Media literacy tips promoting reliable news improve discernment and enhance trust in traditional media

The study addresses whether media literacy tips that promote trust in reliable information can improve news discernment without causing undue skepticism toward true news. Concerns about misinformation have driven interest in scalable, individual-level interventions, yet many such efforts may inadvertently increase skepticism of accurate news. This is problematic because most news people encounter in Western democracies is true, while general trust in news is low and news avoidance is rising. Misperceptions may often stem more from a lack of exposure to reliable information than from belief in misinformation. The authors therefore propose that reducing misperceptions requires both countering misinformation and promoting reliable information. The study makes two contributions: (1) testing three distinct media literacy interventions (skepticism-focused, trust-focused, and mixed) on discernment; and (2) experimentally manipulating the base rate of true versus false news to evaluate how ecological validity affects intervention efficacy.

The paper situates interventions against misinformation in two broad categories: reactive (e.g., fact-checking and labels) and proactive (e.g., inoculation, accuracy prompts, media literacy). Proactive interventions are valued for supporting autonomy and engagement while respecting free press, but much current work disproportionately targets misinformation at the expense of promoting reliable information. Existing media literacy tips often raise awareness of misinformation and encourage skepticism (e.g., Facebook's 2017 tips), which can reduce perceived accuracy of false headlines but sometimes also of mainstream or accurate headlines. Studies report possible negative spillovers where skepticism-inducing interventions reduce acceptance of true news and trust in reliable actors. Longstanding concerns in media literacy warn of fueling cynicism. Methodological improvements suggest measuring effects on both misinformation and reliable information and using discrimination measures that account for response biases. Most prior studies test with a 50/50 mix of true and false items, which diverges from real-world base rates (roughly 5% false, 95% true). This raises concerns that balanced test sets might overstate the efficacy of skepticism-focused interventions and understate benefits of trust-promoting approaches. The authors note prior attempts to adjust effects post hoc but argue for experimental manipulation of the proportion of false to true news to identify causal impacts.

Design: Pre-registered survey experiment with US participants recruited via Prolific between Aug 22–31, 2023. Total N = 3919 (balanced by gender and political orientation); 5 failed attention checks removed. Median completion time ~6 minutes; compensation $1.12. Conditions: Random assignment to one of four conditions: Trust tips (N = 963), Skepticism tips (N = 983), Mixed tips (N = 967), or Control (no tips). Tips were pre-tested; Trust tips rated most positive/trust-enhancing, Mixed intermediate, Skepticism least positive. All tips rated easy to read, informative, and moderately convincing. Stimuli and tasks: Participants viewed 12 political headlines formatted as Facebook posts (headline, image, source). Headlines drawn from a pre-tested pool of 40; true items from mainstream US outlets, false items from fact-checkers (e.g., PolitiFact, Snopes). Dependent variable manipulated between subjects: either accuracy ratings (1 = certainly false to 6 = certainly true) or sharing likelihood (1 = extremely unlikely to 6 = extremely likely). Base-rate manipulation: Participants were randomly assigned to one of three true/false mixes: 75% true (8 true, 4 false), 50% true (6/6), or 25% true (4/8). Pre/post measures: Pre-treatment measures included political news interest, trust in groups of news outlets (mainstream left-leaning; mainstream right-leaning; generally untrustworthy left; generally untrustworthy right), trust in institutions (traditional media, social media, journalists, scientists), social media use, and self-reported media literacy (4 items). Post-treatment measures repeated interest and trust items. Participants also provided self-assessed performance; overconfidence computed as self-assessed minus actual performance. Demographics (age, gender, political identification) were obtained via Prolific. Ethics approval: University of Zurich PhF Ethics Committee (nr. 23.04.17). Materials, survey, data, and code are available on OSF: https://osf.io/73y6c/. Statistical analysis: Alpha = 0.05. Primary analyses at response level (46,992 observations) used linear mixed-effects models with random effects for participants and headlines, controlling for age, gender, political orientation, proportion of true news, response type (accuracy vs sharing), and condition. Attitudinal outcomes analyzed at participant level via OLS regressions. Likert items treated as continuous. Robustness checks included a Poisson GLMM reproducing main results. Exploratory analyses assessed heterogeneous effects and determinants of discernment.

