected offline and at two time points, the team could compute connectivity change scores for every animal, isolating the effects of the stress exposure from individual differences present from the start. Behavioral data were analyzed with linear mixed-effects models, a statistical framework well suited to repeated-measures designs, while network-level connectivity changes were assessed using Network-Based R-Statistics, a method designed to detect coordinated alterations across interconnected sets of brain regions rather than isolated pairwise links.

The behavioral results were strikingly sex-dependent. When the researchers compared change from adolescent baseline to post-stress testing, females showed a significantly greater tendency than males toward negative cognitive bias. In practical terms, stressed females were more likely to interpret ambiguous cues as signaling the unfavorable outcome, a pattern that echoes the pessimistic interpretation of uncertain situations seen in human depression and that has long been noted to be more pronounced in females than in males. Because the same animals had established optimistic baselines before stress exposure, this shift cannot be attributed to pre-existing differences in temperament between the sexes.

Equally important, the brain imaging data revealed that this behavioral divergence was mirrored by divergent network reorganization. Network-based statistics identified a female-specific reorganization within a distributed subnetwork spanning some of the brain’s most emotion-relevant territory: the brainstem, limbic structures, the striatum, the insula, parahippocampal regions, the orbitofrontal cortex, and the retrosplenial cortex. This collection of regions is far from arbitrary. The limbic system and brainstem together govern threat detection and arousal; the insula integrates bodily and emotional states; the striatum drives reward processing and motivated behavior; and the orbitofrontal cortex evaluates outcomes and updates expectations. Parahippocampal and retrosplenial areas contribute contextual memory and spatial framing, helping an organism situate emotionally salient events in the world.

That chronic adolescent stress preferentially rewired connectivity among precisely these regions in females—but not in males—provides a candidate mechanistic bridge between two well-established clinical observations: that negative cognitive bias in depression is stronger in women, and that adolescent stress is a potent risk factor for adult depression, particularly in females. The alignment between the behavioral measure and the network measure in the same longitudinal cohort strengthens the case that the connectome changes are not incidental but functionally meaningful, potentially representing the neural substrate on which pessimistic interpretation of ambiguity is built during a sensitive developmental period.

The authors emphasize that these results highlight the translational importance of considering both sex and distributed network organization when modeling how adolescent stress shapes later vulnerability to depression. That message carries weight for the broader field. Many preclinical studies still analyze only male animals or pool the sexes, and many neuroimaging analyses focus on single regions rather than networks. This study demonstrates that both choices can obscure exactly the effects that matter most: a stress-induced, female-specific reorganization of a multi-region circuit that would have been invisible without sex-balanced design, longitudinal baseline scanning, and network-aware statistics.

For human health, the implications are cautiously encouraging. If a defined adolescent stressor produces measurable, sex-specific reorganization of emotion-processing circuits that persists into young adulthood, then adolescence represents not merely a period of heightened risk but a concrete window of opportunity for intervention. Biomarkers derived from resting-state connectivity, combined with behavioral measures of cognitive bias, could in principle identify young people whose brains have been reorganized by early stress before full-blown depression develops. The rat model established here—linking a quantifiable judgment bias to longitudinal connectome change—offers a rigorous platform for testing precisely such preventive strategies, and for understanding why females and males may need different ones.

Subject of Research: Sex-dependent effects of chronic adolescent stress on cognitive bias and functional brain networks

Article Title: Sex-dependent effects of chronic stress during adolescence on cognitive bias and functional connectome in young adult rats

Article References: Dai, T., Jaeschke-Angi, L., Penrose-Menz, M., Rosenow, T., & Rodger, J. (2026). Sex-dependent effects of chronic stress during adolescence on cognitive bias and functional connectome in young adult rats. Translational Psychiatry. https://doi.org/10.1038/s41398-026-04511-7

Image Credits: AI Generated

DOI: 10.1038/s41398-026-04511-7

Keywords: adolescent stress, cognitive bias, depression, sex differences, resting-state fMRI, functional connectivity, chronic restraint stress, rats, Translational Psychiatry, neuroimaging, limbic system, network-based statistics

The research team built a novel laboratory paradigm designed to isolate exactly how explicit reliability markers shape human choices. Participants were asked to make a simple choice between two options, but before committing to a decision they saw predictions from several advisory sources. Crucially, each source came with an explicit, probabilistic label stating the likelihood that it would provide correct information. Some sources were flagged as highly trustworthy, others as moderately reliable, and some were even labelled as outright liars, meaning their predictions were known to point toward the wrong answer with high probability. This design allowed the researchers to ask whether participants took these stated probabilities at face value, weighting each source’s advice in proportion to its advertised trustworthiness, as an ideal Bayesian observer would.

