L1
·
Quiz
·
Lab
L2
·
Quiz
·
Lab
L3
·
Quiz
·
Lab
L4
·
Quiz
·
Lab
Module Test
Lesson 1 · AI & Misinformation

Synthetic Media and the Deepfake Era

How AI-generated audio, video, and images became tools for large-scale deception
When a video of a real person says something they never said — what breaks first, truth or trust?

On January 1, 2019, the government of Gabon released a video of President Ali Bongo Ondimba delivering a New Year's address. Bongo had been absent from public view for months following a stroke. Critics immediately claimed the video was an AI-generated deepfake. Within days, military officers launched a coup attempt, citing the fabricated video as evidence the president was incapacitated or dead. Independent analysts later disputed whether it was actually a deepfake — but the belief that it might be was enough to destabilize a government. The misinformation ecosystem had matured to the point where the accusation of fakery was as dangerous as fakery itself.

What Is a Deepfake?

The term deepfake emerged around 2017, combining "deep learning" with "fake." Early examples used generative adversarial networks (GANs) to map one person's face onto another's body in video. By 2020, the technology had advanced far enough that consumer-grade tools could produce convincing face swaps in minutes.

Deepfakes are not limited to video. Voice cloning tools can replicate a speaker's vocal pattern from as little as three seconds of audio. Image synthesis models like Stable Diffusion and Midjourney generate photorealistic scenes that never occurred. Text generation models produce fake quotes, fake news articles, and fake social media posts at scale.

DeepfakeAI-generated or AI-manipulated media in which a real person's likeness, voice, or words are fabricated or replaced without consent.
GANGenerative Adversarial Network — two neural networks (generator and discriminator) trained against each other to produce increasingly realistic synthetic outputs.
Voice CloningTechnology that learns the acoustic patterns of a specific person's speech to generate new audio in their voice.

The Scale Problem: Cheap, Fast, Everywhere

In 2018, producing a convincing deepfake video required a powerful GPU, technical skill, and hours of processing time. By 2023, multiple mobile apps could do the same in under a minute. The cost of deception collapsed. This is sometimes called the democratization of synthetic media — but democratization of a dangerous capability is itself a danger.

Documented scale: A September 2023 report by the cybersecurity firm Reality Defender found that deepfake attempts against financial institutions had increased 700% year-over-year. The AI company Sensity AI catalogued over 85,000 deepfake videos online as of 2023, with the number doubling roughly every six months.

Real Event · Slovakia Election, September 2023

Two days before Slovakia's parliamentary election, an audio recording circulated on Facebook appearing to feature liberal candidate Michal Šimečka discussing how to rig the vote by buying votes from the Roma minority. Meta labeled it as potentially AI-generated. Independent audio analysts agreed it showed signs of manipulation. Šimečka's party lost narrowly. Whether the fake audio swung the election is disputed — but it reached hundreds of thousands of voters in a news blackout period when candidates were prohibited from responding publicly.

Why Humans Struggle to Detect Synthetic Media

Human visual cognition evolved to detect threats in motion — predators, falling objects, other humans. It did not evolve to detect GAN artifacts in facial textures at 30 frames per second. Studies by MIT's Media Lab have shown that humans correctly identify deepfake videos only about 50% of the time — no better than chance.

The liar's dividend compounds this problem: even when media is genuine, bad actors can claim it's fake. The very existence of deepfake technology gives plausible deniability to real evidence. A genuine video of a politician accepting a bribe can be dismissed as "probably AI."

Key Concept · The Liar's Dividend

Coined by law professors Robert Chesney and Danielle Citron in 2019: as deepfakes become more common and more believable, anyone can claim any authentic video or audio is fake. The technology undermines trust in real evidence, not just in fabricated evidence.

Detection Technology: An Arms Race

Detection tools do exist. Microsoft's Video Authenticator, launched in 2020, analyzed subtle blending artifacts at pixel edges. Intel's FakeCatcher in 2022 used photoplethysmography — detecting the subtle color changes in skin caused by blood flow, which synthetic video cannot replicate. But every time detection improves, generation models are retrained on detected examples and improve further. It is a genuine arms race, and the offense currently leads the defense.

Content provenance standards offer a different approach. The Coalition for Content Provenance and Authenticity (C2PA), backed by Adobe, Microsoft, Google, and others, embeds cryptographic metadata into media at the point of creation — a kind of digital certificate of authenticity. Adoption, however, remains partial.

