The research team, led by Wei Wang and Boyuan Lu of the University of Wisconsin-Madison and Peking University, together with colleagues at the University of Maryland and Peking University’s Department of Sociology, exploited a natural experiment embedded in the timing of the December 2015 alerts. Beijing’s red haze alerts, the most severe tier in the city’s emergency response system, were triggered by forecasts of sustained heavy pollution and came bundled with dramatic countermeasures, including restrictions on vehicle use, suspension of industrial production, and closures of schools and construction sites. Because the alerts arrived suddenly and applied citywide, the researchers could compare the pollution estimates of people surveyed before the alerts with those of people surveyed after, treating the intervention much like a randomized treatment in a laboratory setting.

The study combined two sources of data. The first was official air-quality monitoring information from the Beijing Municipal Ecological and Environmental Monitoring Center, which provided objective benchmarks for fine particulate matter concentrations, known as PM2.5, and for the frequency of yellow-level haze days, a milder pollution category. The second was a survey experiment involving 947 university students who were randomly assigned to report their perceptions of pollution conditions either before or after the red alerts. By asking participants to estimate pollution levels and then comparing those estimates against monitored values, the researchers could quantify the estimation gap with unusual precision for a social science study of environmental perception.

The headline result is striking. Among students who reported being aware of the red alert, mean overestimation of PM2.5 concentrations fell from 49.90 percent above monitored values to 30.53 percent. The correction was even more dramatic for yellow-level haze days, a category that most residents would find harder to judge from memory: overestimation collapsed from 178.89 percent to 87.43 percent. In other words, before the alerts, students believed mild haze days were nearly three times more common than the monitoring record showed; afterward, their estimates were less than twice the true figure. Multivariate statistical models that controlled for individual characteristics confirmed these reductions, estimating that the red alert cut PM2.5 overestimation by 19 to 27 percent and yellow-level haze-day overestimation by 92 to 147 percent.

These numbers speak to a well-documented puzzle in the literature on environmental risk perception. Decades of research, stretching back to foundational work on judgment under uncertainty by Amos Tversky and Daniel Kahneman, have shown that people rely on heuristics and salient memories rather than statistical records when assessing hazards. Air pollution is a particularly tricky case because its severity varies across space and time, is often invisible at low concentrations, and is filtered through media coverage, personal symptoms, and social conversations. Previous studies in China and elsewhere have found that self-reported pollution levels routinely diverge from station data, sometimes wildly, and that these mismatches can distort protective behavior, undermine trust in government statistics, and complicate emergency management. The Beijing study suggests that a single, highly salient official alarm can serve as an anchoring event that resets the reference point people use when making subsequent judgments.

The mechanism the authors propose is consistent with salience theory in behavioral economics. A red alert is not just information; it is a dramatic, citywide signal that makes pollution cognitively available and provides a concrete benchmark against which past and present conditions can be evaluated. Once residents know that a red alert corresponds to a specific, severe pollution state, they apparently recalibrate their internal scale, becoming more accurate not only about extreme episodes but also about milder yellow-level conditions. This spillover from high-severity signals to lower-severity judgments is one of the study’s most interesting technical contributions, because it implies that extreme warnings do more than communicate the immediate emergency; they teach the public where the whole pollution distribution sits.

Yet the alert’s effects were strikingly selective when it came to evaluations of government mitigation efforts. The researchers examined whether the red alerts changed how students assessed specific policy responses, including vehicle restrictions, factory shutdowns, urban greening programs, and restaurant regulations. Only the vehicle restriction policy showed a significant increase in recognition and evaluation. The authors attribute this asymmetry to measure visibility: odd-even license plate rules remove half the cars from the streets, an effect that any commuter can directly observe, whereas factory closures, tree planting, and restaurant oversight happen largely out of public view. Assessments of those less visible measures did not shift at all in response to the alert. For policymakers, this is a sobering result, because the measures that are hardest for citizens to see are precisely the ones whose public support may depend on deliberate communication rather than ambient visibility.

