A new comprehensive analysis suggests that while conspiracy beliefs are deeply entrenched, they are not entirely resistant to change. The findings indicate that specific intervention strategies, particularly those involving direct fact-checking and alternative explanations, can achieve modest reductions in these beliefs. This meta-analysis was published in the European Journal of Social Psychology.
Conspiracy theories often serve as alleged explanations for complex events, ranging from the COVID-19 pandemic to geopolitical conflicts. Believing in these narratives can lead to tangible negative consequences, such as vaccine hesitancy or disregard for democratic norms.
Lukasz Stasielowicz, a researcher at the Department of Psychology at the University of Salzburg, undertook this study to move beyond individual experiments and provide a systematic overview of what actually works. While many scientists have tested various persuasion techniques, there has been a lack of clarity regarding which methods consistently yield results and which factors moderate their success. Stasielowicz aimed to quantify the average impact of these interventions and identify the specific characteristics that enhance their effectiveness.
“Conspiracy theories play a role in everyday situations, such as deciding whether to get vaccinated after reading alarming statements in a messaging app, arguing with relatives about political news during a meal, and reading social media posts questioning medical advice from health professionals,” Stasielowicz told PsyPost. “However, there is surprisingly little solid evidence on what actually helps people believe less in harmful conspiracy theories.”
“Many research teams have tested various interventions, but these strategies have not yet been compared to determine what works, for whom, and under what circumstances. This study aimed to systematically compare the available interventions and identify features that make interventions most effective.”
To accomplish this, the researcher conducted a systematic literature search using databases such as Web of Science, PsycINFO, and Google Scholar. The inclusion criteria required studies to measure conspiracy beliefs following an intervention and to employ a control group for comparison. Stasielowicz identified 56 independent samples that met these standards. These samples represented a total of 27,996 participants. The studies were primarily conducted in Western nations, with a significant number originating from the United States and Europe.
The analysis utilized a Bayesian three-level meta-analytic model. This statistical approach allowed the researcher to account for the fact that single studies often report multiple outcomes, such as measuring beliefs about several different conspiracy theories simultaneously. The dataset included 273 distinct effect sizes. Stasielowicz examined a wide range of variables to explain differences in outcomes, including the design of the intervention, the demographics of the participants, and the specific nature of the conspiracy beliefs being challenged.
Stasielowicz found that the average intervention effect was small but positive. This suggests that it is possible to reduce conspiracy beliefs, though the typical impact is akin to dimming a light rather than flipping a switch. The overall effect size was estimated at 0.16, which indicates a modest shift in belief rather than a transformative change. However, the data revealed substantial variability between different studies, implying that the quality and type of intervention matter significantly.
“One of the most striking findings was how different the results were across studies,” Stasielowicz said. “Some interventions worked quite well, whereas others barely changed beliefs or even seemed to strengthen conspiracy thinking. Thus, designing and deploying interventions needs to be done carefully to avoid backfiring and pushing people deeper into conspiratorial rabbit holes full of implausible narratives. A bad intervention can be worse than no intervention at all.”
Interventions that included fact-checking of specific conspiracy claims tended to be more effective than other approaches. The analysis suggests that scrutinizing details such as dates, timelines, and numbers directly associated with a conspiracy theory yields better results than general attempts to promote skepticism.
Similarly, the data indicates that providing alternative explanations for the events in question can enhance the persuasive power of the message. This suggests that simply debunking a claim is often insufficient unless a factual narrative replaces the conspiratorial one.
The degree to which the intervention content matched the outcome measure also played a role. Stasielowicz found that effects were larger when the arguments presented in the intervention directly addressed the specific beliefs measured in the questionnaire. For example, countering myths about vaccines proved effective for reducing vaccine-related conspiracy beliefs but would likely have little impact on beliefs regarding government surveillance. This finding supports the idea that belief change is often domain-specific rather than a general shift in mindset.
“The main takeaway is that it is possible to reduce conspiracy beliefs, but the average impact of current interventions is modest rather than transformative,” Stasielowicz explained. “Approaches that directly fact-check specific conspiracy claims and carefully examine dates, timelines, quotations, numbers, and internal contradictions tend to be more effective than approaches such as ridicule or the teaching of general scepticism rules.”
“However, we must be careful. Since not every conspiracy theory is necessarily wrong, public institutions, influencers, and researchers need to avoid automatically dismissing discussions as conspiracy theorizing and instead focus on clearly documenting what is and is not supported by evidence.”
The characteristics of the sample population appeared to influence susceptibility to these interventions. The results showed that interventions tended to be more effective among students compared to the general population. Younger participants also appeared more willing to revise their beliefs than older individuals.
Additionally, the analysis provided some evidence that samples with a lower proportion of college graduates showed larger reductions in conspiracy beliefs. This might suggest that individuals with higher levels of formal education hold their non-standard beliefs more rigidly, or perhaps that the specific interventions used were better suited for general audiences.
The measurement tools used by researchers also affected the observed outcomes. Studies that employed longer questionnaires with multiple items to assess conspiracy beliefs tended to detect larger effects than those relying on single-item measures. This implies that nuanced changes in attitude might be missed by overly simplistic survey instruments. The analysis also indicated that interventions were slightly more effective when the baseline level of conspiracy belief was moderate to high, rather than low.
“As we know from everyday experiences, belief change is hard, and conspiracy beliefs are no exception,” Stasielowicz told PsyPost. “Interventions work like dimming a light rather than completely switching it on and off; they facilitate belief change rather than completely transforming people’s minds. They are unlikely to fully convince firm believers to completely reject conspiracy theories, but even slightly lowering beliefs in implausible conspiracy theories may matter at the population level for outcomes like vaccine uptake or peaceful transfer of power after elections. Still, these modest effects highlight how much room there is to improve interventions.”
Stasielowicz identified several limitations within the existing body of research. A significant portion of the included studies were classified as having a high risk of bias. This is largely because it is difficult to blind participants to the purpose of an intervention when they are being presented with arguments against specific conspiracy theories. If participants guess the goal of the study, they may alter their responses.
The meta-analysis also highlighted that many studies were underpowered, meaning they did not include enough participants to reliably detect small effects. This lack of statistical power can lead to an overestimation of effects in published literature.
Additionally, the vast majority of studies only measured immediate changes in belief. There is a scarcity of data regarding whether these reductions in conspiracy thinking persist over days, weeks, or months. The researcher notes that without longitudinal data, it is difficult to determine if these interventions provide lasting benefits or merely temporary shifts in opinion.
“A common misunderstanding is to treat the ‘average effect’ as if it describes every study, when in reality, some interventions helped more effectively than others,” Stasielowicz noted. “Many existing studies focus on short-term impact, are often underpowered, and are at high risk of bias. More high-quality, long-term studies are needed, especially on specific design features that so far have been examined only in a handful of experiments, such as the use of LLM chatbots to reduce conspiracy beliefs.”
“A key long-term goal is to identify which specific intervention components actually drive belief change so that future interventions can be shorter, less costly, and more tailored. By combining this meta-analysis with my review of the characteristics of people who believe in conspiracy theories, it should be possible to test whether certain groups benefit more from particular intervention strategies and whether personalization adds value.”
“The article is open access, so anyone – including practitioners, journalists, and interested readers – can read the comprehensive results for free,” the researcher added. “The study offers concrete guidance on how to design the most promising interventions, for example, by tailoring messages to specific conspiracy claims.”
The study, “The Effectiveness of Interventions Addressing Conspiracy Beliefs: A Meta-Analysis,” was authored by Lukasz Stasielowicz.
Leave a comment
You must be logged in to post a comment.