Lesson 1 Quiz

Synthetic Media and the Deepfake Era · 4 questions
1. The term "deepfake" combines which two concepts?
Correct. The term originated around 2017, combining "deep learning" — the AI technique used — with "fake," describing the fabricated output.
Not quite. "Deepfake" specifically refers to deep learning (neural networks) combined with fake media production.
2. What is the "liar's dividend" as described by Chesney and Citron?
Correct. The liar's dividend means deepfake technology harms trust even in real evidence — anyone can claim genuine video is AI-generated.
The liar's dividend is about how the existence of deepfakes lets bad actors dismiss authentic evidence as fake — undermining trust in real media.
3. In the 2023 Slovakia election case, what form did the AI-generated misinformation take?
Correct. A fake audio clip of Michal Šimečka circulated on Facebook two days before the election, during a legally mandated campaign blackout period.
In Slovakia 2023, the misinformation was an audio recording — voice cloning, not video — appearing to feature liberal candidate Michal Šimečka.
4. What approach does C2PA (Coalition for Content Provenance and Authenticity) use to combat synthetic media?
Correct. C2PA uses cryptographic metadata — a digital certificate embedded at creation — so platforms and viewers can verify a piece of media's origin and authenticity chain.
C2PA embeds cryptographic provenance metadata at the moment of creation, creating a verifiable chain of authenticity that travels with the media file.

Lab 1 · Deepfake Detection Reasoning

Practice identifying the signals that separate synthetic from authentic media

Your Mission

You are analyzing media for a fact-checking organization. Use this AI assistant to work through the detection signals — visual artifacts, audio anomalies, provenance gaps — that indicate synthetic or manipulated content.

Ask about specific detection techniques, the Gabon or Slovakia cases, how GANs leave artifacts, or how you would verify a suspicious video in practice.

Try: "What visual signs should I look for in a suspected deepfake video?" or "How did analysts determine the Slovakia audio might be fake?"
Detection Lab Assistant
AI & Misinformation M3
Welcome to the Deepfake Detection Lab. I'm here to help you develop practical skills for identifying AI-generated or manipulated media. Ask me about visual artifacts, audio tells, provenance verification, or specific documented cases. What would you like to explore?
Lesson 2 · AI & Misinformation

AI-Powered Disinformation at Scale

Automated propaganda networks, influence operations, and the industrialization of false narratives
When a single operator can generate ten thousand convincing fake personas overnight, what does "public opinion" actually mean?

In 2019, the EU DisinfoLab uncovered a network of over 750 fake media outlets in 116 countries, running since 2005, coordinated from India and designed to support Pakistani isolation at international institutions. The network fabricated quotes from real politicians, created fake NGOs, planted stories in genuine outlets, and had content echoed in the European Parliament. It was patient, professional, and almost entirely human-run. By 2023, operations of similar scope could be stood up in days using large language models — not years.

What Changes When AI Writes the Propaganda?

Traditional disinformation campaigns required writers, translators, cultural consultants, and editors to produce content that felt locally authentic. AI eliminates most of those bottlenecks. A large language model trained on web text can produce locally idiomatic content in dozens of languages at near-zero marginal cost.

In August 2023, Meta removed a network of Facebook and Instagram accounts it attributed to a Chinese influence operation. The operation used AI-generated profile photos (faces that don't exist), AI-written comments, and AI-translated articles to impersonate American political activists across the political spectrum. Meta's threat intelligence team noted this was the first large-scale influence operation it had publicly linked to AI-generated content infrastructure.

AstroturfingCreating the false appearance of grassroots public support using coordinated fake accounts or fabricated individuals.
SockpuppetA fake online identity used to manipulate discussions, amplify specific viewpoints, or harass targets while disguising the operator's true identity.
Influence OperationA coordinated effort, often state-sponsored, to covertly shape public opinion, political outcomes, or international perceptions through deceptive means.

The Narrative Laundering Pipeline

Modern AI-powered disinformation doesn't simply post lies directly. It uses a laundering pipeline: a fabricated claim originates in a fringe forum or state-adjacent website, gets picked up and "reported on" by a fake local news outlet (often AI-generated), then cited by a mid-tier aggregator, and finally picked up by a legitimate outlet as a "report that is circulating."

