Post
JA EN

When 'Not Thinking' Becomes Rational — The Intersection of Individual and Organizational Apathy

Posted
By
21 min read
When 'Not Thinking' Becomes Rational — The Intersection of Individual and Organizational Apathy
  • Target audience: Software engineers, engineering managers, and anyone interested in organizational learning
  • Prerequisites: None
  • Reading time: Approx. 15 minutes

Overview

“Fools learn from experience; wise men learn from history” — a saying attributed to Bismarck. But what do we call those who stop learning altogether?

In Build It, Break It, Learn It, I explored how individuals can accelerate the “build, break, and reflect” learning cycle with AI. In Is “They Might Leave After We Train Them” the Right Question?, I examined the case for organizations investing in safe-to-fail environments. Both articles ended on an optimistic note. Individuals can learn. Organizations can invest.

But in reality, there exists a structure where both not learning and not investing appear rational. And when these two dynamics occur simultaneously, the problem doesn’t merely add up — it multiplies.

Why think for yourself when AI gives you answers? — In the short term, that makes perfect sense. Why invest in people who might leave? — There’s logic to that, too. The problem is that when both are simultaneously “justified,” everyone’s optimal behavior leads to an equilibrium where everyone loses. In game theory, this is called a Nash equilibrium: everyone is playing their best strategy, yet the outcome is suboptimal for all.

This article structurally analyzes why the “rationality of not thinking” and the “rationality of not investing” emerge, and examines what happens when they intersect — backed by evidence. There is no simple prescription for breaking this equilibrium. But recognizing the warning signs before it solidifies, and identifying the options available to individuals already trapped inside it, is possible.

Individual Apathy — The Rationality of Not Thinking

AI Has Driven the Cost of Not Thinking to Near Zero

It used to take time to find answers. You would read manuals, dig through documentation, and arrive at understanding through trial and error. The process itself was learning.

AI short-circuits this process. Ask a question, get an answer. Request code, receive code. The short-term cost of not thinking has never been lower.

This isn’t laziness. It’s a perfectly reasonable judgment. If all you need is to get today’s work done, asking AI is faster. You can spend the time you would have used thinking on other tasks. Tokens are cheap; your time is expensive.

But Learning Disappears

The problem lies in the long-term consequences of this reasonable judgment.

Kapur’s “Productive Failure” research, a meta-analysis of over 12,000 participants across 166 experiments, showed that learners who struggled and failed on their own before receiving instruction performed significantly better on conceptual understanding (effect size g = 0.36; g = 0.58 when design principles were closely followed)1. Crucially, this effect only occurs when learners go through the “generation phase” — the stage where they think and fail on their own. When you ask AI for the answer, this entire phase is skipped.

Anthropic’s randomized controlled trial found that developers who learned a coding library with AI assistance scored 17% lower on comprehension tests (AI group: 50% vs. manual group: 67%)2. Moreover, the time savings from AI use were not statistically significant. Learning was lost, but efficiency didn’t improve.

Gerlich’s (2025) study of 666 participants found a correlation of r = +0.72 between AI tool usage frequency and cognitive offloading (the tendency to delegate thinking to external tools), and r = -0.75 between cognitive offloading and critical thinking ability3. The more you use AI, the more you outsource thinking; the more you outsource thinking, the more your critical thinking declines.

This phenomenon predates the AI era. Sparrow et al.’s (2011) “Google Effect,” published in Science, demonstrated that people are less likely to remember information they believe they can find online4. We had already shifted from remembering content to remembering where to find it. AI is merely accelerating this trend.

The Structure of Individual Rationality

flowchart TB
    A["AI provides<br>instant answers"] --> B["Today's problems<br>solved without thinking"]
    B --> C["Short-term cost of<br>not thinking ≈ 0"]
    C --> D["Not thinking<br>appears rational"]
    D --> E["Judgment doesn't<br>develop (long-term cost)"]
    E --> F["Dependence on AI<br>deepens further"]
    F --> A

The insidious aspect of this loop is that the long-term cost is invisible in the moment. Declining judgment doesn’t announce itself suddenly — it erodes gradually. And as long as AI compensates, the problem never surfaces.

Organizational Apathy — The Rationality of Not Investing

They’ll Leave After Training, So Why Bother

As discussed in detail in Is “They Might Leave After We Train Them” the Right Question?, the structural “rationality” behind organizations not investing in talent development is real.

Edmondson’s research has demonstrated that psychological safety promotes learning and that learning improves team performance5. However, no RCT (randomized controlled trial) has proven a direct causal link from psychological safety to corporate profit. When executives say “there’s no proof, so we can’t invest,” they’re not being dishonest. That’s simply the structure of the available evidence.

