Post
JA EN

Overconfidence Is Not a Bug, It's a Feature — How Evolutionary Psychology Explains Our 'Runaway Confidence'

Posted
By
19 min read
Overconfidence Is Not a Bug, It's a Feature — How Evolutionary Psychology Explains Our 'Runaway Confidence'
  • Target audience: Engineers and knowledge workers curious about the “why” behind overconfidence
  • Prerequisites: None
  • Reading time: 18 minutes

Series article: This is the first installment in a deep-dive series expanding on “The Scientific Risk of ‘Just Be Confident’.” The parent article surveys the difference between fragile and resilient confidence. This article digs into the evolutionary origins of why humans become overconfident in the first place.

Overview

Why do humans so relentlessly overestimate themselves? Extensive research confirms that overconfidence is dangerous. Yet it never disappears from the human population. The reason is simple — overconfidence is not a bug in human cognition; it is a feature that once worked remarkably well.

In 2011, Johnson & Fowler published an evolutionary model in Nature that mathematically demonstrated overconfidence was not weeded out by natural selection but was, in fact, actively selected for. The problem is that the environment this feature was designed for is radically different from the modern world.

This article dissects the mechanism by which overconfidence was evolutionarily advantageous, the feedback structures in ancestral environments that kept it in check, and the process by which those feedback structures have collapsed in modern society.

1. “When Rewards Exceed Twice the Cost, Overconfidence Is the Optimal Strategy” — The Johnson & Fowler Model

1.1 The Core of the Mathematical Model

In 2011, Dominic Johnson (evolutionary biologist, Harvard) and James Fowler (political scientist, UC San Diego) published an evolutionary model in Nature1. The model simulated resource competition between individuals and mathematically identified the conditions under which overconfidence becomes an Evolutionarily Stable Strategy (ESS).

The model’s structure works as follows:

  • Two individuals compete over a resource
  • Each individual estimates their own ability and decides whether to compete based on a comparison with their opponent
  • Overconfident individuals overestimate their own ability, so they enter competitions even when objectively disadvantaged
  • Accurately self-assessing individuals avoid unfavorable competitions

Intuitively, accurate self-assessment seems superior — you avoid wasted effort and focus on winnable contests. But the model’s results were the exact opposite.

1.2 Why “Overconfidence” Beats “Accurate Assessment”

The critical condition identified by Johnson & Fowler’s model is this1:

When the benefit of acquiring a resource (b) is at least twice the cost of competing (c) — that is, b ≥ 2c — overconfidence becomes the optimal strategy.

Why? Accurately self-assessing individuals avoid objectively unfavorable competitions. But some of those apparently unfavorable competitions are actually winnable — the opponent might be ill, injured, or simply distracted. Overconfident individuals enter these “uncertain but potentially winnable” competitions. As a result, they capture more resources overall.

Of course, overconfidence also produces unnecessary defeats. But when the benefit of resource acquisition is at least twice the competition cost, the gains from extra victories outweigh the losses from extra defeats.

flowchart TB
    A["Competition<br>over resources"] --> B{"Self-assessment"}
    B -->|"Accurate<br>assessment"| C["Avoids unfavorable<br>competitions"]
    B -->|"Overconfidence"| D["Enters competitions<br>even when disadvantaged"]
    C --> E["Captures only<br>certain victories"]
    D --> F["Extra defeats<br>(cost incurred)"]
    D --> G["Captures uncertain<br>victories too<br>(extra benefit)"]
    F --> H{"b ≥ 2c ?"}
    G --> H
    H -->|"Yes"| I["Overconfidence is<br>optimal strategy<br>(benefit > cost)"]
    H -->|"No"| J["Accurate assessment<br>is superior"]

    style I stroke:#51cf66,stroke-width:3px
    style F stroke:#ff6b6b,stroke-width:3px

In ancestral environments, the benefits of resources like food, territory, and mates frequently far exceeded the costs of competition (injury, energy expenditure). In other words, the b ≥ 2c condition was regularly met.

1.3 Catastrophic for Groups, Optimal for Individuals

An important caveat: Johnson & Fowler’s model shows that overconfidence maximizes individual fitness, not that it is optimal for the group as a whole1.

In fact, the paper explicitly states that overconfidence can cause wars, market bubbles, financial collapses, and policy failures at the collective level. But natural selection operates at the individual level — if a trait benefits an individual’s genes, it will spread through the population even if it is catastrophic for the group.

This is the classic “Tragedy of the Commons” structure. Each individual maximizes personal gain, and the collective suffers. Overconfidence is a cognitive Tragedy of the Commons.

