If nobody stands outside causality, changing a mind cannot mean appealing to some little sovereign who floats above biology and history. It means adding something to the history.
Table of contents
On this page
Facts Never Travel Alone
Facts matter because an accurate model of the world lets us predict and act. A weather warning can change a journey. A payslip can reveal underpayment. A medical result can open a treatment path. Truth has practical consequences.
But facts never arrive as bare facts. Somebody selects them, puts them in order, gives them a title, and delivers them through a relationship or institution. Even a table has a frame.
Stories are part of that frame. They connect events, supply motives, and let us imagine consequences. A meta-analysis of narrative persuasion found effects across beliefs, attitudes, intentions, and behaviour. That does not make stories magical mind-control devices. It means that form can help determine which facts are noticed, remembered, and made meaningful.
Education works more slowly. It supplies vocabulary, background knowledge, numeracy, methods, and experience with disagreement. It changes the set of explanations a person can consider. A meta-analysis of media-literacy interventions found positive average effects across knowledge, criticism, attitudes, and behaviour, with variation between interventions.
Education is not an immunity badge. Clever people can become very good at defending conclusions they reached for other reasons. Still, education expands the machinery available for checking a claim. That makes it another causal input, not an escape from causality.
Familiar Is Not the Same as True
Repetition matters too.
A repeated statement becomes easier to process. That ease, or fluency, can be mistaken for evidence that the statement is true. Psychologists call this the illusory truth effect. It has appeared across controlled studies and was summarised in a meta-analysis of 51 studies.
Prior knowledge helps, but it is not a perfect shield. In one set of experiments, repetition increased truth ratings even for claims that contradicted knowledge participants already had.
That result deserves care. Much of this research uses short statements, artificial repetition, and immediate ratings in controlled settings. It does not show that every repeated lie converts everyone, or that we can assign one universal effect size to a political campaign. It does show a recurring vulnerability worth taking seriously.
Repetition is also content-neutral. It can reinforce a safety instruction, a mathematical method, or a true correction. The problem is not repetition itself. The problem is allowing money, power, or an engagement system to decide which claims receive it.
People are not simple databases where a false row can be overwritten with a true one. Beliefs sit inside memories, explanations, habits, loyalties, and expectations. Replacing one part may require rebuilding the connections around it.
Trust, Identity, and Emotion
Most knowledge is social. I cannot personally rerun every clinical trial, inspect every bridge, or verify every historical archive. As the Stanford Encyclopedia of Philosophy’s entry on testimony explains, relying on what others tell us is indispensable, even though philosophers disagree about when that reliance is justified.
Trust is therefore not a regrettable substitute for thinking. It is part of how thinking works in a society with specialised knowledge. The useful questions are whether a source has relevant expertise, shows its evidence, admits uncertainty, corrects errors, and has earned trust in this domain.
Mistrust has causes too. A community deceived or mistreated by an institution may reasonably give its next statement less weight. Repeating “trust the experts” does not repair that history.
Identity also changes how information lands. A claim may imply that my family, political group, religion, profession, or entire picture of myself is foolish. Accepting it can carry a social cost. An identity-based review of political belief describes ways partisanship can shape attention, memory, and evaluation.
This does not mean identity always defeats evidence. It means evidence arrives alongside cues about belonging and threat.
Emotion is part of the package as well. Fear directs attention towards danger. Anger prepares conflict. Hope makes action seem possible. In one experimental programme, being prompted to rely on emotion increased belief in false news headlines. That finding is not a reason to treat emotion as failed reasoning. Emotion carries information about what matters. It can also be deliberately exploited.
Influence at Scale
Misinformation is false or misleading information, whether or not the person sharing it intends to deceive. Disinformation usually means the deception is deliberate. Propaganda is broader. It can mix truth, falsehood, selective evidence, symbols, stories, omission, and repetition in service of a political project.
Propaganda becomes especially powerful when joined to control over institutions. The US Holocaust Memorial Museum’s history of Nazi propaganda describes a coordinated system spanning radio, film, schooling, art, books, and the press. It worked with censorship, persecution, and state power. This was not merely a collection of persuasive posters competing in a fair marketplace of ideas.
Today’s information systems organise attention at extraordinary scale. Recommendation algorithms choose what appears, in which order, how often, and beside which signs of social approval. They do not need beliefs of their own to have effects. Their priorities are set through objectives such as predicted engagement, watch time, or retention.
Those systems are not omnipotent. A large Facebook and Instagram experiment found that replacing algorithmic feeds with chronological ones substantially changed use and exposure but did not detect corresponding changes in several political attitudes during the three-month study. Algorithms alter the information environment. That does not mean every alteration produces a measurable conversion.
Truth has no automatic distribution advantage. In one large historical Twitter dataset, false stories travelled farther and faster than true ones. That was one platform, period, and classification method, not a law of nature. It is still enough to kill the comforting idea that truth inevitably rises because it is true.
Correction Is a Cause Too
Fortunately, misinformation is not irreversible.