The NewsGuard 2023 AI misinformation tracker identified 49 websites in 10 languages that appeared to use AI to produce unreliable health and political content at industrial scale, with no human editorial staff. Some were publishing over 1,200 articles per day — impossible for human teams but routine for language models.

Real Event · OpenAI Influence Operation Disruptions, 2024

In May 2024, OpenAI published a report documenting five influence operations from Russia, China, Iran, and Israel that had used ChatGPT to generate social media content, translate articles, and debug code for websites. OpenAI terminated the accounts. The operations ranged from generating fake comments about the Gaza conflict to creating fictional names and bios for fake social media personas. Critically, OpenAI noted none had achieved "significant audience engagement" — but concluded this reflected early-stage AI use in operations, not a fundamental barrier.

Microtargeting and Personalized Propaganda

Perhaps the most troubling development is not mass disinformation but personalized disinformation. AI systems can analyze a target's social media history to infer their fears, values, and persuadable beliefs — then generate tailored messaging designed specifically to move that individual.

Cambridge Analytica's 2016 operations, which used psychographic profiling based on Facebook data to target US voters, were a crude early version of this. The company built models of personality type from "likes" and served different versions of political ads to different psychological profiles. Modern LLMs can do this with far greater nuance and at individual rather than segment scale — one message, one person, optimized for maximum persuasive impact.

Structural Problem · The Engagement Amplifier

Platform recommendation algorithms are optimized for engagement, and research consistently shows that false information spreads faster and farther than true information on social platforms. A 2018 MIT study of 126,000 Twitter stories found false stories were 70% more likely to be retweeted than true ones, and reached their first 1,500 users six times faster. AI-generated disinformation, designed specifically to maximize emotional response, is well-suited to exploit this dynamic.

Automated Amplification: Bot Networks

Creating false content is only half the challenge. Getting it seen requires distribution. Bot networks — coordinated arrays of automated or semi-automated accounts — artificially amplify content to make it appear popular and therefore credible.

The 2020 US election saw estimated bot activity accounting for between 15–20% of political content on Twitter in the weeks before the vote, according to researchers at Carnegie Mellon University, who analyzed 233 million election-related tweets. Modern bot accounts powered by LLMs can respond to replies, engage in multi-turn conversations, and adapt their tone — making them significantly harder to distinguish from human users than the rigid rule-based bots of prior years.

Lesson 2 Quiz

AI-Powered Disinformation at Scale · 4 questions
1. What did the EU DisinfoLab discover in 2019 about the Indian disinformation network?
Correct. The network — known as EU DisinfoLab's "Indian Chronicles" — ran for 15 years, fabricating quotes from politicians, creating fake NGOs, and planting content in legitimate outlets.
The EU DisinfoLab found over 750 fake media outlets across 116 countries, running since 2005, coordinated from India to support Pakistani isolation at international institutions.
2. What is "narrative laundering" in disinformation operations?
Correct. The pipeline: fringe claim → fake local outlet → mid-tier aggregator → legitimate reporting on "the story circulating." Each step adds apparent credibility.
Narrative laundering moves false content through a pipeline — fringe to fake outlet to aggregator to legitimate coverage — so it gains credibility at each step.
3. According to the 2018 MIT study of Twitter, how did false stories compare to true stories in spread?
Correct. The MIT Media Lab study of 126,000 stories found false information has a systematic structural advantage in social media spread, independent of bot amplification.
The MIT study found false stories were 70% more likely to be retweeted than true ones and reached their first 1,500 users six times faster — a structural platform advantage for misinformation.
4. What did OpenAI's 2024 report conclude about AI-assisted influence operations?
Correct. OpenAI terminated the accounts and noted the operations had not yet achieved significant engagement — but assessed this reflected early adoption, not an inherent AI limitation for this use.
OpenAI's 2024 report found five state-linked operations (Russia, China, Iran, Israel) using ChatGPT, but none had achieved significant audience engagement — suggesting early-stage AI use in influence operations.

Lab 2 · Influence Operation Analysis

Dissect the architecture of AI-powered disinformation campaigns

Your Mission

You are a threat intelligence analyst. Use this assistant to map how modern AI-powered influence operations are structured — from content generation to distribution to laundering through credible outlets.

