How to talk to someone who believes something you don't

Series: How minds are moved · Living with it: self-defence for the mind · 3 of 4

In the summer of 2015, 56 canvassers in South Florida knocked on the doors of 501 voters and had a conversation of about 10 minutes each. They did not argue. They asked people to remember a time they had been judged for being different, and to talk about it. Three months later, those voters’ feelings towards transgender people, measured on a 100-point “feeling thermometer”, were still about 10 points warmer than in the control group. The authors described the result as durably reducing transphobia. The shift is larger than the average American’s change in attitudes towards gay and lesbian people over the 14 years from 1998 to 2012. The study, by David Broockman and Joshua Kalla, appeared in Science in 2016.

This post is about the other side of self-defence. Not how to resist being moved, but how to talk with someone whose belief you think is wrong, without manipulating them and without wasting your breath.

Contents · 12 sections

TL;DR #

• Arguments and facts thrown at someone rarely change minds, but the feared “backfire” is also rare. Mostly nothing happens.
• Three approaches have evidence or a track record: deep canvassing (listening and exchanging stories), motivational interviewing (helping people voice their own reasons), and street epistemology (asking how someone knows what they believe).
• They share a core: respect, questions, patience and no mockery.
• The evidence is strongest for deep canvassing and motivational interviewing, and thin for street epistemology.
• The methods have no built-in direction. The studies tested one direction each; the same techniques can move people the other way.
• These are tools for honest conversation, not for stealth conversion. Use them only for what you would say openly.

Why the usual approach fails, and why it is not hopeless #

The usual approach is a fact dump: here is the study, here is the chart, you are wrong. It feels like it should work. It mostly doesn’t, and the reasons are the ones this series has described: people weigh messages by who sends them, and a message from someone who treats them as stupid is discounted.

For a while it was believed that correcting people could make them more certain, the “backfire effect”, after a 2010 study by Brendan Nyhan and Jason Reifler. Later work found that this is rare. In 2019 Thomas Wood and Ethan Porter tested corrections on more than 10,000 people across 52 issues and found that, by and large, corrections moved people towards the facts, even on politically charged topics. Facts do work, a little. What they rarely do on their own is change a belief that is part of someone’s identity.

Deep canvassing: exchange stories, don’t trade blows #

Deep canvassing came from activists in Los Angeles who had lost a ballot campaign and went back to talk to voters who had voted against them. The method is simple to describe and hard to do: ask the person what they think and why, listen without correcting, share a personal story, and ask them for one of theirs, especially a time they needed compassion.

The 2016 Science study is notable for a second reason. It was carried out by the two researchers who, a year earlier, had discovered irregularities in an earlier, famous study of canvassing by Michael LaCour and Donald Green; that study was retracted by Science in 2015. So the 2016 result was examined with particular care, and it has since been extended. In 2020, in the American Political Science Review, Kalla and Broockman reported three more field experiments, on immigration and transgender rights. The key ingredient was what they called a “non-judgmental exchange of narratives”, and the effects lasted at least four months in the experiment where they were followed.

It is not magic. The effects are a few points, from a ten-minute conversation, on one topic. But compare that with the zero average effect of campaign contact described in the previous post in this family. Listening did what broadcasting could not.

Motivational interviewing: let them say it #

Motivational interviewing was developed by the psychologists William Miller and Stephen Rollnick, first for problem drinking, from the early 1980s. Its insight is that people rarely change because someone tells them to. They change when they hear themselves state their own reasons. The counsellor resists the urge to correct, asks open questions, reflects back what the person says, and listens for what the method calls “change talk”.

In my library, Adam Grant’s Think Again tells the story of Arnaud Gagneur, a Quebec neonatologist who used this method with parents hesitant about vaccinating their babies. He did not judge or lecture them; he asked about their concerns and their goals. Grant reports that motivational interviewing has been the subject of more than a thousand controlled trials. Gagneur’s own published work in Quebec found that a session based on the method in a maternity ward raised mothers’ intention to vaccinate by about 15%, and the share of babies vaccinated by seven months rose too. The Quebec government then funded the programme for maternity wards.

The method works because it treats ambivalence as normal. Most people who hold a strong belief also hold doubts. Your job is not to insert doubt. It is to make room for the doubts already there.

Street epistemology: ask how they know #

Street epistemology asks a different question. Not “is your belief true?” but “how did you come to believe it, and how confident are you, on a scale of 0 to 100?” Then: “what would change that number?” The conversation is about the method of knowing, not the content.