In the short term, cost-cutting produces visible numbers. Stakeholders can be satisfied with the mere appearance of action. Going all-in risks destabilizing existing power structures. “Not investing” is, for executives too, a rational short-term choice.

What the Numbers Show

Gallup’s 2026 report shows global engagement has fallen to 20%, the lowest since 20206. Japan’s situation is even more dire: engagement is just 6%. Actively disengaged employees stand at 24% — four times the engaged population. The estimated annual opportunity cost is $10 trillion worldwide (roughly 9% of GDP), and 86 trillion yen in Japan alone6.

Even more concerning: manager engagement has declined more sharply than that of individual contributors. Gallup’s 2025 report recorded a drop from 30% to 27%7, with further decline continuing into 2026. Managers under 35 dropped 5 points; women managers dropped 7 points7. A separate Gallup study found that managers account for 70% of the variance in team engagement. The very people who should be driving change are the first to disengage.

The Structure of Organizational Rationality

flowchart TB
    A["Cannot prove<br>causation"] --> B["Investment approval<br>denied"]
    B --> C["No learning<br>environment provided"]
    C --> D["Employee skills<br>stagnate"]
    D --> E["Engagement<br>drops further"]
    E --> F["'See? Investment<br>was pointless'"]
    F --> A

This too is a self-reinforcing loop. No investment means no results; no results reinforces the conviction that “investing is pointless.”

When They Intersect — The Multiplicative Effect

So far, we’ve examined individual and organizational apathy separately. But the real problem emerges when both occur simultaneously.

Organizational Knowledge as a Public Good

The “public goods game” from game theory describes this structure precisely. Each player decides whether to contribute tokens to a shared pool. The pool is multiplied and distributed to everyone, but each individual is better off not contributing8.

Organizational knowledge is also a public good. When individuals invest time in learning and share their insights, the entire organization benefits. But for each individual, free-riding on colleagues’ knowledge (or AI output) is cheaper than learning independently.

The Nash equilibrium of the public goods game is zero contribution from everyone — universal free-riding8. In real experiments, initial cooperation rates of around 50% are observed, but as participants witness free-riding, conditional cooperators also stop contributing, and investment approaches zero over successive rounds8.

Olson’s (1965) theory of collective action makes this even clearer: as groups grow larger, per-capita marginal benefit shrinks and the incentive to free-ride increases9. Knowledge sharing within teams benefits the whole, but for each individual, “not investing” tends to be more rational — especially when AI can partially compensate for individual productivity, making it possible to get by without depending on others’ knowledge.

Conformity Pressure Locks the Equilibrium

Asch’s (1951) classic conformity experiment found that 75% of subjects conformed at least once to an obviously incorrect majority10. The average conformity rate was approximately 32%. Crucially, a single dissenter was enough to cause conformity to plummet from 32% to 5%10.

Apply this to organizations. As Gallup’s data showed, in an environment where disengaged employees outnumber engaged ones four to one, showing initiative itself becomes a deviation from the majority.

Noelle-Neumann’s (1974) “Spiral of Silence” theory explains this dynamic11. When minorities perceive their views differ from the majority, they fall silent. Silence is taken as agreement, making the majority even louder. The minority silences itself further. Morrison & Milliken (2000) examined this in organizational contexts and showed that organizations are generally intolerant of dissent, and employees are reluctant to speak up about problems12.

Becoming the “Anomaly”

Within this structure, individuals who ask “why” and employees who show initiative both become “anomalies.”

 Individual ApathyOrganizational Apathy
Why it seems rationalAI gives answers — why think?They’ll leave — why invest?
Short-term resultToday’s problem is solvedThis quarter’s costs are cut
Long-term costJudgment doesn’t developOrganizational learning capacity vanishes
Free-riding onAI outputEmployees’ personal motivation
Who becomes the “anomaly”Those who ask “why” → “inefficient”Motivated employees → “stop rocking the boat”

Why verify AI output when you can just use it? Why suggest improvements when everyone else follows orders quietly? Both thinkers and motivated employees look like “people wasting effort on unnecessary things.”

As Asch’s experiment demonstrated, this structure has no self-correcting mechanism. A dissenter — a thinker, a motivated employee — dramatically reduces conformity. But once they’re pushed out, whether by leaving or giving up, conformity operates at full force.

