Someone leans back, half-smiles, and says it: "Let me just play devil's advocate for a second." Everyone relaxes a little. The objection that follows is sharp, maybe even correct, and the room nods along. Then the meeting moves on and decides exactly what it was going to decide anyway. Nobody lied. Everybody did their part. And the one thing the ritual was supposed to give you, a real test of the idea, never happened. The uncomfortable part is that the people in that room usually leave feeling more sure of themselves, not less.

A ritual with a long pedigree

The phrase is old. For centuries the Catholic Church kept an official called the advocatus diaboli, the Promoter of the Faith, whose job was to argue against canonizing a candidate for sainthood. He looked for holes in the miracles and flaws in the character. The office was formalized in the late sixteenth century, and in 1983 Pope John Paul II cut its role sharply. Canonizations then sped up dramatically. Make of that what you will.

The modern office version comes from a reasonable idea. When Irving Janis described groupthink in the 1970s, one remedy he proposed was assigning someone to argue the other side. If a group's danger is premature agreement, then appoint a disagreer. It's tidy and cheap, and it costs nobody any social capital. That last part turns out to be the whole problem.

What the research actually found

The person who has done the most to test this is Charlan Nemeth, a psychologist at UC Berkeley who spent decades studying how dissent affects group thinking. Her early work on minority influence found that when even one person truly disagrees with a majority, the others don't just consider that person's view. They start thinking more broadly overall. They look for more information, weigh more options, and come up with more original solutions, even ones the dissenter never suggested. Disagreement works less like a counterargument and more like a jolt that knocks the group out of its single track.

So the natural question was whether you could fake it, and get the jolt without the friction by assigning the role.

In a set of experiments published in 2001, Nemeth and her colleagues compared groups exposed to authentic dissent, where someone actually believed the opposing position, with groups exposed to a devil's advocate arguing the same position as an assignment. The arguments could be the same. The results weren't. Authentic dissent produced more ideas and more real consideration of alternatives. The role-played version did noticeably less. Nemeth has also argued that the devil's advocate can backfire: when people beat back an objection they know isn't sincere, they may come out more attached to their original view, because now it feels like it has survived a challenge.

It's like a vaccine made from a dead virus. The body learns to fight off something that was never going to hurt it.

Why the fake version fails

The mechanism is something social psychologists have understood for a long time. Harold Kelley's discounting principle holds that when there's an obvious outside reason for someone's behavior, we play down what the behavior says about anything else. If a colleague argues against the plan because they were asked to, the objection explains itself. You don't have to ask whether they're seeing something you missed. You already know why they're saying it.

Authentic dissent can't be filed away so easily. When someone disagrees and you know they mean it, at some social cost to themselves, your mind has to deal with a harder question: why does a reasonable person believe this? That question is what drives the broader thinking Nemeth kept finding. You start building a model of their reasoning, and in doing that you see angles that weren't on the table before.

There's a second, quieter reason. The devil's advocate rarely argues with any conviction. They know the role is temporary. They raise the objection, it gets an answer, and they let it go. Usually there's a little social signal built in, the half-smile or the "just to be thorough," that tells the room not to worry. A real dissenter doesn't let go after one round. They come back with "I hear that, but I still don't think it handles the case where..." That persistence is uncomfortable, and the discomfort is exactly what makes people actually think.

The emotional truth underneath

The more personal reason "let me play devil's advocate" is so popular is that it protects the speaker. Saying "I think this is wrong" in front of your boss and peers is risky. You might be the only one. You might be wrong in public. You might be seen as difficult. The devil's advocate frame lets you voice the thought while keeping deniability: it's not me, it's the role.

That's very human, and often it's the best a person can manage in a room that doesn't feel safe. But it costs something. When you wrap a real concern in a costume, you're telling the room not to take it seriously. Your actual judgment, the thing you're paid for and the thing that might have saved the project, gets delivered in a way that makes it easy to ignore.

The flip side is worth noticing too. If you lead meetings and people regularly reach for the devil's advocate frame, they're telling you something about the room. People don't wear costumes where it's safe to show up as themselves.

What works instead

None of this means structured disagreement is useless. It means the structure has to bring real views to the surface rather than stage pretend ones. A few approaches hold up better.

Ask for the doubt people already have. Instead of assigning a contrarian, ask everyone: what's the part of this you're least sure about? That question presumes doubt exists and makes it normal, and it pulls out reservations people actually hold.

Collect positions before discussion. When people write down their honest view privately before anyone speaks, the room can see disagreement it would otherwise never know about. Then you aren't inventing dissent. You're finding it.

Find the person who actually disagrees, and give them the floor. If someone has pushed back on this plan in a hallway, a doc comment, or a one-on-one, invite them to make the full case, as themselves, with time to do it properly.

Separate the idea from its author. Real dissent is easier when criticism is clearly aimed at the plan, not the person who proposed it. A framing like "let's try to break this so it doesn't break later" gives people permission to be sincere.

Your next moves

  • Retire the phrase for one week. Every time you're about to say "let me play devil's advocate," say "here's what actually worries me" instead, and only say it if something does. Notice how the room responds differently.
  • In your next meeting, add one question before the decision: "What's the strongest reason this could be wrong, from someone who actually believes it?" Then wait a full ten seconds before filling the silence.
  • Before your next big group decision, run a silent poll. Have everyone send their honest view, for or against and how confident, privately or in writing, before discussion opens. Share the spread with the group.
  • Look back at one recent decision. Was there an objection raised only "for argument's sake"? Write it down and ask yourself whether anyone ever tested it for real. If not, take it to the person who owns the decision.
  • If you lead the meeting, thank the first sincere dissenter out loud, especially if they turn out to be wrong. That one reaction teaches the room more than any assigned role.

Seeing the pattern in your own meetings

The hard part is that from inside a meeting, staged dissent and real dissent look almost the same. Both sound like objections. The difference shows up afterward, in whether the objection changed anything, and whether the person who raised it meant it. That's the kind of thing MeetingMortem is built to help you see: a short, honest look back at how a decision actually got made, including whose doubts were voiced, whose were dressed up, and whose stayed in their head. If you want your next meeting to include the disagreement you need and not just the version that's safe to say, you can try it at meetingmortem.lumenlabs.works.