Ask about specific documented operations, the role of LLMs in astroturfing, how platforms detect coordinated inauthentic behavior, or how microtargeting works at scale.

Try: "Walk me through how the narrative laundering pipeline works step by step" or "How did Meta identify the Chinese AI influence operation in 2023?"
Influence Operations Analyst
AI & Misinformation M3
Welcome to the Influence Operations Lab. I can help you analyze how AI is being used to scale and sophisticate disinformation campaigns — from bot networks and sockpuppet infrastructure to narrative laundering and personalized propaganda. What aspect would you like to examine?
Lesson 3 · AI & Misinformation

Health Misinformation and AI

From COVID-19 to vaccine hesitancy: how AI accelerates dangerous falsehoods in healthcare contexts
If an AI confidently tells a patient the wrong thing about their medication — and the patient acts on it — who is responsible?

In March 2020, a message claiming that drinking bleach or consuming high doses of colloidal silver could cure COVID-19 spread across WhatsApp chains in India, Nigeria, and the United States simultaneously. The WHO coined the term "infodemic" — an epidemic of misinformation spreading alongside the disease itself. Iranian state television reported, falsely, that drinking methanol could kill the coronavirus; over 700 Iranians died from methanol poisoning in the following weeks as people acted on the claim. This was human-generated misinformation. By 2023, AI tools were capable of generating such content on demand, with medical-sounding language and fake citations.

How AI Generates Convincing Health Misinformation

Large language models are trained on enormous corpora of text, including medical literature. They can produce text that reads like a clinical study — complete with statistical-sounding figures, plausible mechanisms, and invented citations — without any of the content being true. This is sometimes called "hallucination," but in the context of intentional misuse, it is weaponized fabrication.

A 2023 study in the journal JAMA Internal Medicine tested whether ChatGPT responses to health questions contained misinformation. Researchers found that while most responses were accurate, about 25% contained at least one factual error — and the errors were often stated with the same confident, authoritative tone as the correct information. Users could not reliably distinguish accurate from inaccurate responses.

InfodemicWHO term for an overabundance of information — some accurate, some not — that makes it difficult for people to find reliable guidance, especially during a health crisis.
HallucinationAn AI output that is factually incorrect but presented with apparent confidence — the model generates plausible-sounding text that does not correspond to reality.
Medical MisinformationFalse or inaccurate information about health, disease, treatments, or medical science that can lead individuals to make harmful decisions.

Vaccine Misinformation: A Case Study in AI Amplification

Vaccine hesitancy was already a significant public health concern before AI. The WHO listed it as one of the top ten threats to global health in 2019. But AI tools transformed the production and targeting of anti-vaccine content in measurable ways.

A 2023 analysis by the Center for Countering Digital Hate (CCDH) found that AI image generators were readily producing anti-vaccine imagery — including medically realistic depictions of vaccine "damage" — that would have required graphic design expertise to create previously. These images were then shared on Telegram, TikTok, and Facebook as apparent photographic evidence. Some portrayed faces of children with fabricated injuries described as vaccine side effects.

Real Event · AI-Generated Medical Advice, 2023

In 2023, the National Eating Disorders Association (NEDA) in the United States replaced its human helpline staff with an AI chatbot called "Tessa." Within days, users reported that Tessa was providing advice — including calorie restriction tips — that aligned with eating disorder behaviors rather than countering them. NEDA shut the chatbot down after four days. The incident illustrated a critical gap: AI systems trained on general health content lack the clinical judgment, ethical guardrails, and contextual sensitivity required for high-stakes medical interactions.

The Epistemic Problem: Confident Wrongness

What makes AI-generated health misinformation particularly dangerous is its tone of confidence. Humans naturally interpret fluent, detailed, authoritative text as more credible. An AI that generates a fictional study citing "Nguyen et al., 2021, The Lancet" and describes a mechanism of action for a fake treatment exploits exactly this cognitive bias.

The problem is compounded by the search engine displacement: as AI-generated content floods the web, search results increasingly surface AI-written articles. A study by NewsGuard in late 2023 found that Microsoft's AI-powered Bing Chat was providing misinformation about election topics and health issues in a significant minority of queries — citing its own previous (incorrect) outputs as evidence.