The study also probed whether the alerts changed self-reported protective behavior, such as wearing masks or limiting outdoor activity, and whether gender shaped the accuracy of pollution estimates. On both fronts the results were more muted. Changes in protective behavior were modest and statistically insignificant, and gender had no significant effect on overestimation, a finding that runs counter to some earlier literature reporting systematic gender differences in environmental risk perception and pro-environmental behavior. The authors are careful not to overclaim here; a survey of university students, however well designed, captures a specific population at a specific moment, and self-reported behavior is vulnerable to recall and social desirability biases that the study’s own framing acknowledges.

Those limitations are worth taking seriously, but they do not erase the central contribution. By pairing a randomized survey design with objective monitoring data and a genuine policy shock, the study moves the field beyond the correlational snapshots that have dominated perception research. It demonstrates causally, not just associatively, that high-severity environmental alarms function as reference signals that narrow the perception gap. It also identifies the boundary conditions of that effect: the correction applies to pollution estimates but not automatically to policy evaluations, and it operates through visibility, meaning that invisible mitigation work remains invisible even during a crisis. The practical implication is that warnings and transparency must travel together. An alert can teach the public what severe pollution looks like, but only visible, well-communicated policy action can convert that heightened attention into durable trust in the institutions managing the air.

As cities worldwide grapple with wildfire smoke, dust storms, and industrial smog in a warming climate, the Beijing red alert experiment offers a template for evidence-based risk communication. Emergency warnings are often evaluated solely by whether they prompt immediate protective action, a metric on which this study found only weak effects. But the research suggests a second, subtler function that has been largely overlooked: alarms as calibration devices that align public perception with scientific measurement. If subsequent studies replicate these findings in other cities and populations, environmental agencies may come to see their most severe alert tiers not just as sirens but as teaching moments, brief windows in which the public is unusually receptive to learning what the numbers on the monitor actually mean for the air in their lungs.

Subject of Research: The effect of high-severity air pollution alerts on public perception accuracy and policy evaluation during Beijing's 2015 red haze alerts

Article Title: High-severity environmental alarms as reference signals: Evidence from Beijing’s first red haze alerts

Article References: Wang, W., Lu, B., Zhang, X., Fan, X., & Zhang, W. (2026). High-severity environmental alarms as reference signals: Evidence from Beijing’s first red haze alerts. Air Quality, Atmosphere & Health, 19(10), Article 228. https://doi.org/10.1007/s11869-026-02117-y

Image Credits: AI Generated

DOI: 10.1007/s11869-026-02117-y

Keywords: air pollution, PM2.5, red haze alert, risk perception, Beijing, environmental communication, survey experiment, public trust, protective behavior, salience theory, emergency response, environmental monitoring

The research, published in BMC Medical Education by Merve Candan, Melike İdacı, Feyza Topbaşoğlu, and Alper Çamgöz of the university’s Department of Pediatric Dentistry, surveyed 441 dental students at a single faculty. The response rate of 75.3 percent was high enough to give the sample meaningful weight, and the investigators classified participants into three groups according to their level of patient contact: preclinical students who had not yet treated patients, observer students who shadowed clinical work, and clinical students who performed procedures themselves. This stratification is more than administrative convenience. It allowed the team to test a hypothesis that has long circulated in health-professions education, namely that exposure to real patients and real clinical environments is what converts abstract knowledge about infection into concrete habits of protection.

To measure the outcome, the researchers used the Communicable Diseases Risk Awareness and Protection Scale, a questionnaire instrument that captures both cognitive awareness of disease risks and self-reported protective behaviors, alongside demographic items. The statistical machinery applied to the data was correspondingly layered. Group differences were tested with one-way analysis of variance and Bonferroni-corrected post hoc comparisons, with effect sizes reported so that readers could judge not just statistical significance but practical magnitude. Independent-samples t-tests compared male and female respondents. Then came the analytical core: multiple linear regression to identify independent predictors of protective behavior, and an observed-variable path analysis to decompose the association between each predictor and behavior into a direct pathway and an indirect pathway running through risk awareness.

