Computational Propaganda and the Lie Machines

Philip Howard

In brief

Philip Howard discusses his book The Lie Machines and describes computational propaganda as an industrialized system combining organized disinformation campaigns (social component) with social media algorithms (technical component). He outlines three production stages — creation, dissemination, and long-term marketing — and argues that the threat to democracy comes not just from false claims but from systematic noise designed to degrade collective sense-making and voter selection.

Summary figure for Computational Propaganda and the Lie Machines

Key Contributions

The lie machine framework: computational propaganda integrates human operatives (troll farms, PR firms, lobbyists) with platform algorithms to deliver targeted political misinformation at scale.
Three-stage production pipeline: production (message crafting and narrative building), dissemination (paid ads and organic amplification via bots), and marketing (long-term story sustenance and relaunch cycles).
Broader definition of political lies: beyond false factual claims, includes content designed to provoke unwarranted skepticism, stoke negative emotions, discourage dialogue, or insert doubt without explicit falsehood (e.g., asking misleading questions framed as legitimate inquiry).
Shift from targeting to suppression: many state actors (Russia, China, Iran) use computational propaganda not to elect specific candidates but to confuse voters, polarize discourse, and degrade a target nation's ability to make evidence-based decisions.
Data monopoly as a structural problem: private tech firms control the highest-quality behavioral and political data; breaking this monopoly and enabling data sharing with independent researchers and civil society is foundational to defense.

Method and Examples

Howard draws on verified datasets (including 3,500 Russian fake accounts from the 2016 U.S. election provided by Facebook to the U.S. Senate), academic research at Oxford Internet Institute, and publicly documented campaigns. He discusses:

The evolution from broad direct-mail targeting (Reagan-era micro-targeting via punch-card databases) to real-time algorithmic A/B testing of thousands of ad variants simultaneously, enabling speed and scale previously impossible.
Visual misinformation and out-of-context images used to amplify political narratives without explicit lies (e.g., pictures of Colorado labs paired with questions about COVID-19 origins, asked by Chinese state media without claiming the virus was lab-created).
Organized fake-account networks maintained by governments (Russia, China) and PR firms (examples in Poland and Brazil), used to manufacture consensus on social platforms and infiltrate publications with fake expert personas (e.g., 19 fake geopolitical commentators placed in Newsmax and Washington Examiner opinion sections).
Cost democratization: modern political ad campaigns on social platforms cost a fraction of pre-internet efforts, enabling more actors (candidates, lobbyists, foreign governments) to participate but also allowing extremist and sensational content to dominate for engagement.

Proposed Solutions

Howard proposes structural and institutional reforms:

Data accessibility: share anonymized behavioral and political data from tech platforms with libraries (Library of Congress, British Library), journalists, and civil-society researchers so independent actors can study and counter disinformation.
Device accountability: rebuild consumer devices to transparently report who benefits from users' behavioral data and allow citizens to add organizations (researchers, charities, politicians) to the list of approved data beneficiaries.
Subscription-based platforms: convert social media from ad-driven (incentivizing engagement and controversy) to subscription (fee-based, aligning user and platform interests). Howard argues this could work at $1–2/month per user, enabling competition and reducing algorithmic manipulation.
Collective sense-making: organize community groups (e.g., "Rally Point Alpha" on Facebook) to collaboratively evaluate and verify information, treating information verification as a social activity rather than an individual burden.
Public broadcasters: countries with well-resourced public broadcasters (BBC, CBC, ABC) show better resilience to misinformation because professional editorial gatekeeping reduces signal-to-noise ratios.

Notes

The episode provides an accessible, well-sourced introduction to computational propaganda for general audiences. Howard's framing of misinformation as an industrial system (rather than individual false claims) shifts focus from fact-checking to infrastructure and incentives — a perspective increasingly central to information-integrity research. The emphasis on long-term narrative sustenance and cross-platform recycling (chans → Reddit → mainstream media → political repetition) complements graph-based and temporal studies of misinformation spread.

The proposed solutions are ambitious (requiring regulatory and structural change) but grounded in concrete examples. The subscription-based platform argument is particularly provocative: aligning incentives rather than regulating content is a distinct approach from platform moderation.

Howard's mention of data monopoly and the need for independent research access echoes ongoing debates in the AI safety and data-governance communities, though computational-propaganda-specific applications remain under-explored.

How it connects

Discussed in GUIDE by

Network Propaganda Benkler