The word is contested, and reasonably so. Critics point out that elites have manipulated publics for as long as there have been publics, and that “post-truth” can flatter the present with a false sense of novelty. Keep the objection in view: what follows argues that the infrastructure is new, not the impulse.

The second row is the load-bearing one. Lying is old; what changes a system is whether being caught carries a price. When a falsehood signals loyalty, exposure can raise a speaker’s standing with their own side rather than lower it.

Read the right-hand column as the argument of this whole deck. Every barrier that once slowed a falsehood was a cost — of printing, of licensing, of labour. Each has fallen to roughly nothing, while the cost of verifying a claim has stayed where it was: a person, doing work, taking time.

None of the four requires a lie, which is what makes the diagnosis useful. A public sphere can arrive at truth decay through ordinary commercial and technological pressure, with every individual actor behaving legally and most of them honestly.

The fourth row is the one outsiders miss. Interventions designed for open platforms assume a visible post that can be labelled. In a family group there is nothing to label and nobody outside to label it.

Each of these is a comfortable answer, and comfort is a reason for suspicion. The deck argues instead for resilience — a slower, less satisfying goal that does not require anyone to win an argument or control a platform.

If you have twenty minutes rather than two hours: Section 2 for the vocabulary, Section 5 for why you are not exempt, and Section 11 for what to do. Those three change behaviour; the rest explains why the behaviour is needed.

The two questions that do the sorting are simply: is it false, and was harm intended? Everything in the next four slides follows from those. They are worth memorising, because they can be applied in the moment to a message you have just received.

Sincerity is the propellant. A forward from an anxious relative arrives wrapped in evident care, and care is far more persuasive than argument. This is why the same false message travels further in a family group than on an open platform.

The last row is the most useful and the least used. “Who benefits?” is not proof of anything and is an excellent first question, because engineered falsehood almost always has a beneficiary while an honest mistake usually does not.

Malinformation defeats fact-checking by design. Every individual claim survives verification, so a checker working claim by claim will rate it true. The deception lives in selection and framing, which no per-claim check examines.

Context collapse is the general mechanism, and satire is only its clearest case. Anything written for one audience — an internal memo, a joke among colleagues, a simplification for a class — can be lifted into an audience it was never written for, where its cues mean nothing.

The commonest operational error is treating row one as row two — responding to a worried relative as though they were running an influence operation. It loses the relationship, which was the only channel you had.

Use this as the deck’s worked example. A neat, quotable line attached to a famous name outran a duller and more complicated history, and it did so among exactly the people who spend their days worrying about misattribution. Nobody was lying. That is the point.

The last row is the finding people skip. It is more comfortable to locate the problem in automated accounts than in ordinary users forwarding things, and the study points the other way. Any intervention aimed only at bots is aimed at the smaller half.

Almost all institutional response happens at stage four, because that is when a falsehood becomes visible enough to be noticed by the people authorised to respond. By then the response is a press release into a hurricane.

The third row is why closed-network falsehood is harder, not easier, to counter. The message arrives pre-endorsed by someone the recipient trusts about everything else. No platform label can compete with that, which is why the interventions that work here are social rather than technical.

Bots work on you through other humans. The mechanism is social proof: seeing a view apparently held by many raises the odds that a real person adopts and repeats it. The bot’s target is not the reader but the reader’s estimate of what everyone thinks.

Note what the first row concedes. The classic tell — copy-pasted wording — is exactly what generative text removes. A thousand distinct, fluent, on-message posts now cost about what one used to, so detection is shifting from content to behaviour and infrastructure.

The last row is Pariser’s actual worry, and it is subtler than “the algorithm shows me bad things”. It is that the editing is invisible: a feed presents itself as a window rather than as a selection, and no version of it announces what it dropped.

They compound because each feeds the other. You follow the like-minded, so the model learns you engage with agreement, so it serves more of it, so your sense of what is normal narrows, so the next person you follow is more like the last.

The point of this slide is to stop the argument at the algorithm. Networks here were narrow before any platform existed, and residential and marital segregation shape whose forward reaches whom more strongly than any ranking model does.

Selective exposure is a demand-side problem, which is why supply-side fixes disappoint. Give a person a balanced feed and they will still click what agrees with them; the platform then learns from the clicks and rebuilds the imbalance within a week.

A deck that presented only the strong bubble story would be doing what it warns against. The evidence is genuinely mixed, and saying so is not hedging — it is the difference between teaching a subject and campaigning about it.