The results revealed a systematic distortion. Participants did not use the stated probabilities of source correctness as advertised. Instead of treating a source labelled as unreliable as nearly worthless, they appeared to assign it a moderate degree of trust, as if the label had been softened or discounted. Computational modelling made this pattern precise: the researchers estimated the effective reliability that each participant implicitly assigned to every source, and these estimated values diverged substantially from the explicit markers. Sources described as trustworthy were not rewarded with the full weight their labels justified, while sources described as unreliable were not dismissed as thoroughly as they should have been. In effect, the entire reliability spectrum was compressed, with extreme labels pulled toward the middle.

This compression has a striking consequence: even sources explicitly flagged as unreliable biased participants’ choices. The advice of a known bad source nudged decisions in the direction it pointed, rather than being inverted or ignored. An ideal decision maker who knows a source lies most of the time should either disregard it entirely or, better still, treat its predictions as evidence for the opposite conclusion. Participants did neither. Their behaviour suggested that a source labelled as a liar was processed as if it were merely mediocre, a source of noisy but not systematically misleading information. The explicit warning, in other words, failed to translate into the appropriate inferential strategy.

The picture became even more intriguing when the researchers examined sources known to reliably predict the incorrect answer. The presence of these lying sources did not simply add noise; it impaired performance in a specific and measurable way. Using a drift diffusion framework, in which decisions are modelled as the gradual accumulation of evidence toward one option or the other, the team found that lying sources increased the leakiness of evidence accumulation. In a leaky accumulator, evidence decays over time rather than being preserved, so the decision process becomes less efficient and less able to build a clean, decisive signal. The lying sources did not just push choices in the wrong direction; they degraded the very machinery of the decision, making the integration of all available information sloppier.

Why might the human mind distort explicit reliability in this way? The authors suggest that participants may not take probabilistic labels at face value because the mapping between a stated number and the appropriate weight in a decision is not intuitive. A label saying a source is correct seventy percent of the time must be converted into a likelihood ratio, and that conversion may be systematically miscalibrated. Alternatively, people may harbour a general scepticism about the labels themselves, treating extreme claims, whether of high competence or of outright dishonesty, as exaggerated. Whatever the underlying cause, the result is a decision system that behaves as though the world is flatter and more moderate than the evidence declares it to be.

Yet the study is not simply a story of human irrationality. Alongside each choice, participants rated how much they felt a given source had influenced their decision, and these metacognitive judgements revealed genuine insight. Participants were aware of the pull that unreliable sources exerted on their choices, and they could evaluate how much a given source had increased the likelihood of their response. In other words, even though the integration of reliability information was biased at the level of the choice itself, people retained some access to the process that produced their decisions. They knew, at least approximately, which sources had swayed them and by how much.

This dissociation between biased performance and partially accurate metacognition is one of the most compelling aspects of the findings. It suggests that the distortion occurs early or implicitly in the decision pipeline, while a separate monitoring process tracks the outcome with reasonable fidelity. Metacognitive accuracy of this kind matters because it is the foundation of self-correction. A decision maker who knows that a dubious source nudged her choice is in a position to discount that influence on the next trial, or to seek better information before acting. The study implies that this corrective capacity exists in humans, even when the automatic integration of advice goes astray.

The implications reach far beyond the laboratory. Modern information environments are saturated with explicit and implicit reliability signals: verified badges, fact-check labels, star ratings, community notes, and reputation scores. If people systematically compress the reliability spectrum, treating flagged misinformation as merely mediocre rather than as inverted evidence, then the practical value of these labels may be far lower than designers assume. A warning label that shifts a source from perceived liar to perceived mediocrity still leaves that source exerting a misleading pull on belief and choice. The findings suggest that interventions may need to do more than attach probabilities to sources; they may need to teach people how to convert those probabilities into appropriate inferential weight, including the counterintuitive step of reading a known liar’s claims backwards.