The Rational Apathy Equilibrium

flowchart TB
    A["Organization provides<br>no learning environment"] --> B["Individuals have no<br>opportunity to learn at work"]
    B --> C["Individuals outsource<br>thinking to AI"]
    C --> D["Organization's actual<br>knowledge hollows out"]
    D --> E["Motivated people<br>lose their place"]
    E --> F["Motivated people<br>leave or give up"]
    F --> G["What remains: non-thinking<br>individuals + non-investing org"]
    G --> A

This equilibrium is stable. Everyone is acting rationally. No one complains — because those who would complain are already gone. Metrics look fine — because AI compensates for short-term productivity. And no one inside the system has an incentive to break it.

Economists call this a Nash equilibrium. It’s the same structure as the prisoner’s dilemma: each player is choosing their best strategy, yet the collective outcome is suboptimal13. Everyone would benefit from cooperation, but no individual has an incentive to switch to cooperation unilaterally.

The Augmentation Trap — Why Managers Accelerate the Equilibrium

The analysis so far explains why the equilibrium is stable. But why does it accelerate? Caosun & Aral’s (2026) model, “The Augmentation Trap,” reveals the mechanism14.

Their dynamic economic model shows that adopting AI is rational even with full knowledge of skill degradation. Immediate productivity gains outweigh discounted future skill-loss costs. The critical finding: when managers have shorter planning horizons, organizations deploy AI beyond the optimal level14.

Overlay this with Gallup’s data on plummeting manager engagement7, and the picture becomes clear. Disengaged managers chase short-term numbers. Managers chasing short-term numbers deploy AI for immediate productivity rather than skill development. AI deployed for immediate productivity accelerates skill degradation. Skill-degraded organizations make managers’ jobs harder, further reducing their engagement.

The trap is that everyone inside it is acting rationally.

Before It Happens — Warning Signs and Prevention

Once this equilibrium solidifies, breaking it from the inside is extremely difficult. The very definition of Nash equilibrium is “no single player has an incentive to change strategy alone.” Telling individuals to “think more” doesn’t change the structure that makes not-thinking more rewarding. Telling executives to “invest more” doesn’t conjure an RCT that doesn’t exist.

That’s why acting before the equilibrium solidifies matters.

Recognizing the Warning Signs

Equilibria don’t form overnight — they solidify gradually. When these signs start accumulating, the spiral may already be underway:

  • “Just ask AI” has become the default reflex. Fewer people ask “why does this work this way?” and “if it works, it’s good enough” has become the unspoken standard
  • Improvement suggestions are treated as “unnecessary.” The content of a proposal matters less than the fact that someone made one at all
  • Managers prioritize status quo above all. Challenges aren’t welcomed; not making waves is what gets rewarded
  • Those who learn are becoming isolated. Study groups draw no attendees. Knowledge sharing has become perfunctory

As the public goods game experiments showed, cooperation doesn’t start at zero. But once conditional cooperators begin observing free-riders and withdrawing one by one, the decline accelerates8. By the time the slope is visible, it’s already late.

Designing Against Equilibrium Lock-In

The following is addressed to team leaders, managers, and others in positions to design systems. Preventing the equilibrium from solidifying is far more realistic than trying to break it after.

Embed incentives for thinking into team practices. Anthropic’s research identified “high-scoring patterns” — requesting explanations after code generation, asking only conceptual questions while coding independently — that preserved comprehension even with AI assistance2. Institutionalizing these as team practices rather than leaving them to individual effort structurally counteracts the slope toward “not thinking is easier.”

Build protections for “anomalies” before they’re needed. As noted earlier, Asch’s experiments showed that a single dissenter was enough to break conformity10. Conversely, the moment that person disappears, conformity locks in. Devil’s advocate roles, anonymous feedback mechanisms — these are far less effective when introduced after the equilibrium has already shifted. Build structural protections for dissenting voices while the equilibrium is still healthy.

Consciously extend managers’ planning horizons. Caosun & Aral’s model shows that the “augmentation trap” worsens as managers’ planning horizons shorten14. Managers chained to quarterly numbers will deploy AI for immediate productivity, not skill development. Deliberately designing longer evaluation cycles serves as a prophylactic against the short-termism spiral.

If You’re Already Inside the Equilibrium

Some readers may be thinking, “It’s already too late for my organization.” Honestly, breaking a solidified equilibrium through individual effort is extraordinarily difficult.

Hirschman (1970) categorized individuals’ responses to organizational decline as “Exit,” “Voice,” and “Loyalty”15. In an organization where the spiral of silence is active, Voice is costly and its returns are uncertain. Loyalty amounts to maintaining the equilibrium.