The fourth row is the bridge to Section 9. The damage that matters most is not informational but relational: not that you hear one view, but that you stop encountering the people who hold others as people you could be wrong about.

“Serious people you disagree with” is the load-bearing phrase. Following the other side’s worst performers is worse than following none: it supplies a steady stream of evidence that they are ridiculous, and it feels like open-mindedness.

None of these is a defect. Each was useful where information was scarce, the sources were people you knew, and the cost of being slow was high. They misfire in an environment engineered by people who studied them.

The third row is where it bites hardest for professionals. You are not failing to apply standards; you are applying them asymmetrically, with full rigour to what you did not want to hear. It feels exactly like rigour, which is why it survives training.

Which is why “more information” is a weak programme design. Where an issue has become identity-linked, adding facts to the environment can increase polarisation. That is not an argument against facts; it is an argument for not relying on them alone.

The effect survives knowing better. Studies find repetition raises perceived truth even for claims a person can correctly identify as false when asked directly. Fluency — the ease with which something comes to mind — is doing the work, not belief.

This is the deck’s best self-referential example. A striking research finding travelled further and faster than its own correction, through a community of people whose profession is correcting things. Everything in Section 3 applies to academic and NGO communication too.

A denial leaves a hole where an explanation was. The mind keeps using the discredited account because nothing has replaced it, so the correction is accepted and the reasoning it supported carries on unchanged.

The durable observation underneath it needs no contested effect: confident, unhedged claims travel better than careful ones, and expertise usually arrives hedged. That asymmetry is real and is a communication problem, not a competence problem.

These are habits, not knowledge, which is why a session on cognitive bias so often changes nothing. Knowing the list does not help; having a rule that fires at the right moment does, and rules only fire if they are few and specific.

The candour is the striking part. Bernays used the word “propaganda” as a neutral professional term and defended manipulation as a stabilising social function. The techniques did not go away when the word acquired its later associations; only the label did.

The right-hand column is harder to argue with, because a counter-argument now threatens who the listener is rather than what they think. This is the same mechanism Section 5 described from the inside; here it is being used deliberately from the outside.

The model is contested and worth teaching anyway, because its central question survives the disputes about its answers: what shapes the news before anyone tells a lie? Ownership, advertising and cheap sourcing are observable facts about media economics regardless of what you conclude from them.

Every row is defensible and none is neutral. There is no unframed way to describe a contested thing, which is why “just report the facts” is not available as an instruction — the choice of words is already an argument.

The fourth row is the one development practitioners feel most. Chronic malnutrition kills more people than any single event that will lead a bulletin, and it is never new. Agenda-setting is why slow, large harms lose to fast, small ones for attention and funding.

The last row is why the usual defences underperform. Fact-checking a firehose is answering each of forty claims while the forty-first is posted, and each answer confirms the impression that everything is contested and nothing is knowable.

The aim has not changed in a century. What has changed is the last column: influence at scale has moved from a capability of institutions to a capability of individuals, and the institutions built to check it were designed for the first two rows.

Herbert Simon made the underlying point in 1971: a wealth of information creates a poverty of attention. Everything in this section follows from treating attention, not information, as the thing being fought over.

The right column is not a list of oversights. Those quantities are genuinely hard to measure and lie outside the boundary of the business, which is the textbook definition of an externality. Truth is a cost borne by the public, not by the firm.

None of these is an accident, and none was designed to spread falsehood. They were designed to hold attention, and holding attention in a place where falsehood outcompetes truth is sufficient for the harm.

This is the most important slide in the section and the one that persuades least, because it removes the villain. A system can produce a result nobody chose, that everybody involved would disown, and that persists regardless of who is in charge.

No falsehood is required at any step. A careful person who wants an audience faces a gradient, and following it produces overconfident, under-caveated content by people who would fail no fact-check. The market pays a premium for certainty, including the false kind.

Note what is not on the list: quitting. Absolutist advice is easy to give, rarely taken, and unnecessary. The aim is to restore the decision points that the design removed.

The loss is not of any single photograph but of a default. For a century, “there is footage” ended an argument. What replaces it is provenance — who recorded this, where did it come from, who vouches for the chain — which is slower, harder, and unavailable to most people looking at most images.

The Veles case is worth remembering for its motive. The teenagers running those sites had no politics in the matter; they had discovered which content earned advertising revenue. Falsehood as a straightforward business, with no ideological actor anywhere in it.