By combining a carefully controlled paradigm, computational modelling of choice, and trial-by-trial metacognitive reports, the study offers a rare, quantified window into how humans consume advice in an uncertain world. The message is double-edged. Our decisions are demonstrably distorted by explicit reliability information, in ways that let even confessed liars bias our choices and degrade our evidence accumulation. But we are not blind to our own biases. Some part of the mind keeps watch, registering which voices pulled us and how hard, and that quiet awareness may be the raw material from which better information habits can be built.

Subject of Research: How explicit reliability information about advice sources is integrated into decision-making and metacognition

Article Title: How does explicit reliability guide choices? Effect of trustworthiness on choice and metacognition

Article References: Ota, K., Ciston, A., Haggard, P., Preuss, T. G., & Charles, L. (2026). How does explicit reliability guide choices? Effect of trustworthiness on choice and metacognition. PLOS Computational Biology, 22(10), e1014818. https://doi.org/10.1371/journal.pcbi.1014818

Image Credits: AI Generated

DOI: 10.1371/journal.pcbi.1014818

Keywords: decision-making, metacognition, source reliability, computational modelling, drift diffusion, trustworthiness, Bayesian inference, evidence accumulation, misinformation, cognitive bias, PLOS Computational Biology, psychology

A team of researchers led by Gunnar P. Epping and Jennifer S. Trueblood of Indiana University, working with Andrew Caplin of New York University, Erik Duhaime of Centaur Labs, and Daniel Martin of the University of California, Santa Barbara, has now demonstrated a strikingly simple fix. In a study published in Behavior Research Methods, they introduce what they call cognitive-inspired data engineering: the idea that empirical findings and models from cognitive science can be applied directly to crowdsourced judgments to correct systematic biases before the data ever reaches a machine learning algorithm. Rather than de-biasing models after training or modifying training objectives, the team intervenes upstream, at the level of the individual annotation, and shows that this single correction improves both the quality of crowdsourced datasets and the accuracy of the convolutional neural networks trained on them.

The core insight concerns how annotators express uncertainty. The industry-standard approach asks workers to provide a single binary label for each image, a method that discards valuable information whenever an annotator is unsure. An alternative is to elicit subjective probability judgments, asking annotators how likely they think it is that an image belongs to a given category. Probability judgments convey far more information, but they also open the door to more cognitive biases. Beyond simple response biases, where a rater favors one side of the scale, probability judgments are vulnerable to overconfidence and underconfidence, phenomena cognitive scientists have documented for decades. The hard-easy effect, for instance, leads people to be overconfident on difficult classification tasks and underconfident on easy ones. As a result, subjective probabilities rarely match the actual likelihood of being correct, a mismatch known as poor calibration. Crucially, however, these judgments remain correlated with correctness, which means the information they carry can be rescued if the biases are stripped away.

The researchers used a well-established cognitive model called the linear in log odds, or LLO, function to perform that rescue. The function transforms each annotator’s raw probability judgments into log-odds, fits a logistic regression against ground-truth labels from a small calibration set, and then maps the annotator’s remaining judgments onto recalibrated probabilities. The two free parameters of the function carry direct psychological interpretations. The slope parameter corrects for overconfidence or underconfidence: when its magnitude exceeds one, the function spreads judgments toward the extremes of the scale, indicating the annotator was underconfident, while a magnitude below one compresses judgments toward the center, correcting overconfidence. The intercept parameter corrects overall response tendency, shifting judgments up or down the probability scale to undo a systematic leaning toward one class. Because biases differ substantially across individuals, the researchers fit a unique recalibration function to each annotator separately, effectively normalizing every worker’s judgments before combining them into crowd labels.

