i

Description

La façade de la basilique Saint-Pierre de Rome, sous un ciel bleu, avec la foule au pied des marches.

Source

Gary Todd · original page

Licence

CC0 Public domain

Original photo, unretouched. The background is a version of it stylised by the site.

The devil's advocate

Criticism requested for convenience is easily brushed aside; sincere criticism gets discussed. The method's entire value lies in that difference.

In brief

• What: one person or a small group is tasked with challenging a proposal before the decision, seeking out its weaknesses and defending the opposite option.
• What research says: better performance than consensus without contradiction (Schwenk, 1990); but a played role stimulates less thinking than sincere opposition (Nemeth et al., 2001).
• Three variants: the designated devil's advocate, the red team, independent counter-expertise. The more independent they are, the more valuable they are.
• Condition: the dissenter must be heard, protected from any sanction, and the final decision must respond to their arguments.

On this page

1. Definition
2. The principle: why it works, and when it does not
   1. What research shows
   2. Played role vs. sincere opposition
3. How to do it: step-by-step procedure
   1. Three variants, three levels of independence
4. Examples
   1. The promoter of the faith, and Mother Teresa's lawyer 16th century to 2002
   2. The “Ipcha Mistabra” office in Israeli military intelligence since the 1973 war
   3. Choosing an architecture: adversarial review Illustrative example
5. When to use it, when to avoid it
   1. Use it
   2. Avoid or adapt it
6. Limits and critiques
7. With AI
   1. Detect
   2. Fix
   3. Mitigate
   4. Prompts to use
   5. Precautions
8. Sources and verification
9. See also
   1. Biases this method fights
   2. Complementary methods

Definition#

Fictitious illustration

The devil's advocate is the person who, within a group, is tasked with defending the position opposite to the one taking shape, even without believing in it, so that its robustness can be tested. The expression is now commonplace (“I’m playing devil's advocate”); the method, however, requires rules of engagement: an explicit role, a defined moment, and the group's commitment to respond.

Origin. The expression comes from the Latin advocatus diaboli, the popular name for the promotor fidei (“promoter of the faith”), the officer of the Catholic Church tasked, in canonization trials, with identifying everything that argued against recognizing a saint: the candidate's flaws, doubts about miracles, insufficient evidence. The function took its modern form at the end of the 16th century. It was effectively abolished by John Paul II with the apostolic constitution Divinus Perfectionis Magister of January 25, 1983, which reformed the procedure and made it a less adversarial inquiry. The formal role of the contradictor disappeared on that occasion, and the expression survived in everyday language and management.

The principle: why it works, and when it does not#

A group moving toward a decision tends to look for arguments in its favor and dismiss those that get in the way: this is the terrain of Confirmation bias, Groupthink, and The false consensus effect. Disagreement is socially costly: the person who voices it risks being seen as an opponent. Assigning someone to do it turns a personal risk into a task.

What research shows#

Charles Schwenk (1990) compiled, in a meta-analysis published in Organizational Behavior and Human Decision Processes, experiments comparing, for groups preparing a decision, three approaches: devil's advocacy, “dialectical inquiry” (two opposing plans each defended by a subgroup), and a no-conflict approach based on consensus or expert opinion. Both conflict-based techniques improve group performance compared with the no-conflict approach, with neither one clearly outperforming the other.

Played role vs. sincere opposition#

Charlan Nemeth, Keith Brown, and John Rogers (2001, European Journal of Social Psychology) compared an authentic dissenter (someone who truly defends the opposite opinion) with a designated devil's advocate, in several ways: role known or unknown to the group, opposite position or one consistent with their own. Authentic opposition produced more ideas and better solutions than all devil's advocate variants. Nemeth's earlier work had already indicated that a devil's advocate mainly pushes the group to reinforce its initial position: the objection is rebutted, and the group moves forward more confident, because it knows the objection was “played.” A contradictor with genuine conviction, by contrast, forces reconsideration.

These findings should be read without overextending them. They are laboratory experiments, with short-duration tasks. They do not say devil's advocacy is useless (Schwenk found it helps); they say it helps less than authentic opposition, and that the effect depends on how the group receives it.

How to do it: step-by-step procedure#

Here is devil's advocacy in a decision meeting. Duration: 20 to 45 extra minutes per major decision. Participants: one decision-maker, the group, one designated contradictor (or several, rotating).

1. Choose which decisions are in scope. Not all of them: those that are costly, hard to reverse, or reach unanimity too quickly (architecture, vendor selection, production rollout, launch).
2. Designate the contradictor before the meeting. Someone with enough experience to be credible and who is neither subordinate to the decision-maker nor both judge and party. The role rotates from one decision to the next, to avoid labeling one person as “the complainer.”
3. Give them the case file and time. At least a few days: improvised contradiction during the meeting stays superficial. They can question the plan's authors, look for data, consult outsiders.
4. Ask for a specific deliverable. The three or four most fragile assumptions; the best argument for the opposite option; what would prove the proposal is bad; comparable past failures.
5. Let them speak first, without interruption. The group listens and takes notes. No immediate justification, no tone-based response.
6. Respond point by point, in writing. For each objection: accepted (the plan changes), rejected (with what evidence), or open (with what verification). This is what separates real contradiction from a formality.
7. Decide, then protect the contradictor. The decision belongs to the designated owner. The contradictor faces no consequence for contradicting, and objections not retained remain recorded in the decision file.