2. Three Evolutionary Advantages of Overconfidence

Research complementing Johnson & Fowler’s model has identified the specific mechanisms by which overconfidence enhanced fitness.

2.1 Action Booster — Lowering the Threshold for Engagement

The most direct benefit of overconfidence is that it lowers the threshold for action. It drives you to act under uncertainty, persist through difficulty, and resist giving up when facing obstacles.

In hunter-gatherer societies, uncertain but potentially high-reward opportunities were a daily reality — exploring new hunting grounds, testing unknown plants, negotiating with other groups. Accurately self-assessing individuals would frequently hold back, calculating risks precisely. Overconfident individuals had a lower bar for “just trying it,” allowing them to extract more value from uncertain opportunities.

2.2 Social Signal — The Power of Bluffing

Anderson, Brion, Moore & Kennedy (2012) demonstrated across six studies how overconfidence contributes to acquiring social status2.

Studies 1–3 showed that overconfident individuals gained higher social status in both short-term and long-term settings. The mechanism was straightforward — overconfident individuals were perceived by others as more competent than they actually were. In short, bluffing worked.

Studies 5 and 6 further revealed that the motivation to acquire status promotes overconfidence. Overconfidence raises status, and the desire for status fuels further overconfidence — a positive feedback loop.

In ancestral environments, social status meant access to food, mates, and allies. If bluffing could secure status, whether you were actually competent was secondary.

2.3 Stress Buffer — Suppressing Anxiety

Overconfidence also carries psychological benefits. Taylor & Brown (1988) reported that mentally healthy individuals tend to overestimate themselves, maintain unrealistic optimism about the future, and harbor illusions of control — what they called “positive illusions”3. In uncertain environments, these illusions suppress anxiety and free up cognitive resources for decision-making.

In ancestral environments, paralysis from anxiety was literally life-threatening. When confronting a predator, an individual who froze in fear was less likely to survive than one who believed “I can handle this” and took action. Overconfidence was the psychological mechanism executing the evolutionary command: “Move. Now.”

3. Ancestral Brakes — Why Overconfidence Didn’t “Run Away”

This raises a question: if overconfidence was so advantageous, why didn’t humans become completely overconfident? Why did we retain any capacity for realistic self-assessment?

The answer is that ancestral environments contained powerful feedback mechanisms that kept overconfidence in check.

3.1 Dunbar’s Number — A World Where “Everyone Knows Everyone”

According to anthropologist Robin Dunbar’s “Social Brain Hypothesis,” the natural group size predicted by the human neocortex is approximately 1504. This “Dunbar’s number” has been repeatedly confirmed across 23 studies spanning hunter-gatherer bands, military units (companies), and rural communities — across cultures and eras.

In a group of 150, everyone knows everyone else’s face, name, and reputation. What happens when an overconfident individual bluffs in this environment?

  1. Immediate feedback: Failure is witnessed by everyone. “He said he could do it, and he couldn’t.”
  2. Reputation transparency: The consistency between past performance and claims is constantly monitored
  3. No escape: Moving to another community and “resetting” your reputation is extremely difficult

In a group of 150, the cost of bluffing is very high. Overconfidence offers the benefit of encouraging competitive engagement, but it carries the cost of reputation damage when the bluff is exposed. This balance calibrated overconfidence to an “optimal level.”

3.2 Gossip — Humanity’s Oldest Reputation Management System

Dunbar (1996) argued in Grooming, Gossip, and the Evolution of Language that gossip evolved as a mechanism for maintaining large social groups5. Primates maintain social bonds through grooming, but when group size reaches 150, the time required for grooming becomes prohibitive. Linguistic gossip evolved as an efficient substitute.

Gossip’s function is far from mere entertainment6:

  • Promoting cooperation: Sharing information like “he keeps his promises” or “she cheats” reinforces cooperation and punishes defectors
  • Suppressing free riders: Identifying and sanctioning individuals who take benefits without contributing
  • Correcting overconfidence: Spreading the reputation that “he’s all talk with no substance” raises the cost of bluffing

Wu, Balliet & Van Lange (2016) confirmed that reputation-based indirect reciprocity is the foundation of large-scale human cooperation6. In hunter-gatherer societies, reputation information can propagate through three degrees of separation (friend of friend of friend)7. In a group of 150, that effectively reaches everyone.

