Skip to content

← Thinking Clearly

Disagreeing well

Chapter 34 · The one source you haven't checked

In this chapter 4 sections
  1. Saying it so it sticks
  2. The effect that wasn’t
  3. Keeping the person’s dignity
  4. When to let it go

On 31 August 2024 a video went up on YouTube whose title said that a well-known supermarket chain in Nepal was selling plastic rice, “playing with the health of simple people”. It runs for more than a quarter of an hour. Rice is cooked, pressed into a ball while it’s hot and dropped on the floor, where it bounces. Then it’s held over a gas flame. Then the man who made the video carries the rice back to the shop.

Suppose it turns up in your family’s WhatsApp group, sent by an uncle with a worried line on top about being careful what you buy. (The uncle is our own example. The video is real.) You’re fairly sure it’s nonsense. What do you do?

Most people do nothing. Part of that is manners: he’s older, the whole family is in the group, and nobody wants a quarrel over rice. Part of it is a worry that correcting people only makes them dig in. That second worry turns out to be mostly wrong, and the first has a better answer than silence.

Saying it so it sticks

Start with what’s true, because here it’s simple. Rice that bounces is rice. Rice whose starch is rich in amylose, one part of the starch in a grain, turns firm and springy when it’s cooked and squeezed into a ball while hot, with air caught between the grains. Drop it and it bounces, and the more amylose, the higher. Nepal’s Department of Food Technology and Quality Control has been answering this rumor for years. In 2017 it tested the rice of a trader arrested over it and found no plastic. In 2019, after more than two dozen complaints, it put a notice in the newspaper Gorkhapatra saying its tests had found none. Its spokesperson added that faking rice in plastic wouldn’t even pay. NepalFactCheck.org looked into the 2024 video and called it false, and noted that the same hoax had already gone round Ghana, Senegal, the Gambia and India.

The uncle doesn’t need all of that. What he needs is the right part in the right order, and the order matters more than you’d think.

A false story, it turns out, leaves a hole when you take it away. In a well-known experiment published in 1994, two psychologists, Hollyn Johnson and Colleen Seifert, had people read a string of short news messages about a fire in a warehouse. An early message said that oil paint and gas cylinders had been stored carelessly in a closet. A later one said that, in fact, no such things had been stored there at all. Then people were asked what had caused the explosions and the fumes, and many still answered with the paint and the cylinders, including people who remembered the correction perfectly well. The story needed a cause, and the only one on offer was the one they’d been told to forget. When the messages also said the fire had been set on purpose, people had something to put in the hole, and they leaned on the paint less.

That’s why a bare “this is fake” does so little. A worried uncle who watched rice bounce has an explanation for what he saw. Take it away and leave nothing, and the bounce is still there, needing a reason. So the best corrections, as 22 researchers on misinformation agreed in 2020 in a short guide called The Debunking Handbook, come in four parts. Lead with the fact, if it can be said simply. Mention the myth once, just before the correction, and warn that it’s wrong. Explain how it misleads. Then say the fact again, so that it’s the last thing heard. One of the 22 was Colleen Seifert, of the warehouse fire.

For the rice, that might be four short lines. Rice that bounces is just rice. The video says it must be plastic because it bounces. But any rice rich in amylose bounces when it’s balled up hot, and the government’s tests have never found plastic rice in Nepal. It’s rice. There’s a version of those lines in Correct in four steps.

The third line is where naming the trick comes in. The video’s trick is a real thing with the wrong cause pinned to it. The rice really does bounce. It just doesn’t bounce because it’s plastic. Naming the trick is worth the extra sentence because the trick travels, even though the myth doesn’t. Knowing the truth about rice won’t help with the next scare about eggs or milk, but knowing that a real sight can be given a false cause will.

Doesn’t mentioning the myth at all make it feel truer? Hearing a claim again does nudge it toward feeling true, as the chapter on the feed showed, but what does the harm is repetition on its own, the claim met again and again with nothing to answer it. Said once, and paired at once with why it’s wrong and what’s true instead, the myth makes the correction sharper, and the correction usually outweighs the small nudge. What to avoid is a reply that puts the myth in the headline, or quotes it three times on the way to saying no.

The effect that wasn’t

The fear that correcting does harm has a name, the backfire effect, and a famous beginning.

In the autumn of 2005 two American political scientists, Brendan Nyhan and Jason Reifler, gave students at a university in the Midwest a made-up news article. It quoted President George W. Bush, in a real speech from October 2004, saying there had been “a risk, a real risk, that Saddam Hussein would pass weapons or materials or information to terrorist networks”. For some of the students, chosen at random, the article went on to report the Duelfer Report, which had found that Iraq had no stockpiles of weapons of mass destruction and no active program to make them when the United States invaded in 2003. Then everyone was asked whether Iraq had in fact had such weapons and hidden them.

Among the students who called themselves conservative, the correction went the wrong way. Without it, 32% agreed that Iraq had the weapons. With it, 64% did. Facts, it seemed, could make people more wrong. The paper came out in 2010 as “When Corrections Fail”, and the backfire effect was soon everywhere, in books, articles and talks, often as proof that arguing with anyone is pointless.

Underneath, the numbers were small. That 32% and 64% came from 33 conservative students. When Nyhan and Reifler ran a second version the next spring, with slightly different wording, the correction worked on conservatives: it made them more likely to accept that Iraq had had no such weapons. In an analysis done afterward, the backfire showed up again only among right-of-center students who rated Iraq the country’s most important problem, and there were eight of them. The authors reported all of it. The retelling dropped it.