Three variants, three levels of independence#

• Designated devil's advocate: a group member who defends a viewpoint they do not necessarily hold. Light and low-cost; the least independent, and the most exposed to the “role play” effect described by Nemeth and colleagues.
• The red team: a separate team tasked with attacking the plan as an adversary would (competitor, hacker, regulator) or searching for its weak points. The term comes from military and intelligence exercises; the U.S. Army formalized instruction in the method in the mid-2000s (sources date its creation from 2004 to 2006), with a dedicated center at Fort Leavenworth. In cybersecurity, it is the team that plays the attacker during a penetration test. More costly, more independent.
• Independent counter-expertise: an external expert, with no stake in the decision, examines the case and provides a written opinion (architecture audit, second opinion, peer review). This is the closest to “authentic” opposition: the expert is not acting a part, they say what they think.

Examples#

When to use it, when to avoid it#

Use it#

• For an important and hard-to-reverse decision.
• When unanimity comes too quickly, or when the decision-maker is highly invested in the proposal (see Authority bias).
• When the organization doesn't know how to say no: the role grants permission.
• Before an external commitment (contract, announcement, production release).

Avoid or adapt it#

• Urgent or reversible decision: the cost in time and team climate outweighs the benefit.
• If the decision-maker has already decided: the session becomes theater, and the team understands that quickly. Better not to hold it than to stage it.
• If the dissenter is the only one carrying the burden: they wear out their credibility. Rotate the role, or replace it with external counter-expertise.
• When the goal is to find risks rather than challenge a choice: the The pre-mortem mobilizes the whole group without assigning anyone.
• When the issue is a number: Reference class forecasting provides a comparison point that discussion does not.

Limits and critiques#

• The role-played approach is not very effective. Nemeth and colleagues (2001) show that an assigned devil's advocate stimulates less thinking than sincere opposition, and can strengthen confidence in the initial position.
• The dissenter is quickly dismissed. “That's their role, they're just saying that to say it,” and the objection is discarded. Hence rotation and written response.
• The social cost. Organized criticism can strain the atmosphere and wear out the dissenter. A group must be able to distinguish criticism of ideas from attacks on people.
• The risk of gridlock. An overly active dissenter can paralyze; the decision still belongs to the designated owner.
• An alibi. “We had a devil's advocate” can be used to justify a decision already made. Only a documented response to objections makes the difference.
• Research limits. Many experiments involve students, short tasks, and groups with no real stakes. Schwenk's meta-analysis dates from 1990.

With AI#

Prompts to use#

1. First-line dissenter. Before the session, to prepare the file. The human dissenter then completes it.

Here is a decision being prepared: [description, rejected options, arguments in favor, constraints].
Play devil's advocate, unsparing but not in bad faith.
1. Give the three most fragile assumptions in the proposal, and what would invalidate them.
2. Build the best possible argument for the opposite option, as if you were genuinely defending it.
3. Give ten specific questions to ask the authors.
4. State what you cannot judge due to lack of information, rather than assuming.
Do not invent any precedent or any number.

2. Test the response to objections. After objections, to check whether the response holds up.

Here is a list of objections made to a decision: [list] and the team's responses: [responses].
For each objection, say whether the response truly addresses it, sidesteps it, or ignores it. Classify it: accepted, rebutted with evidence, rebutted without evidence, open.
End with the objections whose response is weakest. Do not try to reassure.

Precautions#

• AI follows your framing: if you present the decision as excellent, its objections will be timid. Describe the decision neutrally.
• Its objections are plausible but not always well-founded: have the important ones checked by someone who knows the topic.
• It may invent precedents or references: retain only what is verified.
• Do not submit confidential files to an unauthorized tool.

Sources and verification#

• Schwenk, C. R. (1990), “Effects of devil's advocacy and dialectical inquiry on decision making: A meta-analysis,” Organizational Behavior and Human Decision Processes, 47(1), 161-176.
• Nemeth, C., Brown, K. & Rogers, J. (2001), “Devil's advocate versus authentic dissent: stimulating quantity and quality,” European Journal of Social Psychology, 31(6), 707-720.
• Nemeth, C. J. (2018), In Defense of Troublemakers: The Power of Dissent in Life and Business, Basic Books — synthesis of her work.
• John Paul II (1983), apostolic constitution Divinus Perfectionis Magister (January 25, 1983) — reform of the canonization procedure.
• Encyclopaedia Britannica, “devil's advocate” — origin of the expression.
• Hitchens, C. (1995), The Missionary Position: Mother Teresa in Theory and Practice, Verso — the critical work; his being contacted by the Church is recounted in his later articles.
• Bar-Joseph, U. (2005), The Watchman Fell Asleep: The Surprise of Yom Kippur and Its Sources, SUNY Press — the 1973 surprise and the reform of Israeli intelligence assessment (to consult for the creation of the Ipcha Mistabra office).
• Janis, I. L. (1982), Groupthink, Houghton Mifflin — groupthink and its remedies, including the dissenter role.

See also#