This is the most common visual falsehood in the region and the easiest to defeat. A real image from elsewhere or from years ago, recaptioned as local and current, needs no technology to make and one search to expose.

The economics have inverted. Convincing synthetic voice now needs seconds of source audio, which anyone who has posted a video has already supplied. Assume the capability exists rather than waiting for it.

This is the more serious half of the deepfake problem, and it needs no deepfake to operate. Merely knowing that convincing fakes exist is enough to make genuine evidence deniable, which is why the harm arrives before the technology is widely used.

For South Asia the language row is the significant one. The main practical protection for smaller-language information ecosystems was that producing fluent falsehood in them took a fluent person. That protection has gone, and detection capacity in those languages has not arrived to replace it.

The last row is the surprising finding. Publics can move closer on actual policy while liking each other markedly less — which means affective polarisation is not simply a symptom of deeper disagreement and cannot be fixed by settling the disagreements.

The third marker is why this section sits inside a deck about falsehood. Affective polarisation is the mechanism that makes disinformation land: a claim about a disliked group needs almost no evidence, because the audience has already decided the group is capable of it.

The first row is the uncomfortable one. In the classic minimal-group experiments the dividing line was meaningless and disclosed as meaningless, and favouritism appeared anyway. Real lines — caste, religion, language — are doing this with far more material behind them.

The first row is the cheapest and least known. People routinely overestimate how extreme the other side’s views are — a gap sometimes measured at several times the real distance. Simply showing the real distribution moves hostility, and it needs no one to change their own position.

This is the slide that gets the budget. Framed as a media-studies concern, post-truth loses to every operational priority. Framed as a risk to uptake, to staff safety and to the community trust a programme runs on, it is on the risk register where it belongs.

The second row is the one that separates competent from incompetent response. Vaccine fertility rumours in South Asia have historically drawn force from real memories of coercive family-planning programmes. A campaign that calls the belief stupid is arguing with a memory, and loses.

What has demonstrably helped: forward limits that slow velocity, police who publish a verified account within hours rather than days, and local volunteers who can say “this photograph is from another country” inside the group where it is circulating.

Recognising the shape is the first defence, because the individual claims are designed to be argued with one at a time. Seeing four steps assemble is quicker than fact-checking each, and it tells you what kind of thing you are looking at.

Timing is the design, not an accident. Each of these is placed where the correction cycle cannot complete. That is why election-period responses have to be pre-built — verified channels, a named spokesperson, prepared material — rather than assembled when the claim appears.

Not one of these requires stating a falsehood, which is what makes the playbook durable. It was documented in the tobacco industry’s own papers — “doubt is our product” — and has since been reused, sometimes by the same consultancies, on acid rain, ozone, lead and climate.

This is why the firehose strategy targets belief itself rather than any particular belief. A population that trusts nothing cannot be mobilised for anything — not a vaccination drive, not a census, not a climate transition — and that immobility serves whoever prefers the current arrangement.

The first three are supply-side and the fourth is demand-side. Only the fourth scales, because it does not need to be repeated for each new falsehood, and it is the only one available to a person or an organisation without platform access.

South Asia has a strong and growing fact-checking ecosystem, and it is a public good worth defending — particularly where checkers face legal and physical pressure. The point of this slide is not to diminish it but to explain why it cannot be the whole strategy.

The research behind this compared historians, undergraduates and fact-checkers on the same task. The fact-checkers reached better judgements in less time, and the difference in method was simply that they opened new tabs instead of scrutinising the page in front of them.

Inoculation theory goes back to William McGuire in the 1960s and has been revived and tested at scale in the digital era, including through short videos shown before content. The results are consistent and modest — real effects that decay over weeks and need boosters, exactly like the vaccine metaphor implies.

Most institutional corrections break rule three. The myth appears in the headline, the graphic and the first line, because it is the interesting part. What the reader retains, a week later, is the vivid claim rather than the correction attached to it.

Media literacy is the long game because it compounds. A cohort taught these five questions at fourteen carries them into forty years of information they will encounter, most of which does not exist yet — which no debunking programme can claim.

Read the 2018 Science paper itself if you read only one. It is the source of the most-cited claim in this field, and knowing what it actually measured — one platform, one period, fact-checker-rated stories — makes you the person who can use it accurately when everyone else is rounding it up.

And the systemic point stands alongside it: individual habits do not fix an incentive structure. Both halves are true, and treating either as sufficient is the error this deck has tried hardest to avoid.