3.3 The Three Conditions of Effective Feedback

The feedback characteristics of ancestral environments can be summarized in three conditions:

ConditionAncestral EnvironmentEffect
ImmediacyShort interval between action and outcomeConsequences of overconfidence visible instantly
TransparencyEveryone is a witnessBluffs are easily exposed
ConsequentialityReputation loss → reduced resource accessThe cost of bluffing is tangible

When all three conditions are met, overconfidence is calibrated to an “optimal level” where the benefits of action and the costs of exposed bluffs reach equilibrium. The problem is that in modern society, all three conditions have collapsed.

4. Feedback Collapse — The Structure That Lets Overconfidence Run Wild in Modern Society

4.1 Anonymity and Large Groups — A World Where “You Won’t Get Caught”

Modern group sizes exceed Dunbar’s number of 150 by orders of magnitude. Companies have thousands to tens of thousands of employees; social media followings reach millions. In this environment:

  • Delayed feedback: Failures take a long time to become visible (or never do)
  • Anonymity: Individuals disappear into the crowd, making reputation tracking difficult
  • Resettable: Transferring departments, changing jobs, or creating new social media accounts lets you “reset” your reputation

Cinelli et al. (2021) demonstrated in PNAS that echo chambers form on social media, exposing users exclusively to information that reinforces their existing beliefs8. Overconfident individuals exist in an environment where they can selectively consume affirming information and block negative feedback.

4.2 Hierarchical Organizations — A World Where “No One Can Speak Up”

Chamorro-Premuzic (2013) argued that many organizations mistake overconfidence for leadership potential in leader selection9. Despite the correlation between confidence and competence being only 0.3 (about 9% shared variance), organizations continue to use confidence as a proxy for ability.

In hierarchical organizations, feedback to superiors is structurally suppressed:

  • Power gradient: Subordinates find it difficult to point out a boss’s overconfidence
  • Distorted success attribution: Organizational successes are credited to leaders; failures are attributed to external factors or subordinates
  • Selective information transmission: Bad news gets filtered out; only good news travels up

As a result, overconfidence accumulates and becomes increasingly resistant to correction the higher you go in an organization.

4.3 The Success Feedback Loop — Overconfidence Breeding Overconfidence

flowchart TB
    A["Overconfident<br>demeanor"] --> B["Perceived as competent<br>by others<br>(Anderson et al., 2012)"]
    B --> C["Promotion /<br>status acquisition"]
    C --> D["Accumulation of<br>success experiences"]
    D --> E["Further<br>overconfidence"]
    E --> F["Dismissing dissent /<br>blocking feedback"]
    F --> A

    G["Ancestral brakes<br>(gossip / immediate feedback)"] -.->|"Non-functional<br>in modernity"| X["Disabled"]

    style A stroke:#ff6b6b,stroke-width:3px
    style F stroke:#ffa94d,stroke-width:3px
    style X stroke:#868e96,stroke-width:3px

As Anderson et al. (2012) demonstrated, overconfidence elevates social status2. Elevated status accumulates success experiences, which breeds further overconfidence. And as overconfidence deepens, the individual begins dismissing dissent and actively blocking feedback.

In ancestral environments, this positive feedback loop was braked by gossip and immediate feedback. In modern society, the brakes don’t work. The result: overconfidence accelerates unchecked.

5. Hubris Syndrome — The “Final Form” of Overconfidence

5.1 Power Changes Personality

The endpoint of unchecked overconfidence is what neurologist and politician David Owen called “Hubris Syndrome”10.

Owen & Davidson (2009) analyzed U.S. presidents and British prime ministers over the past century and described the mechanism by which accumulated power and success transform personality. Their proposed 14 diagnostic criteria include:

  1. A tendency to see the world as a stage for exercising power and seeking glory
  2. A bias toward actions that enhance self-image
  3. Excessive concern with image and appearance
  4. A messianic manner of speaking and an exalted demeanor
  5. Identification of self with the organization
  6. Frequent use of the royal “we”
  7. Excessive confidence in one’s own judgment
  8. Contempt for the advice of others
  9. “Hubristic incompetence” — excessive confidence leading to inattention to detail

5.2 Acquired and Reversible

Two features of hubris syndrome are particularly noteworthy10:

First, it is acquired. Unlike narcissism or antisocial personality disorder, hubris syndrome is not present before power is obtained. Originally psychologically healthy individuals are transformed by the accumulation of power and success.

Second, it is reversible. Symptoms tend to diminish when power is lost. This suggests that hubris syndrome is not a personal character flaw but a structural cognitive distortion produced by the environment (power structure x absence of feedback).

In an engineering context, the process by which a tech lead or architect — after a streak of successes — begins to believe “my judgment is always right,” dismisses code review comments, and overrules team concerns is a day-to-day manifestation of hubris syndrome.