Structural Issue · The Authority Simulation Problem

AI models do not know what they don't know. Unlike a doctor who says "I'm not certain, let me refer you," an LLM without explicit uncertainty-flagging will produce a confident answer in domains where it lacks reliable training data. This authority simulation — appearing expert while being unreliable — is one of the core risks of deploying AI in health-adjacent contexts without robust human oversight.

Responses: Labeling, Guardrails, and Media Literacy

Platforms have implemented health misinformation policies with varying effectiveness. During COVID-19, Twitter, Facebook, and YouTube removed millions of posts that contradicted WHO guidance. Critics argued this over-censored legitimate scientific debate; defenders argued it was necessary triage during an emergency.

AI companies have implemented health-specific guardrails — system prompts that instruct models to recommend professional consultation, add uncertainty hedges, and decline to provide specific medical diagnoses. But these guardrails can be circumvented through prompt injection, jailbreaking, or simply using models without them. The fundamental tension remains: AI tools that are useful for health information are inherently capable of generating health misinformation.

Lesson 3 Quiz

Health Misinformation and AI · 4 questions
1. What happened in Iran in March 2020 related to COVID-19 health misinformation?
Correct. Iranian state TV reported the methanol claim; hundreds died from poisoning as people attempted to use it as a treatment. A direct and lethal consequence of health misinformation.
Iranian state television falsely reported methanol could kill the coronavirus. Over 700 people died from methanol poisoning in the weeks that followed as people acted on this claim.
2. What did the NEDA "Tessa" chatbot incident reveal about AI in health contexts?
Correct. Tessa provided calorie restriction tips to eating disorder sufferers — not through hacking or jailbreaking, but because general health training data lacks clinical contextual sensitivity. NEDA shut it down after four days.
Tessa wasn't hacked — it simply lacked clinical judgment. General health training data doesn't equip AI for high-stakes sensitive clinical interactions, and the chatbot provided eating disorder-aligned advice to vulnerable users seeking help.
3. What is the "authority simulation problem" with AI health responses?
Correct. Unlike a human expert who recognizes the limits of their knowledge and refers accordingly, AI without explicit uncertainty-flagging produces confident text in all domains — appearing authoritative even when unreliable.
The authority simulation problem is that AI generates fluent, confident text regardless of how reliable its information is — it doesn't know what it doesn't know, so it doesn't flag uncertainty the way a cautious human expert would.
4. What did the 2023 JAMA Internal Medicine study find about ChatGPT health responses?
Correct. The 25% error rate was concerning not just for its size but for its presentation — users could not reliably distinguish accurate from inaccurate responses because both were delivered with equal confidence.
The JAMA study found roughly 25% of ChatGPT health responses contained factual errors — and crucially, these errors were delivered with the same confident tone as correct information, making them difficult for users to identify.

Lab 3 · Health Misinformation Evaluation

Practice identifying AI-generated health misinformation and evaluating AI health claims critically

Your Mission

You are a public health communicator tasked with helping communities identify AI-generated health misinformation. Use this assistant to develop frameworks for evaluating health claims, understanding how AI hallucinations appear in medical contexts, and designing effective responses.

Ask about specific documented cases, how to evaluate health claims for AI generation, the authority simulation problem, or what good AI health communication looks like.

Try: "What red flags suggest a health article might be AI-generated misinformation?" or "How should I explain AI health risks to a non-technical audience?"
Health Misinformation Advisor
AI & Misinformation M3
Welcome to the Health Misinformation Lab. I can help you develop frameworks for identifying AI-generated health misinformation, understanding how hallucinations appear in medical content, and designing effective public health communication responses. What would you like to explore?
Lesson 4 · AI & Misinformation

Defending Against AI Misinformation

Media literacy, technical countermeasures, policy frameworks, and the limits of each
If the tools for generating misinformation improve faster than the tools for detecting it, what is the role of the individual?

Since 2014, Finland has run a national media literacy curriculum from primary school through adult education — the result of deliberate investment following Russian information operations during and after the annexation of Crimea. By 2019, Finland ranked first in Europe on the Media Literacy Index, a measure of resilience to misinformation. Studies showed Finnish citizens were significantly more likely to identify false claims, seek primary sources, and reserve judgment on viral content. The program works not by teaching people to detect specific fakes, but by instilling epistemic habits — a disposition to question, verify, and tolerate uncertainty.