That Exit becomes the rational remaining option is not the individual’s failure — it’s a structural consequence. Engaged people leave first because they have options. And the organization left behind only reinforces the equilibrium further. This too is part of the spiral that the structure produces.

Of course, leaving isn’t always the best option. Another choice is to accept being the “anomaly.”

Among the 25% who never conformed in Asch’s experiments, many reported “internal conflict but trusting their own perceptions”10. The structural cost is high — risks of being seen as “inefficient” or “not reading the room.” But this choice preserves something the structure cannot take away: the judgment that comes from continuing to think for yourself. In an age when AI provides answers, the judgment of someone who kept asking “why” will be most valuable when the equilibrium eventually breaks — and it will.

And as noted earlier, a single dissenter is enough for conformity to collapse dramatically. Being the anomaly doesn’t just protect your own judgment — it structurally weakens the conformity pressure around you. Even when it doesn’t feel that way, whether a team has one “thinking person” or none changes the equilibrium’s strength entirely.

A third realistic option is the “build and break outside of work” cycle described in Build It, Break It, Learn It. Maintaining your own skills independently of the organization’s learning environment. However, this is compensating for an organizational problem with individual effort — not a structural solution. Following this article’s analysis, such individual effort may itself become part of the structure in which the organization free-rides on employees’ personal motivation.

Exit, staying the anomaly, learning outside work — none of these are prescriptions for changing the structure. But if you can see the structure, you at least know that it’s not your fault. And whatever option you choose, a decision made with structural understanding is invariably better than one made without it.

Conclusion — Seeing the Structure Itself Has Value

The question this article has repeatedly asked is not who is to blame, but why everyone’s behavior appears rational.

Individuals don’t stop thinking because they’re lazy. They stop because the short-term return on thinking is invisible in an environment where AI provides answers. Organizations don’t stop investing because they’re irresponsible. They stop because no RCT proves causation and the short-term numbers look better without investment. Motivated people aren’t pushed out through malice. They’re pushed out because conforming to the majority is socially safe.

Everyone is acting rationally. Yet everyone is worse off.

This article doesn’t hold a prescription for breaking that equilibrium. By definition, a Nash equilibrium structurally resists internal prescriptions. However, recognizing this structure has value in itself.

“Why is my organization like this?” “Why do I feel like I don’t fit in?” — To these questions, having a structural explanation — “this is an equilibrium resulting from everyone acting rationally” — rather than “you’re not trying hard enough” makes a difference. It gives you the perspective to see structure instead of blaming yourself. And once you can see the structure, you can spot the warning signs early enough to act before the equilibrium solidifies — or calmly assess your options within one that already has.

Build It, Break It, Learn It examined the conditions for individual learning, and Is “They Might Leave After We Train Them” the Right Question? examined the case for organizational investment. This article analyzes what happens when both fail — when individuals and organizations alike rationally choose apathy.

References

References are listed in the order they are cited in the text.

  1. When Problem Solving Followed by Instruction Works: Evidence for Productive Failure - Sinha, T. & Kapur, M., Review of Educational Research, Vol.91, No.5, pp.761–798 (2021). DOI: 10.3102/00346543211019105. Peer-reviewed. Meta-analysis of 12,000+ participants across 166 experiments. Productive failure group showed significant superiority in conceptual understanding and transfer (g = 0.36; g = 0.58 with high design-principle fidelity). [Reliability: High] ↩︎

  2. How AI Impacts Skill Formation - Shen, J. H. & Tamkin, A., Anthropic Research (2026). RCT with 52 software developers. AI-assisted group scored 17% lower on comprehension (AI: 50% vs. manual: 67%). Identified 6 AI usage patterns; cognitively engaged patterns preserved learning. Coverage: InfoQ (2026). [Reliability: Medium-High] (RCT but small sample) ↩︎ ↩︎2

  3. AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking - Gerlich, M., Societies, Vol.15, No.1, p.6 (2025). Peer-reviewed. n = 666. AI tool usage and cognitive offloading: r = +0.72; cognitive offloading and critical thinking: r = -0.75. Younger participants (17-25) showed higher AI dependence and lower critical thinking scores. [Reliability: Medium-High] ↩︎

  4. Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips - Sparrow, B., Liu, J. & Wegner, D. M., Science, Vol.333, No.6043, pp.776-778 (2011). DOI: 10.1126/science.1207745. Peer-reviewed. Four experiments demonstrated reduced recall for information believed to be available online, with enhanced recall for where to find it. [Reliability: High] ↩︎