The results paint a clear developmental gradient. Preclinical students scored significantly lower than both observer and clinical students on the behavior dimension and on the total scale score, with p-values below 0.001, while the observer and clinical groups did not differ from each other. In other words, the jump in protective behavior occurs at the threshold of patient contact, not gradually across the later stages of training. Once students begin observing and then performing clinical work, their protective practices appear to consolidate quickly and then plateau. This pattern is consistent with the idea that infection-control behavior is learned less through lectures about hepatitis B or HIV transmission and more through the embodied experience of sitting chairside, where a splash of saliva or a needlestick near-miss carries immediate, visceral meaning.

Gender emerged as the other dominant axis of difference. Female students scored higher than male students on every dimension of the scale, with all comparisons reaching significance at p ≤ 0.01. This finding echoes a broader literature on gender differences in health-protective behavior, in which women across many populations report greater engagement with precautionary practices, from hand hygiene to mask wearing. In the dental training context, the authors’ data indicate that this gap is present across the awareness and behavior dimensions alike, meaning the disparity is not merely a matter of women knowing more about risks; they also act more consistently on that knowledge. The consistency of the gender effect across all scale dimensions strengthens the case that it reflects a genuine behavioral orientation rather than a measurement artifact.

The regression analysis sharpened the picture considerably. When gender and clinical training stage were entered together as predictors of protective behavior, the primary outcome of the study, only these two variables remained significant. Being female predicted higher protective behavior with p < 0.001, and advancing clinical training stage predicted it with p = 0.004. Risk awareness, the variable one might intuitively expect to be the engine of protective action, did not stand up as an independent driver once these factors were accounted for. For educators who have invested heavily in knowledge-based infection-control curricula, this is an uncomfortable result: the knowledge students acquire appears weakly coupled to what they actually do at the chair.

The path analysis pushed this conclusion further. For gender, the model identified both a direct association with protective behavior and an indirect association operating through risk awareness, with the indirect path significant at p = 0.003. This means that part of the female advantage in protective behavior can be statistically attributed to higher awareness, but a substantial direct effect persists even after that mediation is accounted for. For clinical training stage, the story was different in an instructive way: the total association with behavior was significant at p = 0.032, but neither the direct path nor the indirect path through awareness was individually significant. The authors interpret this as evidence that the overall effect of training stage on behavior is real but diffuse, distributed across pathways too small to reach significance on their own, and crucially not channeled through increased risk awareness.

That last point deserves emphasis because it strikes at a foundational assumption of the classic knowledge–attitude–practice framework that has shaped health-education interventions for decades. The framework presumes a causal chain running from knowledge to attitude to behavior, and it underwrites countless training programs that measure success by gains in knowledge scores. The present study found that increased risk awareness was not directly reflected in protective behavior, and that a strong awareness-mediated pathway to behavior was not supported in the data. Awareness rose only weakly with advancing clinical training stage, while behavior rose substantially. The two constructs, in this population, appear to move on largely separate tracks, driven by different forces.

The practical implications follow naturally. If protective behavior improves mainly with hands-on clinical exposure, then infection-control education should be embedded in clinical practice rather than delivered as preclinical theory alone. Simulation-based training, early chairside mentoring, and structured feedback on personal protective equipment use may do more to build durable habits than additional didactic content. If the gender gap is robust, as it appears to be here, then targeted interventions for male students, perhaps emphasizing the personal and professional stakes of consistent protection, could close a disparity that currently leaves one half of the trainee population less protected. And if awareness alone is a weak lever, then measuring awareness as a proxy for safety culture in dental schools may be misleading, giving educators false comfort while actual chairside practices lag behind.

The authors are appropriately cautious about the limits of their design. A cross-sectional study captures a single moment in time, so it can identify associations but cannot establish that clinical exposure causes improved behavior, however plausible that mechanism appears. The sample came from a single institution, which raises fair questions about generalizability to dental faculties with different curricula, patient populations, or infection-control resources. Self-reported protective behavior is also vulnerable to social desirability bias, since students may overstate their compliance with protocols they know they should follow. Yet the study’s methodological rigor, including effect-size reporting, multicollinearity checks via variance inflation factors, and adherence to STROBE reporting guidelines, gives its findings credibility. For a profession whose daily work unfolds at the boundary between clinician and contagion, the message is clear: the safest dental students are not necessarily the best informed, but the most experienced and, in this dataset, the female ones. Building protective instinct, it seems, requires the clinic itself as the classroom.

