CHAPTERS
- 0:00 – 1:30
Renée’s entry point: anti-vaccine communities and early signs of algorithm gaming
Renée DiResta explains how her research began in 2015 while investigating anti-vaccine activism in California as a new parent. She describes how small groups can disproportionately amplify narratives by learning to “game” platform algorithms and hashtags.
- •Motivation came from navigating preschool communities and vaccine norms
- •Observation: small coordinated groups can create outsized “share of voice”
- •Hashtag gaming and automation as early manipulation tactics
- •Distinguishing organic advocacy from coordinated amplification
- 1:30 – 2:31
From ISIS propaganda to state actors: realizing the information ecosystem is exploitable
Her work broadened from domestic health misinformation to terrorist propaganda, as researchers noticed ISIS using bots and cross-platform coordination to build a “digital caliphate.” This period helped frame manipulation as an ecosystem problem rather than a single-platform issue.
- •ISIS used small, coordinated networks to dominate narratives
- •Information hops between platforms in both coordinated and contagious ways
- •Policy questions: how should governments respond to propaganda on US platforms?
- •Vulnerability is structural—any actor can exploit it
- 2:31 – 4:18
How Russia entered the picture: early reporting on the Internet Research Agency (IRA)
DiResta connects the rise of concern about manipulation to 2015, when Adrian Chen’s reporting exposed the IRA in St. Petersburg. The conversation shifted from terrorism alone to whether nation-states were also running influence operations on US platforms.
- •Adrian Chen’s exposé as an early public milestone
- •Parallel tactics: propaganda, fake personas, coordinated narratives
- •Emerging goal: detect and stop manipulation regardless of the actor
- •Early bot/troll hunting began around 2015
- 4:18 – 7:40
Why consolidation matters: from the open web to a handful of dominant platforms
They discuss how the internet evolved from decentralized blogs and forums to consolidated social platforms with massive audiences. That concentration makes it easier for propagandists to focus efforts and achieve scale quickly.
- •Internet 1.0: speech without guaranteed attention
- •Social platforms created network effects and audience consolidation
- •A few companies became the primary infrastructure for public discourse
- •Consolidation reduces the cost of targeting and manipulation
- 7:40 – 9:52
Troll accounts, “cyborgs,” and the pivot from Crimea to American audiences
DiResta explains that many “bots” are actually semi-automated sock puppets run by real people. The IRA tested tactics around Crimea and MH17, then expanded outward to interact with Americans and mimic domestic trolling/harassment styles.
- •Not just bots: real operators + automation (“cyborg” behavior)
- •2013–2014: Russian/Ukrainian focus; Crimea and MH17 narratives
- •Late 2014/early 2015: pivot to US targets
- •Harassment and @-mentioning public figures to gain visibility
- 9:52 – 12:03
Shifting community norms: how trolls change tone without ‘swinging’ elections
They explore the idea that impact isn’t only about election outcomes—it’s also about altering the culture of online communities. Infiltration can normalize harsher language, expand tolerated behavior, and intensify polarization.
- •Critique addressed: impact is hard to measure, but tone shifts are real
- •Trolls operate inside communities and subtly reshape norms
- •Normalization of viciousness and new linguistic/behavioral baselines
- •VPNs and rapid account recreation reduce enforcement effectiveness
- 12:03 – 14:50
Inside the IRA: quotas, standups, microtargeted authenticity, and cultural training
DiResta cites whistleblowers and legal filings describing the IRA like a marketing shop fused with intelligence tradecraft. Operators had quotas, ran “standups,” and were coached on community-specific authenticity cues (e.g., imagery for niche identities).
- •Whistleblower accounts: high-volume posting targets and engagement quotas
- •Millennial workforce fluent in memes and internet culture
- •Mueller filings: standups with detailed targeting guidance
- •One-on-one relationship-building crosses into intelligence-style operations
- 14:50 – 19:20
Building ‘tribes’ across platforms: long-game community cultivation
Using Senate-supplied platform datasets, DiResta describes how IRA pages built identity-based communities (LGBT, Black communities, right-wing subtypes, etc.). Most posts were non-political rapport-building, with occasional prompts to reinforce in-group beliefs.