The Three Layers of Defense

Defending against AI-powered misinformation requires action at three levels: technical (detection tools, content provenance, platform architecture), structural (policy, regulation, platform incentives), and individual (media literacy, critical thinking, information hygiene).

No single layer is sufficient. Technical tools can be gamed. Regulations lag technology. Individual resilience varies enormously. Effective defense requires all three working together — and accepting that even then, some misinformation will succeed.

PrebunkingInoculation-based approach: exposing people to weakened forms of misinformation techniques before they encounter them, making the brain more resistant to manipulation.
Lateral ReadingFact-checking technique where instead of evaluating a source by reading it deeply, you open new tabs to see what other sources say about the original source's credibility.
Content ProvenanceA cryptographic record of a media file's origin, creation tool, and modification history, enabling verification of authenticity across platforms.

Technical Countermeasures: Detection and Provenance

Detection tools — classifiers trained to identify AI-generated text or synthetic video — are a front line of defense. OpenAI's AI Text Classifier, launched in early 2023, was designed to detect ChatGPT-generated text. OpenAI shut it down in July 2023 due to "low accuracy." More recent classifiers achieve 70–85% accuracy in lab conditions — but accuracy drops sharply on short texts, multilingual content, and text that has been lightly edited by a human.

Content provenance — the C2PA standard described in Lesson 1 — offers a more robust approach than after-the-fact detection. If cameras, phones, and publishing tools embed cryptographic provenance at creation, verification becomes a check of the metadata chain rather than a guessing game. Adobe, Canon, Nikon, Sony, and major news agencies have begun implementing C2PA in professional equipment. Consumer device adoption remains nascent.

Real Initiative · Google's Prebunking Campaign, 2022

In 2022, Google and the University of Cambridge's Social Decision-Making Lab ran a large-scale prebunking experiment in Eastern Europe (Poland, Czech Republic, Slovakia) ahead of Russian disinformation campaigns related to the Ukraine war. Short YouTube videos taught viewers to recognize specific manipulation techniques — emotional language, fake experts, conspiracy thinking. A randomized controlled trial found the campaign significantly increased viewers' ability to identify manipulation in novel misinformation they hadn't seen before. The inoculation effect persisted weeks after viewing.

Policy and Regulatory Approaches

Governments have moved to regulate AI-generated content and platform handling of misinformation, with mixed results. The European Union's AI Act (finalized 2024) requires that AI systems capable of generating synthetic content must label outputs as AI-generated. The EU's Digital Services Act requires large platforms to assess and mitigate "systemic risks" including manipulation of elections and dissemination of illegal content.

In the United States, regulatory progress has been slower. Executive Order 14110 (October 2023) directed agencies to develop standards for AI-generated content labeling, but legislative action remained stalled as of 2024. Several states — including California and Texas — passed laws requiring disclosure of AI-generated content in political advertising, though enforcement mechanisms remain limited.

Framework · SIFT (Stop, Investigate, Find, Trace)

Developed by media literacy educator Mike Caulfield, SIFT is a practical decision framework: Stop before sharing or believing — pause and notice your emotional reaction. Investigate the source — who is behind this? Find better coverage — what do authoritative sources say? Trace claims to their origin — where did this actually start? SIFT was adopted by the State Department's Global Engagement Center and integrated into curricula across more than 20 countries.

The Limits of Defense and the Role of Platform Architecture

Every defensive measure has limits. Prebunking works but doesn't reach everyone. Classifiers work but are beaten by adversarial editing. Provenance standards work but require universal adoption. Regulation works but lags technology cycles.

The deepest structural problem is platform incentive misalignment: engagement-optimized algorithms reward content that triggers strong emotional responses, which is exactly what effective misinformation is designed to do. Until platform business models are altered — through regulation, liability reform, or user pressure — the architecture will continue to advantage misinformation over accurate but less emotionally compelling information.

Research by Renée DiResta at the Stanford Internet Observatory has documented that algorithmic amplification is a greater driver of misinformation reach than organic sharing. Content doesn't spread because users seek it out; it spreads because recommendation systems surface it. Fixing the pipeline — not just the content — is the challenge that remains largely unaddressed.