Then Thomas Wood and Ethan Porter went looking for the effect on a large scale. In five experiments with more than 10,000 Americans, they corrected false or misleading claims made by politicians, on 52 issues chosen because people felt strongly about them. None of their corrections backfired. On most, people of every political leaning moved toward the facts. Other searches came up just as empty, and some of the early backfires may simply have come from measuring belief with a single question, which is a shaky way to measure anything.

What happened next is a small picture of how this is supposed to go. In 2021 Nyhan wrote that later research and the press had seized on his finding, “distorting its generality and exaggerating its role”. Reifler, Wood and Porter were all among the 22 authors of the Debunking Handbook, which tells its readers not to hold back from correcting misinformation “out of fear that doing so will backfire”. The people who had found the effect, and the people who failed to find it, ended up writing the same advice.

The good news comes with a catch. Corrections work, usually somewhat rather than completely, and they fade. The effect can last for weeks, but people tend to drift back toward what they believed before, especially when the people they look up to keep repeating the myth. So a correction is rarely one message. It’s worth doing, and it may be worth doing again.

Keeping the person’s dignity

Nearly all of that was measured on Americans reading articles by strangers. A family group is different. You know the person, they know you, and they’re often older than you.

The closest test to it comes from India. Sumitra Badrinathan and Simon Chauchard showed about 5,100 Hindi speakers, found through Facebook, pictures of what looked like WhatsApp group chats. In each, someone posted a rumor about politics, health or sport, and someone else replied. Sometimes the reply was just “thank you”. Sometimes it disagreed, and sometimes it disagreed and pointed to a fact-check. Disagreeing made people believe the rumor less, often by a good deal, even when the reply was a bare doubt with no source, and even on beliefs bound up with who people are. Party loyalty didn’t blunt it the way many Western studies would lead you to expect. Pointing to a source helped only now and then, and not by much. So in a family group, you don’t have to be an expert, or win, to make a difference. The study measured what people said they believed, right after reading invented chats, and not what they did later, but it is the nearest thing to a family group anyone has tested.

Who says it matters too. A correction is believed more when it comes from someone trusted, and trust may count for more than expertise. In a family, that person is often you. A grandchild who checks things can be trusted in a way that a website run by strangers may never be.

The rest is manners, and Nepali manners mostly point the right way. We found no study of how best to correct a Nepali grandparent, so what follows is ordinary good sense rather than research. In most families a younger person doesn’t contradict an elder in front of everyone. Doing it in the group, before cousins and in-laws, turns a question about rice into a contest the elder can’t afford to lose. Done in private, it’s just a conversation.

Here is how one might go. (The family is made up.) Sabina’s grandfather, her हजुरबुबा, has sent the rice video to the family group. She calls him that evening instead of replying.

“हजुरबुबा, I saw the rice video you sent. It worried me too, so I looked it up.”

He asks what she found.

“Rice like that bounces because of its starch. When it’s hot and pressed into a ball, air gets trapped inside. The food department tested it years ago and found no plastic. It’s just rice. Shall I send you the page?”

He might say the man in the video seemed honest, and she might agree that he did. He might send the page to the group himself the next morning, or never mention it again. Either way, nobody has called him foolish, and he has heard the facts from someone he trusts.

What she left out matters as much. She didn’t say “this is fake” and stop, which leaves the bounce unexplained. She didn’t say “you always fall for these”, which turns a question about rice into a question about him. She began from the worry they share, and that isn’t only kind. It’s probably accurate: a person who forwards a scare about food is usually trying to protect the family. And when they talk in Nepali, every “you” she says to him is हजुर, the most respectful form in everyday speech. Respect in the words leaves room for disagreement in the sentence. The same steps are set out in Talk to an elder in private first.

The group still matters. Everyone in it saw the video, and in studies on social media, people who only watched someone else being corrected came away with more accurate views. The opposite holds too: when nobody speaks, a loud and mistaken few can set the story while everyone else stays quiet. So if a rumor could hurt someone, one calm line in the group, a fact and a link with nothing about the person who posted it, does good well beyond the person you’re talking to. With an elder, it’s often best of all if they post it themselves.

When to let it go

Not every false thing needs an answer. Nobody has the time, and in 2013 an Italian programmer, Alberto Brandolini, put the reason in one line: the energy needed to refute nonsense is an order of magnitude bigger than the energy needed to produce it. (His word was ruder than “nonsense”.) A thirty-second forward can take an evening to untangle.

So two questions come before the reply. Could it hurt someone, through their health, their money, their safety, or what people think of a person or a group? And is it spreading, or about to? If the answer to both is no, let it be. The Handbook adds that the less said about a myth nobody has heard of, the better, since a correction also advertises what it corrects: “If no one has heard of the myth that earwax can dissolve concrete, why correct it in public?”

Some arguments can’t be finished with facts at all. If a disagreement is really about what matters more, freedom or safety, the old way or the new, then no fact-check will settle it, and correcting someone’s values as though they were a wrong date only gives offense. Check the one fact that can be checked, and leave the rest to a longer conversation.

And sometimes you’ll be the one who’s wrong. You’ll correct someone and find, when you look harder, that they were right, or that the matter is less settled than you thought. The decent thing then is to say so, quickly and plainly, in the same place you made the correction. It costs a little pride. It buys the thing that makes every later correction work, which is being someone whose word is good.

The rice will come round again. It already has, in other countries and other years, and some new video will show something real and give it a frightening cause. Disagreeing well was never about making the uncle say you were right. It’s about the next time a ball of rice bounces on his phone, when he might stop and wonder before he forwards it. That happens slowly, one call at a time, and nobody ever gets to be completely sure it worked. Acting anyway, without being sure, is most of what any decision asks.

Found a mistake? Please tell us.