An optimisation model for minimising fake news dissemination: the role of user demographics, content characteristics, and intervention strategies

This study investigates how user characteristics and content formats influence fake news dissemination and proposes an optimisation framework for its minimisation. Logistic regression analysis reveals that age and education are statistically significant predictors of fake news sharing behaviour, whereas gender has no significant effect. Older individuals and those with higher educational qualifications are less likely to share fake news. Political affiliation significantly increases the propensity to disseminate misinformation. The study further examines content-type virality, finding that image-based fake news spreads at significantly higher rates than text or video formats. To identify actionable countermeasures, topic modelling is applied to triangulated data comprising social media discourse, expert commentary, and published literature, yielding five platform-level strategies: algorithmic demotion, content labelling, fact-checking partnerships, user education, and incentive reduction and three governmental strategies: legislation, transparency and accountability mechanisms, and independent fact-checking with international cooperation. These findings are integrated into an optimisation model that minimises expected fake news sharing across intervention levers under budget, capacity, and feasibility constraints, providing social media platforms and government agencies with a prescriptive resource-allocation tool to mitigate misinformation dissemination. This study addresses a critical operational gap: the absence of empirically grounded frameworks for allocating finite resources to counter fake news dissemination. The observational data reveal that younger age, lower educational attainment, and political affiliation significantly increase the propensity to share fake news. Gender, however, exerts no significant effect. These demographic patterns provide practitioners with an empirical risk-stratification taxonomy for targeting media literacy interventions toward the most vulnerable segments. Further analysis shows that image-based fake news propagates at significantly higher rates than text- or video-based formats. Topic modelling applied to social media data, including expert opinions and published literature, yields both platform-level and government-level strategies. The principal operational contribution of this study is optimisation model that integrates these empirical inputs. The model estimates baseline sharing probabilities for each user segment using logistic regression coefficients, assigns weights to content types based on empirically derived virality coefficients, and minimises expected fake-news sharing across five intervention decision variables, subject to budget, capacity, and feasibility constraints. Because effectiveness and cost parameters are exogenous, practitioners can calibrate the model through A/B testing or pilot programs, thereby transforming descriptive findings into operationally actionable resource-allocation prescriptions.