Lesson 4 Quiz

Defending Against AI Misinformation · 4 questions
1. What makes Finland's media literacy approach distinctive compared to teaching people to spot specific fakes?
Correct. Teaching recognition of specific fakes becomes obsolete as fakes evolve. Epistemic habits — question, verify, tolerate uncertainty — are durable skills applicable to novel misinformation formats.
Finland's strength is instilling epistemic habits rather than specific fake-spotting skills. Habits like questioning, seeking primary sources, and tolerating uncertainty work against misinformation formats that don't yet exist.
2. What is "lateral reading" as a fact-checking technique?
Correct. Professional fact-checkers leave suspicious sites quickly and check what others say about them — rather than spending time evaluating the site's internal content, which can be convincingly constructed.
Lateral reading means leaving a site and opening new tabs to see what external authoritative sources say about that site's credibility. It's the key technique professional fact-checkers use — don't evaluate content from within; check the source from outside.
3. What did Google and Cambridge's prebunking experiment in Eastern Europe find?
Correct. The inoculation effect transferred to novel misinformation (not just examples shown in the videos) and persisted weeks after viewing — making prebunking a scalable, durable intervention.
The Google/Cambridge study found prebunking transferred to novel misinformation not shown in the videos, and the effect persisted weeks after — both crucial findings for scalable public resilience programs.
4. According to Stanford Internet Observatory research, what is the primary driver of misinformation reach on social platforms?
Correct. DiResta's research shows recommendation algorithms are a greater driver than organic sharing or bot networks — meaning platform architecture, not just content moderation, must be addressed.
Renée DiResta's Stanford Internet Observatory research found algorithmic amplification — recommendation systems surfacing content — is a greater driver of misinformation reach than organic user sharing. The pipeline, not just the content, is the problem.

Lab 4 · Misinformation Defense Strategy

Build practical frameworks for defending individuals and communities against AI-powered misinformation

Your Mission

You are advising a media literacy organization developing a curriculum for high school students in a country recently targeted by influence operations. Use this assistant to design and pressure-test a defense strategy combining SIFT, prebunking, lateral reading, and technical awareness.

Ask about specific techniques, how to adapt them for different audiences, what policy levers exist, or how to prioritize when resources are limited.

Try: "Design a 30-minute lesson on SIFT for 16-year-olds" or "What are the limits of teaching lateral reading and how do we address them?"
Media Literacy Strategy Advisor
AI & Misinformation M3
Welcome to the Defense Strategy Lab. I can help you design media literacy interventions, evaluate the strengths and limits of different defense approaches (prebunking, SIFT, lateral reading, technical tools, policy), and think through how to reach specific audiences effectively. What challenge would you like to tackle?

