Benkler, Y. R., Farris, R., & Roberts, H. (2018). Network propaganda: Manipulation, disinformation, and radicalization in American politics. Aquila Digital Community (University of Southern Mississippi).
Summary
Network Propaganda is a large-scale empirical investigation into the sources of disinformation and epistemic dysfunction in American political communication, centered on the 2016 election and the first year of the Trump presidency. Drawing on roughly four million political stories analyzed over three years, Benkler, Faris, and Roberts argue that the crisis is fundamentally asymmetric and partisan — driven by a radicalized right-wing media ecosystem rather than by technology, Russia, or Facebook alone. Their central analytical device, the propaganda feedback loop, explains how the right-wing media sphere became insulated from truth-correcting journalistic norms and thus uniquely susceptible to lies and half-truths. The book adopts a “political economy view of technology,” insisting that institutions, culture, and politics — not algorithms — shape outcomes.
Key Contributions
- Introduces the concepts of network propaganda and the propaganda feedback loop as frameworks for how media-ecosystem architecture shapes susceptibility to manipulation.
- Provides data-driven evidence for the asymmetric (rather than symmetric) nature of American media polarization, challenging both “the internet polarizes” narratives and false-equivalence framings.
- Offers revived, precise definitions of propaganda, disinformation, misinformation, bullshit, and disorientation, rooted in the intellectual history of propaganda studies.
- Supplies a transferable, institution-sensitive method for empirically studying any country’s political media ecosystem.
- Reframes disparate political events as elements of a single conflict between Trumpism and Reaganism over control of the Republican Party.
Methods
Computational analysis of ~4 million political stories (April 2015–March 2018) using hyperlink analysis plus Twitter and Facebook sharing data. The authors built on the Media Cloud platform for collection and parsing, used Gephi for network mapping and visualization, and integrated TV Archive and GDELT data. Findings are grounded through case studies (immigration/Islamophobia framing, Fox News coverage, the propaganda pipeline) and cross-checked against polling data, existing reports, and academic literature with multiple robustness specifications.
Findings
- Influence in the right-wing ecosystem is highly skewed to the far right and insulated from the rest of the network; no comparable asymmetry exists on the left.
- Left-leaning sites are tightly integrated with mainstream, norm-abiding outlets that serve a corrective function absent on the right.
- Mainstream 2016 coverage was dominated by horserace and negativity, largely adopting the agenda and framing of the right-wing media and the Trump campaign.
- Facebook appears more polluted than Twitter or the open web, with Facebook-prominent sites more prone to false and hyperpartisan content on both sides.
- Radical sites (Breitbart, Infowars, Gateway Pundit) drove radicalization independently of decentralized alt-right networks, which were relatively isolated.
- Disinformation campaigns targeted even core Republican figures (e.g., the smearing of Jeb Bush), justifying “radicalization” over “polarization.”
- The asymmetric architecture predates Trump, tracing back to Rush Limbaugh’s 1988 talk-radio model; Trump acted as “catalyst in chief.”
Connections
This book is a foundational statement of the asymmetric polarization thesis and directly informs empirical work on partisan media diets and misperceptions such as Osmundsen2021-et and Iyengar2019-jj. Its ecosystem-level, cross-platform mapping anticipates later network and sharing studies of disinformation flows like Freelon2020-yp, Starbird2019-qv, and Vosoughi2018-at, while its “political economy of technology” stance offers a corrective counterpoint to the algorithm- and echo-chamber-centric accounts examined in Bakshy2015-rn, Flaxman2016-lm, Bail2018-fk, and Guess2023-ai. Its cross-national argument that the same technologies yield different effects also resonates with comparative resilience work such as Humprecht2020-gd.