Irving Janis coined the term groupthink in 1972 following his analysis of a sequence of catastrophic foreign policy decisions — the Bay of Pigs invasion, the failure to anticipate Pearl Harbor, the escalation of the Vietnam War — that shared a structural similarity. The decision-makers were intelligent. The information needed to avoid the mistake was, in most cases, available within the group. And the consensus was wrong. Janis identified the mechanism: in groups with high cohesion and high external pressure, the social cost of dissent becomes high enough that members suppress their reservations, validate one another’s assumptions rather than challenging them, and arrive at a convergence that none of them, individually, would necessarily have endorsed.

The insight that has transferred imperfectly from Janis’s original research into the management literature is that groupthink is not primarily a phenomenon of weak or unsophisticated groups. It is a phenomenon of cohesive, high-functioning groups operating under pressure — which is an accurate description of most senior leadership teams. The conditions that make leadership teams work well — trust, shared commitment, aligned values, high stakes — are also the conditions that make them susceptible to the specific failure Janis described. Cohesion is both the foundation of effective collective decision-making and the precondition for groupthink.

What Groupthink Actually Requires

Janis identified eight symptoms of groupthink, but the underlying mechanism can be compressed to two. First: the group develops an implicit consensus before the deliberation has concluded, and subsequent discussion functions to confirm that consensus rather than to test it. Second: the social cost of voicing a dissenting view increases as the implicit consensus solidifies — not because anyone explicitly punishes dissent, but because challenging the emerging direction requires more emotional and social capital than most group members are willing to spend on an uncertain outcome. The person who has reservations about the direction the group is moving performs a rapid calculation: the probability that my objection changes the outcome, multiplied by the value of preventing the bad decision, versus the social cost of being the one who held things up, expressed doubt, or seemed not to trust the group’s judgment. In most leadership teams, most of the time, the calculation resolves in favour of compliance.

Philip Tetlock’s research on expert prediction and forecasting adds a precise term for the behaviour that calculation produces: epistemic cowardice. In his studies of forecasting accuracy across thousands of predictions by domain experts, Tetlock found that forecasters who gave vague, hedged, non-committal predictions were performing a social manoeuvre rather than an epistemic one — protecting themselves from being demonstrably wrong while also protecting their standing in the group from the risk that comes with staking out a clear, falsifiable position. The collaborative forecasting research Tetlock and colleagues subsequently conducted found that groups which explicitly reward calibrated, specific predictions and track prediction accuracy over time outperform expert panels substantially — because they change the social incentives around precise epistemic commitment. The groups where it is safe to be wrong are the groups that are most often right.

The Conditions for Genuine Collective Intelligence

James Surowiecki’s analysis of when crowds are wise and when they are not identifies three conditions that need to be present for a group to outperform its most capable individual members: diversity of information, independence of judgment, and a method of aggregating individual views that doesn’t destroy the independence before the aggregation happens. The third condition is the one that most leadership teams violate most systematically. The standard meeting format — in which views are expressed sequentially, in the presence of a leader whose position is known, with the expectation that the discussion will resolve into a direction rather than remain in productive ambiguity — destroys the independence that would allow genuine collective intelligence to emerge. Each person in the room adjusts their expressed view toward what they believe the group and the leader want to hear before they express it. The aggregation is corrupted before it begins.

The research on what separates leadership teams that make genuinely good decisions from those that make consistently mediocre ones finds a consistent differentiator: the degree to which the team has built structures that protect independent judgment before aggregation. This includes pre-meeting written positions that are shared simultaneously rather than stated sequentially, explicit red team assignments in which a designated member’s role is to challenge the emerging direction, and the specific practice of making the most senior person’s view known last rather than first. None of these are exotic. All of them require deliberate structural design, because the default meeting format produces the inverse of each.

What Psychological Safety Adds — and What It Doesn’t

Amy Edmondson’s research on psychological safety, developed across studies of medical teams, manufacturing teams, and strategic leadership teams, establishes that teams with high psychological safety — where members believe it is safe to speak up, disagree, and admit uncertainty without social penalty — make fewer errors, learn faster, and perform better on complex tasks than those without it. The relationship to the groupthink analysis is direct: psychological safety lowers the social cost of dissent, which is the mechanism that groupthink requires to operate.

What the psychological safety literature does not fully address is the distinction between the kind of safety that allows low-stakes candour and the kind that allows high-stakes dissent. Most leadership teams that describe themselves as having high psychological safety are describing a context in which it is safe to raise operational issues, give honest feedback on presentations, and express uncertainty about execution details. This is genuine safety, and it is genuinely valuable. It is not the same as the safety required to challenge the strategic direction the CEO visibly favours, to express doubt about a major acquisition in the week before it closes, or to argue against the consensus of a senior leadership team that has already invested six months in a particular path. The situations where groupthink is most destructive are exactly the situations where the social cost of dissent is highest — and those situations require a more deliberate structural intervention than the general norm of psychological safety provides.

How the Consensus Trap Plays Out in Practice

The most common manifestation in senior leadership teams is not the dramatic, identifiable convergence on a catastrophically wrong decision. It is the sustained avoidance of the consequential question. The item that keeps appearing on meeting agendas and keeps getting displaced by operational discussion. The strategic uncertainty that everyone in the room is aware of and that no meeting conversation has directly addressed. The difficult personnel situation involving a senior person with strong sponsor relationships. The business model assumption that the plan depends on and that no one has tested explicitly.