5.3 The Daedalus Trust — 14 Symptoms

Owen founded the Daedalus Trust to raise awareness of hubris syndrome and research organizational countermeasures against overconfidence in positions of power11. A diagnosis of hubris syndrome requires at least 3 of the 14 symptoms, including at least one of 5 unique symptoms specific to the condition.

The critical point is that these symptoms arise not from individual malice or incompetence, but from the structure of the environment. Anyone given power in a feedback-free environment can develop hubris syndrome.

6. Group-Level Catastrophes — Consequences of Evolutionary Mismatch

6.1 Historical Cases Reinterpreted

The catastrophes caused by overconfidence at the group level are innumerable12:

CaseManifestation of OverconfidenceOutcome
World War INational leaders: “It will be over by Christmas”4 years, ~20 million dead
2008 Financial CrisisFinancial institutions: “Risks are under control”Global economic collapse, trillions in losses
Sydney Opera HouseEstimate: 4 years / AU$7MCompleted in 14 years / AU$102M
Challenger Disaster“The O-ring issue is minor”Launch failure, 7 crew killed

What all of these share is a structure in which decision-makers ignored feedback (warnings, dissent, data). And that ignoring was not born of malice — it was the result of overconfidence degrading the capacity to receive feedback.

6.2 Everyday Runaway in Engineering

Project estimates consistently miss. Review comments are dismissed as “minor issues.” You are certain your design is correct and refuse to consider alternatives. You have failed the same way three times before, but you believe “this time is different.”

All of these are the result of evolutionarily advantageous overconfidence running wild in a modern environment starved of feedback. In the world of software engineering, the Planning Fallacy first articulated by Kahneman & Tversky (1979) occurs daily13 — you are “confident” you can estimate the next task accurately, regardless of your track record.

7. Prescriptions — Rebuilding Ancestral Brakes for the Modern World

Overconfidence is an evolutionary feature, and willpower alone cannot overcome it. Still, we can consciously rebuild the three feedback conditions that existed in ancestral environments — immediacy, transparency, and consequentiality.

7.1 The Premortem — Start with “This Project Has Failed”

Gary Klein’s (2007) premortem analysis asks the team to imagine, before a project begins, that “this project ended in spectacular failure” — then list the reasons why14. The underlying research by Mitchell, Russo & Pennington (1989) found that simply imagining an event has already occurred improves the ability to identify causes by 30%.

This is an artificial substitute for ancestral “immediate feedback.” By simulating failure in advance, it front-loads feedback that would normally only arrive after a project’s conclusion.

7.2 Red Teams — Institutionalizing the Role of “Official Dissenter”

The red team methodology, developed in the U.S. military, embeds an “official dissenter” into the decision-making process. Specific members are explicitly assigned the role of “find flaws in this plan.”

This is the institutionalization of the ancestral “gossip” function. In small groups, when someone made an overconfident claim, others would apply the brakes: “Didn’t you say the same thing last time and fail?” In modern organizations, power gradients disable this brake. Red teams bypass the power gradient by granting official “authority to brake.”

7.3 Intellectual Humility — Upgrading Metacognition

Intellectual Humility — defined as “the degree to which one recognizes that one’s beliefs may be wrong” — is a cognitive vaccine against overconfidence15. Leary et al.’s research found that people high in intellectual humility were more sensitive to the strength of persuasive arguments and possessed more general knowledge.

The practice of intellectual humility is straightforward — regularly ask yourself: “If I were wrong, what would be the most likely reason?” This is a method of providing yourself with the “Are you sure about that?” feedback that ancestral environments delivered automatically through the social group.

7.4 Placing Trust Outside the Ego — Means Efficacy

The most structural countermeasure is shifting the object of trust from “myself” to “my methods.” Means Efficacy, a concept proposed by organizational psychologist Dov Eden, refers to the belief that “this tool/method works” rather than “I can do it”16.

Why means efficacy prevents overconfidence from spiraling is clear — the ego is not involved. When you believe “I’m great,” failure becomes an identity crisis. When you believe “this method is good,” failure is just a signal to change methods. The self remains unscathed. (Means efficacy is explored in depth in article 4 of this series.)

Conclusion

Overconfidence is not a human bug. It was a feature honed by evolution — a “booster rocket for action,” a “social status elevator,” and an “anxiety brake” — that worked effectively in ancestral environments.