Discernment (H1): All three tip conditions significantly increased overall discernment (true minus false ratings) compared to control. General discernment effects: Mixed b = 0.23 [0.17, 0.29], p < 0.01; Trust b = 0.22 [0.16, 0.29], p < 0.01; Skepticism b = 0.19 [0.13, 0.25], p < 0.001. Effects held for both sharing and accuracy outcomes. Between-treatment differences in general discernment were not statistically significant, except Trust was slightly less effective than Mixed for accuracy discernment (b = -0.09 [-0.17, -0.01], p = 0.025). True vs false ratings (H2): Trust tips increased ratings of true news (b = 0.12 [0.04, 0.21], p = 0.004) and had no significant effect on false news (b = -0.04 [-0.11, 0.03], p = 0.27). Skepticism tips reduced ratings of false news (b = -0.10 [-0.16, -0.027], p = 0.006) and did not significantly affect true news (b = 0.067 [-0.02, 0.15], p = 0.11). Mixed tips both increased true news ratings (b = 0.10 [0.02, 0.19], p = 0.018) and reduced false news ratings (b = -0.07 [-0.14, -0.01], p = 0.034). Exploratory contrasts found no significant differences among treatments for effects on true or false news. Base-rate moderation (H3): Contrary to preregistered expectations, Trust and Skepticism tips were most effective at increasing discernment under a 50/50 true-false mix, largely via increases in true news ratings. Mixed tips were most effective under 75% true and 25% false, mostly via reductions in false news ratings. Three-way interactions: Trust more effective at 50% true than 75% true (b = 0.25 [0.11, 0.40], p = 0.001). Mixed more effective at 75% true than 25% true (b = 0.18 [0.021, 0.34], p = 0.027). Other contrasts not significant; authors note limited power for small three-way effects. Attitudes (H4–H6): No significant effects on interest in news, trust in journalists, or trust in specific outlet groups, except Trust tips increased trust in traditional media vs control (b = 0.08 [0.02, 0.14], p = 0.008) and more than Skepticism (b = 0.092 [0.03, 0.15], p = 0.003) and Mixed (b = 0.076 [0.02, 0.14], p = 0.014). Exploratory: Tips tended to increase confidence and reduce overconfidence slightly, but effects were small and inconsistent. More exposure to false news increased confidence in recognizing made-up news without affecting overconfidence. Heterogeneous effects: Skepticism more effective among Independents (b = 0.34 [0.19, 0.48], p < 0.001); Mixed more effective for men than women (b = 0.15 [0.03, 0.28], p = 0.014); Skepticism more effective for women than men (b = 0.13 [0.01, 0.25], p = 0.042); Mixed and Skepticism slightly less effective among heavier social media users (b = 0.06 [-0.11, -0.01], p = 0.02). Determinants of discernment: higher political news interest, higher self-reported media literacy, older age, being female, and identifying as Democrat (vs Independent) predicted greater discernment.

The findings demonstrate that brief media literacy tips can increase discernment between true and false news across both accuracy judgments and sharing intentions. Trust-focused tips preferentially increased acceptance of true news; skepticism-focused tips reduced acceptance of false news; and mixed tips achieved both, though effect sizes were similar across treatments. Importantly, there was minimal evidence of adverse spillovers on acceptance of true news or broad trust metrics, with the exception that trust-focused tips slightly increased trust in traditional media. These results suggest platforms and policymakers need not focus exclusively on misinformation; promoting trust in reliable information is also effective. The relative emphasis on skepticism versus trust should be context-dependent—greater focus on trust in environments where misinformation is rare but news trust is low, and greater skepticism where misinformation is more prevalent. Base-rate manipulations indicate mixed tips may align well with more ecologically realistic environments (75% true), while trust and skepticism tips showed strongest effects under 50/50 conditions. Given limited power for complex interactions, further testing is warranted. Overall, tips should complement longer-term literacy programs and systemic solutions, and be targeted to content or contexts (e.g., pairing trust tips with reputable sources, skepticism tips with suspect sources).

Media literacy tips can enhance news discernment without requiring exclusive emphasis on skepticism. Interventions that promote trust in reliable information, skepticism toward misinformation, or both, all improved discernment, with mixed strategies achieving increases in true news acceptance and reductions in false news acceptance. Given the low prevalence of misinformation in many Western contexts and concerns about cynicism, organizations and platforms should avoid exclusively skepticism-inducing tips and instead combine skepticism with promotion of reliable information. Future research should: (1) replicate these findings in larger and more diverse samples; (2) test interventions under more ecologically valid conditions (realistic base rates and content mixes); (3) unbundle and isolate specific tips to identify the most effective components and mechanisms; and (4) tailor tip deployment to information environments and audience subgroups.

The study did not identify the precise mechanisms by which tips improved discernment. While priming of trust or skepticism is a plausible explanation consistent with observed specificity (e.g., trust tips increasing traditional media trust), it does not fully align with base-rate moderation expectations. Given the brevity of the interventions and evidence that one-shot effects may be short-lived, durable learning is unlikely. The experiment may have been underpowered to detect small three-way interaction effects. Findings should be replicated with larger samples and more diverse populations, and future work should experimentally test mechanisms, including unbundling tips to assess which components drive effectiveness.