A meta-analysis of political fact-checking found that corrections improved factual accuracy on average, although effects varied with format, topic, prior beliefs, ideology, and knowledge. A broad experimental search also found that the feared backfire effect was elusive. The claim that correcting people routinely makes them believe the falsehood more strongly is too sweeping.
Correction is not perfect deletion. A major review of misinformation and correction describes continued influence, where corrected information can still affect later reasoning. Corrections may fade, come from a distrusted source, or remove a story without supplying a better explanation. Much of the evidence also measures short-term judgements rather than lasting behaviour.
The sensible conclusion is neither “corrections do not work” nor “just show people the facts”. Corrections are causal inputs with average benefits and real limits.
Imagine a parent receives a family-group voice note claiming that the MMR vaccine causes autism. It comes from a cousin who has been reliable about childcare, includes a vivid personal story, and appears again after a worried search. The claim is false. A nationwide Danish cohort of 657,461 children found no increased autism risk after MMR vaccination, including in the higher-risk groups it examined.
Sending that paper may help. A better response may also explain what the researchers compared, acknowledge the parent’s concern, and offer a conversation with a trusted clinician. Humiliation would attach the correction to a threat against identity. Calm explanation gives it a better route in.
Even then, information may not settle what happens. The parent may lack a nearby clinic, paid time off, transport, childcare, or support in their language. Past medical mistreatment may make reassurance harder to trust. A correct belief does not manufacture an appointment.
Knowing is not the same as being able. The capability approach makes this distinction explicit: a resource becomes a real opportunity only through personal, social, and environmental conditions.
Capacity, incentives, trauma, power, and material conditions all constrain action. A worker can understand that a job is damaging them and still need the wage. A traumatised person may experience a message about danger differently from someone who feels safe. A citizen can know they are being exploited while the exploiter controls housing, employment, media, or law.
Information matters enormously. It is not sufficient.
Persuasion Without Puppetry
If all communication enters a causal chain, the ethical distinction cannot be between influencing somebody and leaving them untouched. I am trying to influence you by writing this.
The better distinction is in how influence works. Persuasion offers reasons openly. It identifies its source and purpose, represents uncertainty honestly, permits checking, answers objections, and leaves room to refuse. By respecting agency here, I mean engaging a person’s capacity to consider reasons through their own values, not appealing to an uncaused soul.
Manipulation tends to conceal, deceive, pressure, or exploit a vulnerability in ways that make sound deliberation harder. The exact boundary is disputed, as the Stanford Encyclopedia’s discussion of manipulation makes clear. Stories and emotional appeals are not automatically manipulative. A story can illuminate a reason. A frightening image can accurately show danger.
Power changes the ethics. A friend making an argument and a platform invisibly optimising millions of feeds for engagement are not equivalent. Useful questions are who knows the influence is happening, whose interests it serves, what alternatives remain, and whether mistakes can be challenged.
Tending the Information Environment
Nobody can fact-check every sentence. We can still improve the conditions under which information reaches us.
-
Add friction before sharing. Pause when a message produces an immediate urge to warn, punish, or perform loyalty. Ask what the exact claim is and whether the linked source supports it.
-
Read sideways. Professional fact-checkers often leave a page to investigate its publisher and claims elsewhere, a practice called lateral reading. A polished website should not get to supply all the evidence for its own credibility.
-
Choose better defaults. Use direct subscriptions, chronological feeds, or RSS where practical. Turn off needless notifications. These choices do not guarantee truth, but they reduce the power of moment-by-moment engagement ranking.
-
Reward correction. Prefer sources that publish methods and visible corrections. When you get something wrong, correct it clearly. This makes updating a normal social act rather than a humiliating defeat.
-
Correct with an alternative. Lead with the accurate account, explain why the false claim fails, and provide a better explanation. The research-based Debunking Handbook is a useful practical guide.
-
Protect capacity. Exhaustion and panic are poor conditions for evaluating claims. Sometimes improving an information environment means translation, time, safety, healthcare, education, or a relationship in which questions are allowed.
These are not techniques for becoming invulnerable. Nobody gets administrator access to their own brain. They are modest ways of changing the inputs and making truth easier to reach.
Not Quite a Battle
The battle metaphor is partly justified. Governments, companies, campaigners, and propagandists compete for attention. Some deceive deliberately. Budgets, censorship, surveillance, and ownership give certain speakers far more force than others.
But battle language can also make every disagreement look like enemy action. It turns people into targets or territory, makes listening resemble surrender, and tempts each side to excuse its own manipulation.
Most days, I prefer to think of an information environment: something with inputs, feedback loops, pollution, maintenance, and public goods. The metaphor is less thrilling, which may be one of its virtues.
Truth does not automatically win. It needs access, attention, comprehension, trust, repetition, protection, and institutions willing to correct themselves. Building those conditions is one way a society can deliberately alter the causes shaping its future.
That still leaves a harder question. If nobody ultimately authored the causes that formed them, what should society do when those causes produce harm?
That is where Part 6, Can Justice Survive Without Free Will?, begins.