To test the approach, the team chose a task with real clinical stakes: classifying white blood cells as blast cells or non-blast cells, a critical step in diagnosing malignant blood diseases such as leukemia and lymphoma. The stimuli were 549 digital images of Wright-stained cells from anonymized patient blood smears at Vanderbilt University Medical Center, captured with an automated cell morphology instrument and ground-truthed by three hematopathology faculty members who had to agree unanimously on each classification. The task is an ideal sandbox for studying high-skill perceptual decision-making: novices can reach roughly 65 percent accuracy with minimal training, the task remains challenging even for experts, and machine learning models can be trained on only a few hundred images rather than the thousands required for comparable problems.

In the first experiment, 400 Amazon Mechanical Turk workers recruited through CloudResearch judged the probability that each image contained a blast cell. After training with labeled example images and practice trials with feedback, participants completed testing blocks in which a small subset of images with known labels served as the calibration set for the LLO function. The results were revealing. Recalibration barely changed individual accuracy, which hovered around 65 percent before and after the transformation. But it dramatically improved calibration. Before correction, participants clustered their judgments at the extremes of the scale, and those extremes were badly miscalibrated: images judged to have a zero percent chance of being a blast cell were in fact blast cells roughly 28 percent of the time, while images judged to be certain blasts were truly blasts only 68 percent of the time. After recalibration, the response distribution spread across the scale and the calibration curve moved much closer to the ideal identity line. Most fitted slope parameters fell between negative one and one, confirming that the typical annotator was overconfident, and a few negative slopes even flipped the judgment ordering of participants who performed below chance.

The payoff appeared when individual judgments were aggregated into crowd labels. Using all available judgments, recalibration lifted crowd-label accuracy from 81.6 percent to 85.1 percent, even though no individual annotator had become more accurate. The researchers attribute this to an implicit variance-weighting mechanism: by compressing the judgments of overconfident annotators toward 50 percent, recalibration reduces their outsized influence after binarization, while stretching the judgments of underconfident annotators toward the extremes increases theirs. Recalibrated labels were also more efficient, reaching any given accuracy level with fewer judgments per image, and the benefit of calibration leveled off surprisingly quickly. Just six judgments on calibration images, three per class, captured most of the improvement, raising accuracy by 2.2 percentage points, with only a further 0.4 percentage point gain from extending to ten judgments.

The second experiment moved the study into a real-world setting using DiagnosUs, a crowdsourcing platform specializing in medical data annotation where skilled users, many of them medical students and healthcare professionals, compete in contests for prize money. The platform’s existing structure, in which gold-standard images with known labels are shuffled into unlabeled sets to provide feedback, supplied a natural calibration set at no extra cost. A total of 175 participants took part, split between a probability-judgment condition and an industry-standard binary-choice condition. Individual accuracy was higher than in the novice experiment, around 69 percent, and the two elicitation methods produced crowd labels of nearly identical accuracy, 88.3 percent each. Recalibration, however, pushed the probability-based crowd labels to 96.7 percent, an 8.4 percentage point gain that exceeded the 3.5 point gain seen with novices. The researchers attribute the larger benefit to the more accurate judgments available for fitting the recalibration function, not to the larger number of judgments, since gains again plateaued at roughly ten calibration judgments per annotator. Notably, the fitted intercept parameters skewed negative in this experiment, suggesting that medically trained annotators erred toward diagnosing cells as cancerous when unsure, consistent with the clinical convention that false positives are less costly than false negatives.

The improvements propagated directly to the machine learning models. The team trained GoogLeNet convolutional neural networks via transfer learning on crowd-labeled datasets, repeating the full training and evaluation procedure across 100 independently resampled datasets with five-fold cross-validation. At every number of judgments per image, models trained on recalibrated labels outperformed those trained on raw probability judgments or binary choices with respect to the true expert labels. Remarkably, models trained on just five recalibrated judgments per image matched the accuracy of models trained on the full set of uncalibrated judgments, a result with immediate economic implications. Because annotation cost scales with the number of judgments collected, recalibrated datasets achieve any target accuracy at a fraction of the annotations, directly attacking what the authors identify as the primary bottleneck in developing medical AI: the shortage of large, accurately labeled datasets in high-skill domains.

