The Essence of
Scientific Civilization

I The Epistemological Position of Science: Institutionalized System 2 과학의 인식론적 위치: 제도화된 시스템 2

What is science? Within the framework of Cognitive Architecture Theory, science has a precise positioning: it is the institutionalized form of System 2’s abstraction function. System 2 executes classification, ordering, analogy, distillation, alignment, and internalization operations on information — scientific methodology standardizes these operations into the procedure of hypothesis-experiment-verification, and anchors them through academic institutions, journal systems, and peer review systems so they can operate across individuals, across time, and across space.

This paper’s critique must unfold across three levels, which are not the same object:

This institutionalization has produced enormous achievements in human history. But it has also inherited a fundamental tendency of System 2: System 2’s instinct is to find patterns, build deterministic models, and eliminate random noise. When this tendency is institutionalized as the sole admission criterion for scientific methodology, it becomes an epistemological constraint — reproducibility becomes the entry threshold for scientific knowledge.

The more precisely science controls, the less it sees the real world, and the more it sees the artificial world created by the act of control. This is not to say that science is “wrong,” but that science is a tool with a specific scope of applicability — it can precisely describe reproducible regularities in the quantitative world, but when facing fundamentally irreproducible quantum randomness, its methodological axioms constitute an insurmountable epistemological ceiling.

II The Quantitative World versus the Variable World 정량적 세계 대 변수적 세계

This paper proposes a core distinction: the physical world has two faces — the quantitative world and the variable world. The quantitative world is the stable low-dimensional projection obtained after the physical world is controlled, measured, and modeled. The variable world is closer to the actual operational state of open systems — a high-dimensional dynamic system jointly constituted by multi-variable coupling, historical dependence, nonlinear feedback, and random perturbation.

Science excels at handling the quantitative world. When a scientist designs an experiment, the first thing they do is “control variables” — freezing most variables, allowing only one or two to change, then observing their relationship. This operation is methodologically legitimate, but ontologically it is an intentional simplification: it projects a high-dimensional dynamic system into a low-dimensional static cross-section.

A qualification is needed: science is not entirely incapable of handling the variable world. Complex systems science, statistical physics, chaos theory, non-equilibrium thermodynamics, evolutionary biology, and stochastic process theory are precisely the tools through which science attempts to reach the variable world. The problem is not that these tools don’t exist, but their status within the research institution — their methodology (simulation rather than precise prediction, probability rather than certainty, emergence rather than reduction) exists in persistent tension with the mainstream reproducibility paradigm, and occupies a relatively weak position in funding structures, publication incentives, and disciplinary power. Scientism — treating precisely controllable experiments as the sole source of “rigorous” knowledge — is the true target of this paper’s critique.

The significance of this distinction is: science’s success derives precisely from its limitations. Because it only handles the reproducible portion, it can produce reliable technical applications — bridges don’t randomly collapse, drug dosages can be standardized, airplane engines work the same way every time. But this success also creates an epistemological illusion: people begin to believe that the quantitative world is the entire world, that reproducible regularities are the entire truth. Irreproducible creative generation — insight, cross-domain analogy, unpredictable conceptual recombination (which the companion paper in this series, “Dark Channels and the Intelligence Evaluation Formula,” terms the “dark channel,” used here as metaphor without presupposing its physical mechanism) — is excluded from the definition of “knowledge,” not because it doesn’t exist, but because science’s net cannot catch it.

A further qualification: the “variable world” in this paper is not a mystical domain, but refers to the high-dimensional state space of open systems that cannot be fully captured by a single controlled experiment, including multi-variable coupling, path dependence, feedback loops, random perturbation, and historical contingency. It can be partially modeled (complex systems science is doing this), but is difficult to fully reduce to a low-dimensional quantitative cross-section. The variable world is not the opposite of the quantitative world, but the larger space in which the quantitative world is embedded.

III The Structural Degradation of the Research System 과학연구 체계의 구조적 퇴화

If scientific methodology’s limitations are the ceiling, then the institutional arrangements of the modern research system further lower the space beneath that ceiling. This section uses data to reveal the scale of this degradation.