  5. Psychological Safety and Learning Behavior in Work Teams - Edmondson, A., Administrative Science Quarterly (1999). DOI: 10.2307/2666999. Peer-reviewed. 51 manufacturing teams. Demonstrated the mediation model: psychological safety → learning behavior → team performance. No RCT exists proving direct causation from psychological safety to corporate profit. [Reliability: High] ↩︎

  6. State of the Global Workplace 2026 - Gallup (2026). Global engagement at 20% (2025 measurement; lowest since 2020). Productivity loss: $10 trillion annually. Japan: 6% engaged, 24% actively disengaged, 86 trillion yen in opportunity cost (2023). First-ever two consecutive years of decline. [Reliability: High] (Robust survey methodology; GDP-equivalent costs are estimates) ↩︎ ↩︎2

  7. Synthesis of multiple Gallup reports on manager engagement. (1) Global Employee Engagement Falls for Second Time - Gallup (2025). Manager engagement dropped from 30% to 27% (2024 measurement). Under-35 managers: 5-point drop; women managers: 7-point drop. (2) Global Employee Engagement Continues Decline - Gallup (2026). Continued decline in manager engagement. (3) The finding that managers account for 70% of variance in team engagement is from Gallup Business Journal. [Reliability: High] ↩︎ ↩︎2 ↩︎3

  8. Synthesis of multiple sources on public goods games. Nash equilibrium is zero contribution, but experiments observe ~50% initial contribution declining over rounds. Punishment mechanisms significantly improve cooperation rates. See: Public goods game - Wikipedia; The dynamics of human behavior in the public goods game with institutional incentives - Scientific Reports (2016). [Reliability: High] (Standard game theory findings) ↩︎ ↩︎2 ↩︎3 ↩︎4

  9. The Logic of Collective Action: Public Goods and the Theory of Groups - Olson, M., Harvard University Press (1965). Classic work on collective action. Rational individuals will not voluntarily contribute to non-excludable public goods. Free-riding incentive increases with group size. Selective incentives (rewards/punishments available only to participants) as the solution. [Reliability: High] ↩︎

  10. Asch, S. E. (1951). Effects of group pressure upon the modification and distortion of judgment. In H. Guetzkow (ed.) Groups, leadership and men. Pittsburgh, PA: Carnegie Press; Asch, S. E. (1956). Studies of independence and conformity: I. A minority of one against a unanimous majority. Psychological Monographs, 70(9), 1-70. 75% conformed at least once. Average conformity rate: 32%. Single dissenter: 32% → 5%. Bond & Smith (1996) meta-analysis confirmed higher conformity in collectivist cultures. See: Simply Psychology. [Reliability: High] ↩︎ ↩︎2 ↩︎3 ↩︎4

  11. Noelle-Neumann, E. (1974). The Spiral of Silence: A Theory of Public Opinion — Our Social Skin. Journal of Communication, 24(2), 43-51. Minorities who perceive their views differ from the majority fall silent; majority becomes louder in a self-reinforcing cycle. [Reliability: High] ↩︎

  12. Morrison, E. W. & Milliken, F. J. (2000). Organizational Silence: A Barrier to Change and Development in a Pluralistic World. Academy of Management Review, 25(4), 706-725. Peer-reviewed. Theoretically analyzed how organizations are generally intolerant of dissent and employees are reluctant to voice concerns. Coined the “organizational silence” concept. [Reliability: High] ↩︎

  13. Standard definition of the prisoner’s dilemma. Mutual defection is the only strong Nash equilibrium and is not Pareto efficient. In organizational contexts: The prisoner’s dilemma in the workplace — showed that unconditionally cooperative managers significantly improve organizational performance and reduce stress. [Reliability: High] (Standard game theory findings) ↩︎

  14. The Augmentation Trap: AI Productivity and the Cost of Cognitive Offloading - Caosun, M. & Aral, S. (2026). Dynamic economic model. Even with knowledge of skill degradation, rational decision-makers adopt AI because immediate productivity gains outweigh discounted future costs. When managers have shorter planning horizons than workers, organizations deploy AI beyond optimal levels (“the augmentation trap”). Preprint. [Reliability: Medium-High] (Theoretical model; no field validation yet) ↩︎ ↩︎2 ↩︎3

  15. Hirschman, A. O. (1970). Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States. Harvard University Press. Classic work categorizing individual responses to organizational decline as “Exit,” “Voice,” and “Loyalty.” Theoretically demonstrated that Exit becomes the rational choice when Voice is costly and its returns are unclear. [Reliability: High] ↩︎

Career
AI organizations learning game-theory psychology engagement
This post is licensed under CC BY 4.0 by the author.