It is fair to say where it comes from. The term was popularised by the philosopher Peter Boghossian in A Manual for Creating Atheists (2013), and it began as a tool for questioning religious faith. Since then, its practitioners have applied it to every kind of belief, political and scientific as well as religious, and the best of them publish their conversations on video. As far as I can find, there is little controlled research on its effects. It is a set of good questions with a strong following, not a proven intervention, and I would treat it that way.

Its questions are useful on any side, including your own. Try them on a belief you hold strongly.

A machine can do some of it, which should make us think #

In 2024, Thomas Costello, Gordon Pennycook and David Rand reported in Science that conversations with an AI model reduced belief in a chosen conspiracy theory by about 20% among 2,190 Americans, and the effect lasted at least two months. The AI was patient, specific and never insulting. It answered the person’s actual reasons rather than a generic version of them.

That is a hopeful finding about people: many can revise a belief when they get a respectful, specific answer. It is also a warning, which the authors themselves gave: the same technique could be used to talk people into false beliefs. Good conversation is a power, and power can be pointed anywhere.

The methods have no direction of their own #

Every worked example in this post moves people one way: towards warmer feelings about a minority, towards vaccination, away from a conspiracy theory. That is because those are the outcomes the researchers chose to test, not because the methods only work that way. Deep canvassing is a way of holding a conversation, and campaigns on any side of an issue can use it. Motivational interviewing helps people voice whatever change they are already ambivalent about. Street epistemology’s questions apply to any belief, including the questioner’s. The AI study’s authors warned explicitly that the same approach could talk people into conspiracies. My inference from the mechanism: a respectful, listening conversation is persuasive in whatever direction the person holding it is going. That is why the honesty rule below matters more than the technique.

The same knowledge works as a defence. If a stranger, a campaigner or a salesperson listens unusually warmly, trades a personal story for yours and then asks you to rate your certainty from one to ten, you are probably inside one of these methods. That is not sinister in itself, but it is worth noticing, so you can ask yourself whether you are being heard or being steered.

Respect, questions, no mockery #

The common core of all three methods is not a trick. It is a stance.

• Respect: the other person has reasons, and their reasons make sense from where they stand.
• Questions: you learn their reasons before you answer them.
• Patience: one conversation rarely ends anything. It opens a door.
• No mockery: ridicule turns a belief into a badge of loyalty. Nobody leaves a group that is being laughed at, because leaving would mean joining the people laughing.

And one rule that matters most for this series: use these methods only for what you would say openly. A conversation that hides its purpose is manipulation, whatever the cause.

What it feels like from the inside #

From your side, it feels like losing. You do not get to deliver the knockout argument. You sit with views you think are wrong and you ask about them. You may leave without any sign of change.

From their side, it may be the first time someone who disagrees has listened. That is often what makes the difference. People who feel heard can afford to doubt. People who feel attacked cannot.

How to notice it is working, or not #

• They are explaining their reasons in more detail, not repeating slogans.
• They mention a doubt, a “but”, or an exception without being pushed.
• The temperature drops: shorter sentences, fewer absolutes.
• If you catch yourself scoring points, waiting for your turn to talk or enjoying their discomfort, stop. It has become a debate.
• If you would be embarrassed for them to know your goal, change your goal.

Frequently asked questions #

What is deep canvassing? #

A conversation method in which a canvasser listens to a person’s views without judgement, shares a personal story and invites one in return. Field experiments by Broockman and Kalla found that ten-minute conversations of this kind durably reduced transphobia (2016) and, in their words, “exclusionary attitudes” towards transgender people and immigrants (2020), with effects lasting three to four months. The method itself can be used for any position.

What is motivational interviewing? #

A counselling approach developed by William Miller and Stephen Rollnick that helps people explore their own reasons for change through open questions and reflective listening. It has been tested in more than a thousand controlled trials, according to Adam Grant, and has been used for vaccine hesitancy in Quebec.

What is street epistemology? #

A questioning method, popularised by Peter Boghossian in 2013, that asks people how confident they are in a belief and how they reached it. It began as a way to question religious faith and is now used on all kinds of beliefs. Controlled evidence on its effects is limited.

Does correcting someone make them believe the falsehood more? #

Rarely. The “backfire effect” was reported in 2010, but a 2019 study of more than 10,000 people across 52 issues found that corrections generally moved people towards the facts.

Should I ever mock a belief? #

The evidence and experience point the other way. Mockery makes a belief part of a person’s identity and makes leaving it harder. It also makes you less likely to notice when they have a point.

The lesson #

The methods that work are the ones that treat people as capable of thinking. They are slower than an argument and less satisfying than a joke, and they work better than both. Ask before you answer, listen to the reasons and not just the claim, tell a story of your own, and never use these tools for anything you would not say openly. That is the difference between persuading someone and moving them like an object.