The Extinction of Disruptiveness

A sixty-year analysis of 45 million papers and 3.9 million patents shows that scientific and technological disruptiveness has been in continuous decline (Park, Leahey & Funk, Nature 2023). Paper disruptiveness scores from 1945 to 2010 dropped by 91.9% in the social sciences and 100% in the physical sciences. A methodological caveat is necessary: some researchers question whether the CD disruptiveness index may be subject to systematic bias from citation inflation (Petersen et al., 2023), and some analyses find that the absolute number of highly disruptive patents may have actually increased since 2008. Therefore, the robustness of the “proportional decline” trend in disruptiveness remains under debate. Meanwhile, paper output has grown explosively — the annual growth rate of scientific papers is approximately 8–9%, rising from 1.3 million papers in 2000 to 4.6 million in 2020. A large number of papers have extremely low citation impact, suggesting a possible decoupling between output growth and knowledge impact.

The tension between exponential growth in paper output and the trend decline in disruptiveness — even as the latter’s measurement methodology remains disputed — suggests that the research institution may be systematically biased toward predictable incremental work.

Misaligned Incentive Systems

Research institutions globally tie promotion and bonuses closely to publication statistics, rather than to those publications’ impact on real human lives. This incentive system handsomely rewards observable research output — publications and citations — without adequately rewarding hard-to-measure exploratory research. Researchers are incentivized by the dominant scientific evaluation paradigm to pursue high output and incremental, low-disruptiveness work.

Disciplinary Silos: The Institutionalization of Deep-Well Intelligence

Self-reinforcing disciplinary silos and academic publishing norms prevent new insights from reaching potential collaborators beyond disciplinary boundaries. Peer review processes ossify disciplinary norms. Pre-tenure researchers are discouraged from pursuing interdisciplinary and “applied” research. By the time researchers obtain tenure, they are usually past 37, often in their early 40s, and the peak of creativity has already passed.

In 1980, over 18% of NIH principal investigators were under 35; by 2014 this proportion had plummeted to just 2%. The funding mechanism has transformed from catalyst to bottleneck. A typical modern scientist repeatedly runs the same abstraction program on an extremely narrow channel, with inputs from existing literature within their own discipline and outputs that are fine-tunings of existing conclusions. Their System 2 is indeed working, but only on an extremely narrow frequency band. All other channels are closed.

The Systematic Marginalization of Basic Research

Basic research — the scientific activity closest to unpredictable creative generation — is being systematically marginalized. Between 2012 and 2023, the federal government’s share of basic research funding fell from 52% to 41%. In fiscal year 2023, only 25% of federal R&D appropriations went to basic research, 29% to applied research, and 46% to experimental development. A distinction must be drawn: the total U.S. R&D pie continues to expand (estimated at $993 billion in 2024), and absolute investment in basic research is also growing (estimated at $145 billion in 2024). What is discussed here is the continuous decline of basic research’s relative position within the R&D system — growth is concentrated in experimental development (estimated at $668 billion in 2024), and the priority of exploratory research in resource allocation is continuously declining.

The current NIH director has described the agency’s reduced support for “edge science.” Projects must pre-define expected outcomes, funding must demonstrate predictable returns, and researchers must state in their applications what they will discover — this amounts to requiring that one know the outcome of creative exploration before it begins. This is a logical paradox and an institutional self-limitation.

IV The Self-Referential Loop of the Research System 과학연구 체계의 자기 순환 폐쇄 루프

The modern research system has internally formed a perfect self-verification loop:

  Publish paper → Peer review (conducted by people within the same silo)
      ↓
  Receive citations → Citations from other researchers in the same silo
      ↓
  Citation count becomes performance metric → Metric determines funding
      ↓
  Funding determines who can remain in the system
      ↓
  Those remaining continue publishing papers → Cycle continues
      ↓
  No stage of the entire cycle requires contact with external reality
  

In certain highly metric-driven academic fields, knowledge production can self-circulate with weak external reality feedback — publishing papers, peer review, receiving citations, citation counts becoming performance metrics, performance metrics determining funding allocation, funding determining who can remain in the system. In application-oriented fields such as engineering, medicine, and climate science, external reality feedback remains powerful and the loop is less closed. But in certain foundational disciplines and highly specialized theoretical fields, the risk of this self-circulation is real.