Tetlock’s research on forecasting is relevant here: the questions that forecasting teams perform worst on are not the genuinely unpredictable ones but the ones where the answer is uncomfortable and where the social cost of stating the honest prediction is high. Leadership teams avoid the uncomfortable strategic question for the same reason that forecasters hedge on the uncomfortable prediction — not because the question is too complex to address but because the social dynamics around addressing it are too costly to navigate without structural protection.

The cost is not always visible at the time. Strategic questions that are avoided rather than resolved do not produce immediate consequences. They produce consequences when the environment changes in the direction the avoided question would have anticipated, or when the competitive or operational context makes the unaddressed assumption no longer sustainable. By then, the leadership team has typically forgotten that the question was avoided. The outcome reads as an external event rather than a predictable consequence of a pattern of collective decision avoidance.

The Structural Intervention

The leadership team that wants to escape the consensus trap needs to change the conditions under which difficult questions are raised and addressed, not the people raising them. This means structural changes to how the team generates and aggregates views, rather than cultural exhortations to be bolder or more candid. The most useful structural changes are the ones that lower the social cost of independent judgment before the group’s view has solidified: pre-read positions committed to paper before the meeting, red team assignments with explicit permission to argue the case against, and the specific practice of having the most powerful person in the room express their view last.

These structural interventions are more effective when the executive who leads the team has the physiological regulation to tolerate genuine dissent without the defensive activation that makes challenge feel like a threat. The leader whose cortisol response to challenge is well-regulated can receive a dissenting view as information. The leader whose autonomic system is chronically elevated experiences the same challenge as a social threat and responds accordingly — which changes the information available to the dissenter before they have said a word. The structural change and the physiological change are not alternatives. They work in the same direction, and they compound when they are present together. Four slots available monthly. Apply here.

Frequently Asked Questions

How do you distinguish productive team cohesion from groupthink?

Janis’s original framework identified this as one of the most practically difficult distinctions to make from inside the group. The cohesion that produces groupthink feels, from the inside, like the cohesion that produces effective collaboration — the mutual trust, shared language, and aligned commitment that make high-functioning teams work. The diagnostic that most reliably distinguishes the two is not the quality of the relationship but the fate of dissenting information. In a cohesive team that is not in groupthink, dissenting observations are engaged — challenged, examined, tested — and the process of engaging with them changes the group’s view at some measurable rate. In a groupthink-susceptible team, dissenting observations are acknowledged and then effectively ignored: the implicit consensus continues to develop as if the challenge had not been raised, and members who raised it learn not to raise similar challenges again. The behaviour of the group in response to the first few expressions of genuine dissent is the most diagnostic signal available about which dynamic the team is in.

Does the research support having an explicit devil’s advocate role in leadership meetings?

The evidence is mixed and the nuance matters. Assigning a rotating devil’s advocate role does improve decision quality in laboratory studies compared to no structural intervention, but it performs worse than genuine minority dissent — situations where a member actually holds and articulates a contrary view rather than performing one. Charlan Nemeth’s research on minority influence found that genuine dissent, even when the minority is ultimately wrong, produces more thorough consideration of the problem and more original thinking than devil’s advocacy, because the group processes genuine dissent differently than assigned performance of dissent. The practical implication is that the devil’s advocate role is useful as a floor — better than nothing — but that the more powerful intervention is building the conditions in which genuine dissent occurs organically. That means the structural changes described above, combined with the leader’s visible track record of engaging with rather than punishing views that challenge their own.

What is the relationship between decision speed and decision quality in leadership teams?

The research does not support the intuitive view that faster decisions are necessarily lower quality or that more deliberation is always better. Kathleen Eisenhardt’s research on strategic decision-making in fast-moving environments found that the highest-performing leadership teams in her studies made decisions both faster and better than their competitors — and the mechanism was not less deliberation but more structured deliberation. They tracked more alternative options simultaneously, sought more outside perspectives, and used conflict resolution mechanisms more consistently than slower-deciding teams. The consensus trap typically produces slow decisions, not fast ones — because the avoidance of the difficult question means the decision keeps being deferred rather than made. The intervention that allows the team to address the difficult question directly tends to produce both faster resolution and better outcomes, because the bottleneck was the avoidance, not the complexity of the question itself.

How does the SEAM protocol specifically address consensus dynamics in leadership teams?

The SEAM approach engages the consensus problem through the individual executive’s physiological regulation rather than through team process design. The reasoning is that the most significant lever available to a single executive is their own capacity to tolerate the discomfort of genuine disagreement — both to express it and to receive it. An executive who can hold their own dissenting view calmly in the face of emerging consensus, and who can receive a challenge to their position without the defensive activation that shuts the challenge down, changes the information dynamics of every room they are in. The Clarity Index includes a dimension specifically assessing the executive’s self-reported experience of high-stakes disagreement — how they feel going into conversations where the risk of conflict is high, how they tend to behave in the moment of challenge, and what happens physiologically in those situations. The 90-day protocol’s HRV biofeedback component directly builds the autonomic regulation capacity that makes genuine tolerance of dissent — as both giver and receiver — physiologically sustainable rather than effortful.