The relationship between digital media and democracy is a subject of intense debate, with conflicting findings hindering progress. This study aims to synthesize existing research to provide an evidence-based understanding of this complex relationship. The dual-use dilemma, where technology serves both positive and negative purposes, is central to this issue. Historical examples, such as the use of radio during World War II and the Rwandan genocide, illustrate this. Digital media presents a similar dichotomy: it can empower citizen movements (Arab Spring, Fridays for Future, #MeToo) while simultaneously fueling destructive behaviors (polarization, populism, the January 6th Capitol attack). The review focuses on the pre-registered question: "If, to what degree and in which contexts, do digital media have detrimental effects on democracy?" The study acknowledges the predominance of correlational evidence due to the limitations of experimental designs in studying democracy. A two-step approach is used: synthesizing a broad set of articles examining associations, then conducting an in-depth analysis of the subset reporting causal evidence. The review adopts a broad definition of digital media, encompassing various platforms and content types, while focusing on political attitudes and behaviors relevant to democratic principles. The review specifically examines variables like trust in institutions, political participation, political knowledge, and phenomena such as populism and polarization. While acknowledging the nuances within each variable, the review classifies effects as beneficial or detrimental to democracy. Existing systematic reviews focus on specific aspects (radicalization, polarization, hate speech, participation, echo chambers); this review integrates these findings into a comprehensive analysis.

Several systematic reviews have examined subsets of the relationship between digital media and political behaviors, focusing on issues like radicalization, polarization, hate speech, political participation, and echo chambers. However, these reviews did not integrate a broad range of politically relevant variables into one analysis. This study aims to fill that gap by providing a comprehensive synthesis of the existing literature, contrasting and integrating evidence across diverse political variables, and examining different methodologies and their limitations. The review contrasts correlational and causal evidence, providing a nuanced perspective on the complex interplay between digital media and democracy. In particular, it addresses the lack of comprehensive analysis across multiple politically relevant variables, and seeks to identify research gaps through a comparison of the variables and methods employed across existing studies.

This systematic review followed the MOOSE guidelines. A pre-registered protocol, including the research question and search strategy, was available on the Open Science Framework (OSF). The search was conducted on Scopus and Web of Science databases. The search query included keywords related to digital media (social media, online news, Twitter, Facebook), political outcomes (polarization, participation, knowledge, attitudes), and methodologies (causal and descriptive methods). The initial search yielded 3,509 articles. A two-stage selection process was conducted, starting with title screening, followed by abstract-based screening, resulting in 1,349 potentially relevant articles. Subsequent full-text screening and the application of inclusion/exclusion criteria (detailed in Supplementary Information) resulted in a final sample of 496 articles. Inter-coder reliability was assessed at multiple stages of the coding process. Articles were coded based on several factors: types of digital media, political variables, content features (misinformation, selective exposure), methods used, and reported directions of associations. Due to heterogeneity in methods and measures, a meta-analysis was not conducted. Instead, the results are presented in a descriptive way, visualizing patterns of associations between digital media use and various political variables, categorized into those that are beneficial or detrimental to democracy. For articles using causal inference methods (matching, instrumental variables, panel designs, field experiments), an in-depth analysis was conducted to summarize findings using simplified path diagrams. Geographical distribution of effects was examined, considering established and emerging democracies/authoritarian regimes. Sampling methods and potential sources of bias were also examined, including sample size, sampling strategy (behavioral data, probabilistic samples, randomized control trials), transparency indicators (competing interests, open data practices, pre-registration). The study notes limitations of the method relating to data heterogeneity, variable categorization, inability to perform quantitative comparisons, unequal numbers of studies for different outcome variables, and limitations associated with data collection and analysis from the selected studies.

The review analyzed 496 articles, finding predominantly beneficial associations between digital media use and political participation (e.g., voting, protesting) across diverse contexts. However, the effects were mostly concentrated in emerging democracies and authoritarian regimes. For established democracies, the associations largely indicated that digital media use was detrimental to democracy. A total of 354 associations were reported for the top ten most frequently studied political variables. These associations were categorized as beneficial or detrimental to democracy. The findings revealed overwhelmingly detrimental associations between digital media use and trust in institutions (government, politics, media, social, and institutional trust). Similarly, detrimental associations were observed with hate speech, political polarization, and populism, particularly among established democracies. Digital media use was widely linked to increased exposure to misinformation. The review also notes that digital media use is linked to increased homophily (like-minded connections) in social networks, mostly observed on social media, especially Twitter. The analysis separated correlational and causal evidence. The 24 articles providing causal evidence largely supported the trends identified in the correlational evidence. Positive associations of digital media use with participation and political knowledge were found in both emerging and established democracies, though less consistently for knowledge. Detrimental associations with trust, polarization, and populism were consistent across both types of evidence. Notably, the study finds that the beneficial effects are more prevalent in emerging democracies and authoritarian regimes, while detrimental effects are more prominent in established democracies. This highlights the context-dependency of the effects of digital media on democracy. An analysis of the sampling methods indicates a tendency for behavioral data to be more prevalent for studies focusing on detrimental outcomes, while surveys were more common for beneficial outcomes. The review suggests a need to combine both types of data to improve reliability and address potential limitations of each method. The study found limited evidence of systematic bias due to publication bias, temporal trends, or specific authors. The review also noted several research gaps, such as the under-exploration of specific regions and method-variable combinations, as well as the need for more research on specific digital media platforms.

The findings indicate a clear association between digital media use and key variables crucial to democratic functioning. While digital media can enhance participation and knowledge, the detrimental effects on trust in institutions, the rise of populism and polarization, and increased exposure to misinformation pose significant threats to democracy. The double-edged nature of digital media is emphasized, with effects varying depending on the political system. The review highlights the need for more causal evidence, particularly regarding the apparent contradiction between diverse news exposure and increased homophily in social networks. The study emphasizes the context-dependency of the effects of digital media on democracy, urging caution against generalizing findings from specific political contexts, such as the United States, to other settings. Methodological advancements that link different data types are crucial for understanding this complex relationship. The review underscores the significant influence of large commercial platforms on the current information ecosystem. Their control over information flow impacts political attitudes and behaviors. The study suggests that future research and policy should address this power imbalance and prioritize open access to data, potentially necessitating regulation to enable independent researchers to fully understand the impact of digital media on democracy.

This systematic review reveals a complex and context-dependent relationship between digital media and democracy. While digital media can enhance political participation and knowledge, they also present significant threats, particularly through their impact on trust in institutions and the rise of polarization and populism. Addressing the detrimental effects while harnessing the potential benefits requires further research, methodological innovation, and potentially regulatory interventions to ensure open access to data and a more balanced information ecosystem. Future research should focus on bridging research gaps, including refining methodologies and expanding research to understudied regions and populations.

The broad scope of this review necessitated the categorization of variables and effects into broader categories, limiting the granularity of analysis and preventing quantitative comparisons. The unequal number of studies examining different outcome variables also presents limitations. The interpretation of results was sometimes hindered by the lack of appropriate baseline measures to compare digital media effects to offline behaviors and norms. The study's reliance on published literature might be susceptible to publication bias. While efforts were made to address this, the possibility of a file drawer problem remains. Additionally, restrictions on data access from platforms pose a significant challenge for future research.