This is the precise mechanism of “self-gratification mode.” The system is not failing to operate — it operates with extreme efficiency — but the goal of its operation has been silently swapped from “understanding the world” to “maintaining the system’s own existence.” This is a classic organizational degradation pattern: means become ends, tools become products, process becomes result.

V Science Aligns with Survival, Not Development 과학은 생존에 정렬되지, 발전에 정렬되지 않는다

The outputs of science — technology, medicine, engineering systems — are all cases of forcibly compressing a local portion of the variable world into a quantitative system, making it predictable, reproducible, and mass-replicable. What need does this satisfy? Within the Cognitive Architecture Theory framework, the answer is clear: it satisfies the needs of the enteric-brain system and the endocrine system — eating, safety, stability, comfort. These are survival-layer needs, not development-layer needs.

Counterexamples and Expansion Effects: Science Also Creates Variable Space

It must be acknowledged that science and technology not only compress variable space, but have also opened entirely new variable spaces on multiple occasions. The microscope revealed the microbial world to humanity; the telescope revealed cosmic scales; particle accelerators allowed detection of subatomic structures; the internet enabled global collaboration networks; gene sequencing allowed reading the code of life. Each breakthrough in observational tools was a dramatic expansion of variable space — new channels opened, new information dimensions became accessible.

Standardization is sometimes not the compression of creation, but the lowering of creation’s threshold. Universal mathematical language enabled more people to enter the scientific creation system; open-source tools enabled more people to participate in technological innovation; preprint platforms enabled knowledge circulation to break through journal silos. These are all instances of scientific civilization expanding variable space.

Therefore, the more accurate judgment is not “scientific civilization only compresses variable space,” but: scientific civilization simultaneously possesses both expansive and compressive effects. In its early and breakthrough stages, the expansion effect dominates — new tools open new worlds. In its mature and institutionalized stages, the compression effect gradually strengthens — standardization, efficiency optimization, and the demand for predictability begin to close the exploration space that had been opened. The current research system’s problem is not a defect of the scientific method itself, but that the institutional compression effect is overtaking the expansion effect.

And what does “development” mean? It means opening new cognitive channels, establishing cross-domain connections, and obtaining knowledge never before systematized through unpredictable exploration. This precisely demands tolerance for unpredictability, encouragement of non-standard paths, and institutional space for high-failure-rate exploration. The scientific method itself does not exclude these — complex systems science and evolutionary biology are attempting them. But the institutional inertia formed after the scientific method’s success may bias toward reinforcing predictable, publishable, manageable paths, thereby reducing institutional tolerance for unpredictable exploration.

GPS relies on Einstein’s general relativity and 19th-century geometric thought — the latter was considered utterly useless by contemporaries. Teflon, saccharin, and the cardiac pacemaker were all products of accidental creation. These “accidents” are not accidents within this framework — they are products of unpredictable creative generation, occurring in window periods before the institutionalized demand for predictability had come to dominate the research ecosystem. The contemporary research institution’s incentive structure is systematically compressing the institutional space in which “accidents” can occur.

VI Industrialization: The End of Natural Selection 산업화: 자연선택의 종결

The biological consequences of scientific civilization began to manifest with the Industrial Revolution. A precise formulation is needed: industrialization did not “shut off” natural selection — natural selection never stopped; it merely changed direction. Before industrialization, selection pressures came primarily from famine, infectious disease, predators, and climate extremes — these forces selected for physical endurance, immunity, and stress response. After industrialization, these selection pressures weakened dramatically, but new selection pressures emerged: the ability to adapt to high-sugar diets, sedentary environments, information overload, microplastics, and endocrine disruptors became new adaptive variables. Only eight to nine generations have passed since the Industrial Revolution, and only three since the post-WWII acceleration of industrialization. Researchers describe 21st-century humans as an “indoor-urban species” — in the US, UK, and Canada, people spend 93% of their time indoors.

Therefore, the more accurate judgment is not “natural selection has stopped,” but “the specific forms of selection pressure directly related to cognitive ability and self-control (famine mortality, survival competition intensity) have dramatically weakened, while the screening intensity and direction of new selection pressures (educational attainment, social networks, health behaviors) remain unclear.” Population genetics evidence for this inference is currently insufficient and should be treated as a theoretical hypothesis.