But that feature was designed under the environmental conditions of small groups of 50–150, immediate feedback, and reputation transparency. Modern society — with its anonymous mega-groups, delayed feedback, hierarchical information filtering, and social media echo chambers — has disabled the brakes on overconfidence.

Without brakes, overconfidence accelerates to its terminal state: hubris syndrome — power transforming personality. But this is not an individual defect; it is a structural cognitive distortion produced by the environment.

The problem is not overconfidence itself, but the collapse of the feedback structures that once kept it in check.

This is why the prescription is not “reduce your confidence” but “rebuild your feedback.” Premortems, red teams, intellectual humility, and means efficacy — placing trust outside the ego — are all attempts to consciously reconstruct, in modern society, the brakes that ancestral environments provided automatically.

Overconfidence was a feature. But the environment it was designed for no longer exists. Understanding the feature and redesigning the brakes for our new environment — that is the wisdom for coexisting with our evolutionary inheritance.

For more on this topic, see the following articles:

References

References corresponding to the citation numbers used in this article, listed in order.

Additional References (not cited by number in the text)

On citation accuracy: The studies cited in this article have been verified through the following methods:

  • Confirmation via academic databases (PubMed, Google Scholar, ScienceDirect, etc.)
  • Verification of paper details on official journal websites
  • Cross-validation through multiple independent sources (academic media, official research institution announcements, etc.)

For some papers, direct access to the full-text PDF may be restricted, but the abstract, DOI, author information, and key findings have been confirmed through official academic databases and trusted secondary sources.

  1. The Evolution of Overconfidence - Johnson, D.D.P. & Fowler, J.H., Nature, 477, 317-320 (2011). [Reliability: High] ↩︎ ↩︎2 ↩︎3

  2. A Status-Enhancement Account of Overconfidence - Anderson, C., Brion, S., Moore, D.A. & Kennedy, J.A., Journal of Personality and Social Psychology, 103(4), 718-735 (2012). [Reliability: High] ↩︎ ↩︎2

  3. Illusion and Well-Being: A Social Psychological Perspective on Mental Health - Taylor, S.E. & Brown, J.D., Psychological Bulletin, 103(2), 193-210 (1988). [Reliability: High] ↩︎

  4. The Social Brain Hypothesis - Dunbar, R.I.M., Evolutionary Anthropology, 6(5), 178-190 (1998). [Reliability: High] ↩︎

  5. Grooming, Gossip, and the Evolution of Language - Dunbar, R.I.M., Harvard University Press (1996). [Reliability: High] ↩︎

  6. Reputation, Gossip, and Human Cooperation - Wu, J., Balliet, D. & Van Lange, P.A.M., Social and Personality Psychology Compass, 10(6), 350-364 (2016). [Reliability: High] ↩︎ ↩︎2

  7. The Impact of Gossip, Reputation, and Context on Resource Transfers - Research on reputation transmission across multiple cultures including hunter-gatherers, Evolution and Human Behavior (2023). [Reliability: High] ↩︎

  8. The Echo Chamber Effect on Social Media - Cinelli, M. et al., PNAS, 118(9), e2023301118 (2021). [Reliability: High] ↩︎

  9. Why Do So Many Incompetent Men Become Leaders? - Chamorro-Premuzic, T., Harvard Business Review (2013). [Reliability: Medium-High] ↩︎

  10. Hubris Syndrome: An Acquired Personality Disorder? - Owen, D. & Davidson, J., Brain, 132(5), 1396-1406 (2009). [Reliability: High] ↩︎ ↩︎2

  11. The 14 Symptoms in Full - Daedalus Trust (14 diagnostic criteria for Hubris Syndrome). [Reliability: Medium-High] ↩︎

  12. Overconfidence Effect - Wikipedia (overview article integrating multiple peer-reviewed studies). [Reliability: Medium-High] ↩︎

  13. Planning Fallacy - The Decision Lab (explainer based on Kahneman & Tversky, 1979). [Reliability: Medium-High] ↩︎

  14. Performing a Project Premortem - Klein, G., Harvard Business Review (2007). [Reliability: Medium-High] ↩︎

  15. Cognitive and Interpersonal Features of Intellectual Humility - Leary, M.R. et al., Personality and Social Psychology Bulletin, 43(6), 793-813 (2017). [Reliability: High] ↩︎

  16. Augmenting Means Efficacy to Boost Performance: Two Field Experiments - Eden, D. et al., Journal of Management, 36(3), 687-713 (2010). [Reliability: High] ↩︎

Psychology
overconfidence evolutionary psychology cognitive bias evolutionary mismatch hubris syndrome
This post is licensed under CC BY 4.0 by the author.