Module 3 Test

AI & Misinformation · 15 questions · Pass at 80%
1. GANs (Generative Adversarial Networks) work by:
Correct. The generator creates synthetic content; the discriminator tries to detect it. They improve each other iteratively until outputs become very difficult to distinguish from reality.
GANs use two networks in competition: a generator creates synthetic content, a discriminator tries to detect it. Their adversarial training iteratively improves output realism.
2. Intel's FakeCatcher detection tool used what biological signal to identify deepfakes?
Correct. Photoplethysmography detects the rhythmic color changes in skin as blood pulses — a real biological signal that current generative video models cannot accurately reproduce.
FakeCatcher used photoplethysmography — detecting the subtle skin color changes caused by blood flow. This biological signal is present in real video but not in AI-generated synthetic video.
3. The 2019 Gabon deepfake incident demonstrated which specific risk?
Correct. The Gabon video may not have been a deepfake at all — but the belief that it might be triggered a coup attempt. Doubt in authenticity itself became the weapon.
In Gabon, whether the video was actually a deepfake was disputed — but the accusation of fakery was enough to trigger a coup attempt. This is the liar's dividend: undermining trust in real content is as dangerous as creating fake content.
4. What was notable about the 49 news websites identified by NewsGuard's 2023 AI misinformation tracker?
Correct. 1,200 articles per day is impossible for any human team — but routine for language models. This industrialization of content production is a qualitative shift in the misinformation threat.
The 49 sites published 1,200+ articles daily in 10 languages with apparently no human staff — a volume only possible with AI generation, representing a genuinely new scale of misinformation production.
5. Cambridge Analytica's 2016 operation pioneered what technique that AI has since made more powerful?
Correct. Cambridge Analytica used Facebook data to build personality models and serve different ad versions to different psychological profiles. Modern LLMs can do this at the individual rather than segment level.
Cambridge Analytica pioneered psychographic profiling — using Facebook likes to infer personality and serve tailored political ads. AI makes this vastly more nuanced and capable of individual-level personalization.
6. The WHO's term "infodemic" refers to:
Correct. The infodemic concept recognizes that information overload itself — not just false information — is a public health problem, because it prevents people from finding and acting on reliable guidance.
Infodemic = information epidemic: too much information of all kinds during a crisis, making reliable guidance hard to find. The problem is overabundance, not just falsity.
7. OpenAI's AI Text Classifier, launched to detect ChatGPT-generated text, was shut down because:
Correct. OpenAI acknowledged the classifier had low accuracy and shut it down in July 2023 — an honest admission that detection is genuinely hard, especially for short texts or lightly edited AI output.
OpenAI shut down its AI Text Classifier in July 2023 due to low accuracy. Even the model's own creator could not build a reliable detector — illustrating the fundamental difficulty of AI text detection.
8. What does the SIFT framework's "T" (Trace) step involve?
Correct. Tracing to origin often reveals that quotes are out of context, statistics are misrepresented, or images are from unrelated events — because misinformation frequently recontextualizes real material.
SIFT's Trace step means finding the original source of a claim, quote, or image. Much misinformation uses real content in false context — tracing to origin exposes the manipulation.
9. The EU AI Act's requirement regarding AI-generated synthetic content is:
Correct. The EU AI Act (finalized 2024) mandates disclosure labeling for AI-generated synthetic content — a transparency requirement rather than a prohibition.
The EU AI Act requires AI systems that can generate synthetic content to label their outputs as AI-generated — a mandatory disclosure approach rather than a prohibition on creation.
10. Sensity AI's tracking data on deepfake videos found approximately what growth pattern?
Correct. Doubling every six months — a roughly exponential growth trajectory — reflects both improving generation tools and their increasing accessibility to non-technical users.
Sensity AI found deepfake volumes doubling roughly every six months — exponential growth driven by improving generation tools and falling technical barriers to use.
11. The NEDA Tessa chatbot controversy is best understood as evidence of:
Correct. Tessa wasn't hacked and wasn't necessarily poorly built for general purposes — it lacked the specific clinical sensitivity eating disorder support requires. The domain gap is the key lesson.
Tessa illustrates the domain gap: general health training ≠ clinical judgment for sensitive high-stakes interactions. An AI competent in general health information can still cause serious harm in specialized clinical contexts.
12. Meta's August 2023 removal of a Chinese influence operation was notable because:
Correct. Meta identified AI-generated profile photos (faces that don't exist), AI-written comments, and AI-translated articles — marking a documented first for AI-infrastructure-supported influence operations at scale.
Meta's August 2023 report was the first time the company publicly linked a large-scale influence operation to AI-generated content infrastructure — AI profile photos, AI comments, AI translations — a documented threshold crossing.
13. Carnegie Mellon research on the 2020 US election found what about bot activity in political Twitter content?
Correct. Analyzing 233 million election-related tweets, Carnegie Mellon researchers estimated bots accounted for 15-20% of political content in the pre-election weeks — a significant and measurable distortion of organic discourse.
Carnegie Mellon's analysis of 233 million election tweets found bots accounted for an estimated 15-20% of political content in the weeks before the 2020 election — a substantial and measurable synthetic component of political discourse.
14. Why does platform algorithmic amplification represent a deeper structural problem for misinformation than content moderation alone can address?
Correct. DiResta's research shows the architecture itself advantages misinformation. Removing content after the fact treats symptoms; the incentive structure of engagement optimization is the disease.
The structural problem: engagement algorithms reward emotionally triggering content, and misinformation is designed to trigger emotion. Content moderation removes individual pieces; the algorithm's incentive structure continuously surfaces more. The pipeline is the problem.
15. The C2PA content provenance standard differs from AI detection tools because it:
Correct. C2PA is a provenance-at-creation approach, not a detection-after-the-fact approach. If creation tools embed cryptographic certificates, verification becomes checking the chain rather than guessing from artifacts.
C2PA embeds provenance at creation — a cryptographic certificate of origin — rather than trying to detect manipulation from artifacts after the fact. This is a fundamentally different and more robust approach than post-hoc detection.