The authors are careful to frame the work as a proof of concept rather than a universal solution. The experiments used a single set of stimuli sampled with balanced class prevalence, so future work must test whether recalibration can also correct prevalence-induced errors that arise when target categories are rare, a notorious source of mistakes in visual search. The team also binarized crowd judgments before training, and training directly on continuous probability estimates may preserve even more of the uncertainty information that recalibration cleans up. Still, the broader message is hard to ignore. Cognitive scientists have spent decades cataloguing the systematic distortions in human probability judgment, and this study shows that knowledge can be converted into a simple, label-efficient transformation applicable to virtually any crowdsourcing task involving subjective probabilities. As the authors put it, the approach requires no changes to model architectures or training procedures; it simply fixes the data. In an era when the quality of AI systems is increasingly limited by the quality of human labels, teaching machines to think may begin with correcting how humans report what they think.

Subject of Research: Using cognitive-science-based recalibration of crowdsourced probability judgments to improve AI training data quality

Article Title: Improving crowdsourcing for AI through cognitive-inspired data engineering

Article References: Epping, G. P., Caplin, A., Duhaime, E., Holmes, W. R., Martin, D., & Trueblood, J. S. (2026). Improving crowdsourcing for AI through cognitive-inspired data engineering. Behavior Research Methods, 58(10), Article 283. https://doi.org/10.3758/s13428-026-03160-4

Image Credits: AI Generated

DOI: 10.3758/s13428-026-03160-4

Keywords: crowdsourcing, machine learning, cognitive bias, recalibration, medical imaging, data annotation, probability judgments, wisdom of crowds, neural networks, training data, human judgment, AI

The study, led by Zhifang Liu of Shaanxi Normal University and Hunan First Normal University, together with Chengyu Zhang, Jing Li, and Bihua Yan, was motivated by a practical concern that has grown acute in the era of viral campus controversies: university administrators and researchers can observe online public opinion erupting in real time, but they lack validated tools to identify which students or which patterns of discourse signal genuine risk. Existing instruments in psychology measure related constructs such as cognitive distortions or impulsivity in general contexts, but the researchers argue that online discourse has its own dynamics—anonymity, immediacy, audience effects, and the amplifying architecture of platforms—that demand a purpose-built measurement tool. Their answer was to construct one from the ground up using a mixed-methods design that mirrors the gold-standard sequence of modern scale development.

The qualitative phase began with the raw material of student experience. The team conducted semi-structured interviews and collected 153 valid open-ended questionnaires, asking students to describe how they and their peers behave irrationally in online discussions. Rather than imposing a pre-existing theory, the researchers let the students’ own language drive the process. Five independent coders worked through the responses, systematically extracting recurring themes and translating them into candidate questionnaire items. This coding exercise generated an initial item pool organized around four theoretical dimensions, a structure that would later be tested statistically. The draft items were then scrutinized by a cross-disciplinary expert panel, whose job was to sharpen wording, eliminate ambiguity, and ensure that each item captured a distinct facet of the construct without overlapping its neighbors.

That four-dimensional architecture is the conceptual heart of the instrument. The first dimension, cognitive bias, covers the systematic errors in thinking that distort how students interpret online information—jumping to conclusions, accepting claims without evidence, or framing disagreement as hostility. The second, emotional sensitivity, captures how strongly online interactions trigger affective reactions, from taking criticism personally to being swept up in collective outrage. The third, self-regulation deficiency, measures failures of impulse control: the inability to pause before posting, to disengage from escalating arguments, or to moderate one’s own emotional state. The fourth, behavioral deviance, describes the observable endpoint—posting behavior that departs from reasoned norms, such as aggressive attacks, rumor spreading, or deliberate provocation. Together, the dimensions trace an arc from distorted thinking through emotional reactivity and failed self-control to problematic action.

With the item pool refined, the quantitative phase put the draft scale through the standard gauntlet of psychometric testing on two independent samples. The first, comprising 500 undergraduate students, was used for item analysis and exploratory factor analysis, the statistical procedure that reveals whether the items naturally cluster into the hypothesized dimensions or whether the underlying structure is different from what the qualitative work suggested. The second, a separate sample of 476 students, was reserved for confirmatory factor analysis, a stricter test in which the hypothesized four-factor structure is specified in advance and the data are asked whether they fit. Splitting the samples this way matters: a scale that only fits the data used to build it is a statistical artifact, not a measurement tool. The two-sample design gives the validation genuine out-of-sample credibility.