Subject of Research: Determinants of communicable disease risk awareness and protective behaviors among dental students

Article Title: Determinants of communicable disease risk awareness and protective behaviors among dental students: a path analysis-based cross-sectional study

Article References: Candan, M., İdacı, M., Topbaşoğlu, F., & Çamgöz, A. (2026). Determinants of communicable disease risk awareness and protective behaviors among dental students: a path analysis-based cross-sectional study. BMC Medical Education. https://doi.org/10.1186/s12909-026-10543-w

Image Credits: AI Generated

DOI: 10.1186/s12909-026-10543-w

Keywords: dental education, infection control, communicable diseases, risk awareness, protective behavior, path analysis, cross-sectional study, clinical training, gender differences, health professions education, dental students, public health

The research team, led by Lei Lin of the Southwest University of Political Science and Law, together with Jing Tan of Chongqing University and Di Zheng of the Sichuan Academy of Social Sciences, recruited 599 urban residents in China for two randomized online experiments. The choice of population was deliberate. China has more than 900 million urban permanent residents, a group that overlaps heavily with the audiences of official government microblogs, weather applications, and other digital warning channels, and these users are among the most frequent adopters of AI-enabled services. The experimenters used Credamo, a survey platform with verified users, and applied strict quality controls: eligibility was restricted to the highest credit ratings, duplicate submissions from the same IP address or device were blocked, minimum completion times were enforced, and attention and comprehension checks filtered out careless respondents. Of 636 initial responses, 37 were excluded under preregistered criteria, leaving a final sample whose demographic profile shifted negligibly after cleaning.

The scenario at the heart of both experiments was a heavy rainfall warning issued during peak flood season in a Chinese city, a hazard chosen for its familiarity, its low political sensitivity, and the fact that both government agencies and AI platforms plausibly issue such alerts. Warning severity was communicated through the standardized color-coded signals of the China Meteorological Administration, ranging from blue for general risk through yellow and orange to red for extreme risk. All experimental materials were identical in content, length, and style; only the source cue varied. Pilot testing confirmed that participants could accurately identify who was said to have issued each warning and perceived the scenarios as realistic.

The first experiment employed a three-group between-subjects design comparing a government-only warning, an AI-only warning, and a consistent joint warning issued by both. Three hundred valid participants were randomly assigned, 101 to the government condition, 100 to the AI condition, and 99 to the joint condition, and balance checks confirmed no pretreatment differences across groups on gender, age, education, income, occupation, or disaster experience. A one-way analysis of variance revealed a substantial effect of warning source on information trust, with a partial eta squared of 0.215, a medium-to-large effect. Post hoc comparisons using Tukey’s honestly significant difference test showed that trust in government warnings significantly exceeded trust in AI warnings, with a mean difference of 0.29 points on the five-point scale and a Cohen’s d of roughly 0.45.

More striking still was the performance of the joint condition. Warnings attributed jointly to the meteorological station and an AI-driven warning platform produced the highest trust of all, exceeding the government-only condition by 0.26 points and the AI-only condition by 0.55 points, the latter corresponding to a Cohen’s d of approximately 0.85, a large effect. The authors interpret this as evidence that credibility emerges not from any single source but from the combination of institutional authority and technological channels. When AI alerts appear alongside official government communication, they are more likely to be perceived as complementary rather than autonomous, easing concerns about algorithmic opacity and unclear responsibility. The finding echoes the long-standing ‘speak with one voice’ principle in risk communication, which emphasizes coherent messaging across sources to sustain public confidence.