- •Cross-platform brand building (Facebook + Instagram + YouTube)
- •Identity-targeted communities: many sub-tribes and niche segmentation
- •Mostly cultural/affirmation content; politics introduced strategically
- •In-group persuasion: ‘as members of this tribe, we believe…’
- 19:20 – 26:41
Scale and rebranding tactics: sleeper accounts, repurposed pages, and meme experiments
They cover follower ranges, hit-or-miss page performance, and how operators repurposed failed accounts. DiResta’s examples (Kermit→Homer→Army of Jesus; Black Guns Matter→Anonymous) illustrate “sleeper” groundwork and rapid rebrands.
- •Long tail of small pages; a few reached hundreds of thousands of followers
- •Repurposing underperforming pages like a growth team would
- •‘Sleeper’ aging: making accounts look established before politicizing
- •Red flags: sudden pivots vs years-long groundwork
- 26:41 – 33:58
Election-era objectives: pro-Trump, anti-Clinton, depressing turnout, and polarizing both sides
DiResta lays out the political pattern: sustained pro-Trump messaging on right-leaning pages and strong anti-Clinton content across the board, with tactics to siphon or suppress votes on the left (e.g., Jill Stein promotion). The goal extended to eroding mainstream conservative figures and pushing audiences toward extremes.
- •Early flirtation with Rand Paul, then rapid pivot to Trump
- •No pro-Hillary posts found on Facebook/Instagram datasets
- •Left-leaning pages: anti-Trump messaging paired with pro–Jill Stein nudges
- •Right-leaning pages: pro-Trump plus anti-establishment GOP attacks
- 33:58 – 39:29
Measuring impact is hard: engagement stats, missing comment context, and platform differences
They discuss why quantifying persuasion is difficult: datasets show posts and engagement but not downstream effects or user reactions. DiResta provides platform-level counts and explains how Twitter differed (conversation hijacking) versus Facebook/Instagram (relationship building).
- •Key limitation: cannot see full response threads or offline effects
- •Platform totals: 10.5M tweets; 133 Instagram accounts; 81 Facebook pages; 17 YouTube channels
- •Instagram performance was especially strong, heavily meme-driven
- •Twitter: joining and nudging ongoing discourse more than tribe-building
- 39:29 – 56:00
Repurposing American content: copying local news, memes, and native vernacular to avoid detection
DiResta explains how linguistic tells (ESL errors, overly formal voice) helped identify fabricated posts—so operators increasingly copied authentic American text and memes. This made propaganda feel native while also laundering narratives through real cultural artifacts.
- •Vernacular often came from copied sources, not original writing
- •Original writing revealed grammar slips and cultural incongruities
- •Meme laundering: reposting Turning Point USA, Occupy Democrats, etc.
- •Authenticity camouflage: slap on new logos and redistribute
- 56:00 – 1:14:03
From online posts to offline action: funding activists, creating protests, and staged confrontations
The conversation escalates to real-world operations described in indictments: operators contacted activists, paid for materials, and used Facebook events and ads to mobilize opposing rallies—sometimes at the same place and time to provoke conflict.
- •Direct outreach via Messenger to real activists and supporters
- •Material support: money for posters, organizing help, paid trainers
- •Coordinated opposing rallies (e.g., Texas anti-Islam vs pro-Muslim)
- •Events show manipulation isn’t ‘just memes’—it’s operational
- 1:14:03 – 1:56:46
Platform accountability, moderation debates, and regulation: Section 230, ads transparency, and privacy
They turn to institutional responses: initial platform resistance, congressional hearings, user notifications, and growing pressure for oversight. The discussion spans Section 230, ad disclosure reforms, privacy (GDPR/California), identity verification proposals, and the difficulty of moderating at scale without political backlash.
- •2016–2017 shift: platforms went from denial to partial transparency
- •Notifying users who engaged with troll content (email and in-platform tools)
- •Section 230: right to moderate without obligation; hard balance to strike
- •Policy paths: Honest Ads Act, privacy regulation, and potential FTC oversight
- 1:56:46 – 2:07:54
The next wave: synthetic faces, deepfakes, VR, and disinformation as a real-world security risk
They conclude by looking forward: AI-generated personas (GAN faces), deepfakes, and immersive tech will make authenticity harder to verify. DiResta cites earlier IRA hoaxes (e.g., the Louisiana chemical plant story) to warn how future media fabrications could trigger panic or escalation beyond social media.
- •GAN-generated profile photos eliminate reverse-image red flags
- •Deepfakes and synthetic audio raise stakes for public figures and citizens
- •VR/AR could intensify persuasion and blur reality further
- •Hoax operations show a path from online deception to real-world crises