The numbers that emerged are, by conventional standards, solid. The confirmatory factor analysis returned a chi-square to degrees-of-freedom ratio of 2.094, comfortably within the accepted threshold of below 3. The comparative fit index reached 0.934 and the Tucker-Lewis index 0.928, both above the 0.90 benchmark that indicates good fit, while the standardized root mean residual of 0.057 and the root mean square error of approximation of 0.058 fell within the ranges generally interpreted as acceptable to good. Perhaps most striking is the internal consistency coefficient of 0.95, a value indicating that the 28 items of the final scale hang together with remarkable coherence. In practical terms, the statistics say that the four dimensions are real, that the items measure them reliably, and that the scale behaves consistently across the independent validation sample.

The final instrument therefore comprises 28 items spanning cognitive bias, emotional sensitivity, self-regulation deficiency, and behavioral deviance. But the researchers did not stop at validation; they applied the scale to see whether it detects meaningful differences between groups. It did. Scores varied significantly by gender, by the duration of internet use, and by family background, suggesting that irrational online behavior is not uniformly distributed across the student population. Intriguingly, no significant difference emerged across grade levels, implying that the tendency toward irrational discourse is not simply a matter of academic maturity or year of study—first-year students and near-graduates appear broadly comparable in this respect. The researchers suggest that identifying these at-risk subgroups could help universities target interventions more precisely than blanket awareness campaigns.

The implications reach beyond campus administration. The study was funded under Chinese research programs on risk prevention and guidance mechanisms for group polarization in university network public opinion, and on the interactive evolution of online and offline opinion dynamics—signaling that the scale is intended as a tool for understanding how irrational discourse can escalate into polarized collective phenomena. Group polarization, in which like-minded individuals push each other toward increasingly extreme positions, is one of the most studied dangers of online communication, and a validated individual-level measure of the psychological precursors could give researchers a way to study that escalation at its source. If cognitive bias and self-regulation deficits can be quantified before they harden into deviant posting behavior, interventions might be designed to strengthen exactly those weak points.

The work also carries the usual caveats that accompany any new psychological instrument. The scale was developed and validated with Chinese undergraduate samples, and cross-cultural generalization remains to be tested; constructs like emotional sensitivity and behavioral deviance in online discourse may manifest differently in other linguistic and platform environments. Self-report measures, however carefully constructed, depend on respondents’ willingness to admit to irrational tendencies, and the authors themselves describe the psychometric support as preliminary. Still, the methodological rigor of the two-sample validation, the expert refinement process, and the strong fit statistics give the instrument a credible foundation. The study was approved by the Ethics Committee of Hunan First Normal University and conducted with informed consent under the principles of the Declaration of Helsinki.

What makes the research resonate beyond the academy is how precisely it captures a phenomenon that anyone who has spent time online will recognize: the moment when a discussion stops being an exchange of reasons and becomes a contest of reactions. By decomposing that moment into measurable components—biased cognition, emotional triggering, failed restraint, and deviant expression—the scale turns a vague cultural anxiety about online irrationality into something that can be counted, compared, and tracked over time. Whether it ultimately helps universities defuse volatile opinion storms or simply gives psychologists a sharper lens on digital behavior, the instrument marks a step toward treating online discourse not as an unknowable torrent but as a measurable expression of human psychology, one 28-item questionnaire at a time.

Subject of Research: Development and psychometric validation of a scale measuring irrational behavior in undergraduate students' online discourse

Article Title: Development and validation of a scale for measuring irrational behavior in undergraduate students’ online discourse

Article References: Liu, Z., Zhang, C., Li, J., & Yan, B. (2026). Development and validation of a scale for measuring irrational behavior in undergraduate students’ online discourse. BMC Psychology. https://doi.org/10.1186/s40359-026-05678-w

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05678-w

Keywords: online discourse, irrational behavior, undergraduate students, scale development, psychometrics, confirmatory factor analysis, exploratory factor analysis, cognitive bias, self-regulation, group polarization, BMC Psychology, social behavior