The experiment also confirmed that trust translates into action. In a hierarchical regression controlling for gender, age, education, prior disaster experience, AI usage, institutional trust, and risk perception, adding information trust to the model raised the explained variance in protective intentions from a nonsignificant 2.9 percent to 12.7 percent. Information trust was the strongest predictor of intentions to monitor hazard information, move vehicles and belongings to safety, identify emergency shelters, and stockpile supplies, with a standardized coefficient of 0.416. Risk perception retained a smaller but significant effect. Protective intention items showed acceptable internal consistency, with Cronbach’s alpha of 0.72, while the trust and anticipated regret composites reached 0.82 and 0.84 respectively.

The second experiment turned to the harder question: what happens when government and AI disagree? Four conditions were constructed. In one, both sources issued a low-level blue warning; in a second, both issued an orange warning, serving as a reference and replicating the joint condition from the first study. In the two conflict conditions, one source issued orange while the other issued blue, with the order of severity reversed between the two versions. After excluding invalid responses, 387 participants were distributed almost evenly across the four groups, and randomization checks again showed no meaningful imbalances. A pretest had verified that a one-level discrepancy was perceived as a genuine conflict and that orange was consistently read as more severe than blue.

Conflict, it turned out, produced a powerful emotional response. A one-way analysis of variance revealed a large effect of warning consistency on anticipated regret, with a partial eta squared of 0.349, and a nonparametric Kruskal-Wallis test corroborated the result. Participants in both conflict conditions reported significantly higher anticipated regret than those in either consistent condition, exceeding the consistent-low group by roughly 0.85 to 0.91 points and the consistent-high group by 0.33 to 0.39 points, all differences highly significant. Crucially, the two conflict groups did not differ from each other, meaning it did not matter whether the government or the AI issued the more severe alert. When sources diverged, people temporarily set aside questions of institutional authority and focused instead on avoiding the cost of underestimating the threat, a pattern consistent with worst-case thinking described in regret theory and loss aversion research.

Anticipated regret also proved to be a formidable driver of behavior. In a hierarchical regression, adding anticipated regret to the controls raised explained variance in protective intentions from a nonsignificant 1.7 percent to 26.8 percent, with a standardized coefficient of 0.502, meaning a one-unit increase in expected regret corresponded to roughly half a unit more protective intention, all else held constant. Together, the two experiments support a dual-pathway framework of decision making in multi-source warning environments: under consistent messaging, responses are structured by cognitive trust grounded in institutional authority and technological support; under conflicting messages, responses are instead guided by the emotional motivation to avoid future regret.

The implications for warning system design are substantial. Because joint government-AI releases enhance confidence, the authors suggest that policymakers develop coordination mechanisms between meteorological agencies and AI service providers, aligning update schedules, severity thresholds, and release timing to reduce discrepancies before they reach the public. At the same time, they caution against exploiting regret as a persuasive tool, which could breed anxiety or distrust; instead, communication strategies should explain why different systems produce different alerts and offer graded recommendations for how to respond. The study acknowledges limitations: the simulated scenarios cannot fully reproduce the urgency of real disasters, the sample of urban online users limits generalizability, and the two experiments examined trust and regret in separate randomized conditions rather than estimating a unified causal pathway. Nonetheless, as AI becomes an increasingly visible participant in public warning dissemination, the research makes clear that institutional credibility remains the anchor of public confidence, and that when machines and governments speak in harmony, people listen most closely of all.

Subject of Research: Public trust and behavioral responses to disaster warnings issued by governments versus artificial intelligence systems.

Article Title: Public Responses to Disaster Warnings from Government Versus AI: Evidence from Behavioral Experiments

Article References: Lin, L., Tan, J., & Zheng, D. (2026). Public Responses to Disaster Warnings from Government Versus AI: Evidence from Behavioral Experiments. International Journal of Disaster Risk Science. https://doi.org/10.1007/s13753-026-00766-2

Image Credits: AI Generated

DOI: 10.1007/s13753-026-00766-2

Keywords: disaster warnings, artificial intelligence, government trust, anticipated regret, risk communication, behavioral experiments, early warning systems, protective behavior, algorithm aversion, flood warnings, China, decision making