VII Specialized Education: The Designed Closure of Channels 전문화 교육: 채널의 설계적 폐쇄

The channel closure brought by the human division of labor and the career path fixation brought by educational specialization are the core mechanisms among postnatal factors that make individuals increasingly unable to exercise self-control and lose the possibility of thinking.

Specialization locks a person’s cognitive channels onto an extremely narrow frequency band. An assembly line worker who spends an entire lifetime using only System 1 to process repetitive motions may see their System 2 abstract processing capability atrophy due to prolonged disuse — neuroplasticity research supports the “use it or lose it” principle (long-term unused synaptic connections weaken), but this paper currently lacks direct neuroimaging data (such as longitudinal cortical grey matter density changes in specific populations) to precisely quantify this effect.

It must be acknowledged that specialization also has an expansion effect: deep accumulation enables individuals to achieve extremely high technical precision and complex collaboration capability in specific domains. Many of the higher-order achievements of modern civilization — from chip manufacturing to gene editing — are inseparable from extremely deep specialization. The problem lies not in specialization itself, but in premature, excessively narrow, irreversible specialization, combined with the absence of cross-domain circuits and the evaluation system’s systematic punishment of exploratory behavior, which jointly closes the individual’s multi-frequency channels. The American labor market is progressively removing requirements for general education degrees, demanding only specialized knowledge. Most students do not enjoy learning, and even with extensive incentives have only narrow interest ranges.

VIII The Three-Stage Convergence Engine: Internet, Mobile Internet, AI 3단계 수렴 엔진: 인터넷, 모바일 인터넷, AI

After the internet became humanity’s information channel, behavioral homogenization accelerated severely. It accelerated again after mobile internet, and again after AI integration. These three stages constitute an accelerating convergence process.

Stage One: The Internet

Information filter bubbles are not manufactured by algorithms alone, but emerge through the recursive interaction of motivated cognitive processing, identity-based social network structures, and algorithmic amplification of behavioral and emotional cues. The internet gave humanity its first shared information pool. On the surface, information volume exploded; in reality, what each person sees is increasingly similar. Algorithms reinforce convergence tendencies by amplifying prior behavior.

Stage Two: Mobile Internet

Smartphones bound the information homogenization system to the human body, 24 hours a day, never leaving the hand. It is no longer something people actively connect to, but something that continuously pushes content to people. Users in homogenized digital networks encounter inconsistent viewpoints at a probability 64% lower than a decade ago. Simultaneously, at the physical level, it continuously generates endocrine noise — blue light disrupts melatonin, dopamine loops hijack attention, short-form video trains the brain to accept only information packets that have already been reduced to the extreme.

Stage Three: AI

AI’s cognitive impact is more fundamental than the previous two stages, and it is bidirectional. On one hand, AI is beginning to replace portions of System 2’s abstraction operations — classification, ordering, summarization, draft generation. When humans chronically outsource these core operations to AI, the related cognitive capabilities may weaken due to lack of training. Algorithmic systems may structurally amplify ideological homogeneity, reinforce selective exposure, and limit viewpoint diversity.

On the other hand, AI may also augment System 2 — liberating humans from lower-order abstraction operations to handle higher-order abstract problems. As a tool, AI can expand cognitive channels: searching larger information spaces, simulating more complex systems, assisting cross-domain analogy. Whether AI ultimately compresses or expands the human cognitive variable space may depend on the mode of use: passive dependence may lead to atrophy; active collaboration may lead to augmentation. This is an empirical question requiring task-specific measurement, not a direction that can be determined a priori.

IX Generation Z: The First Output of the Reverse Flynn Effect Z세대: 역 플린 효과의 첫 번째 산출

The cumulative effect of all the aforementioned mechanisms has produced its first measurable output in Generation Z. In January 2026, cognitive neuroscientist Jared Cooney Horvath testified before the U.S. Senate: Generation Z (born approximately 1997–2012) may be the first generation in modern history to score below their parents’ same-age levels in cognitive performance. This decline is not anecdotal or cultural pessimism, but measurable in IQ, memory, literacy, numeracy, attention, and problem-solving ability.

Generation Z scores 2 to 4 points lower on standardized IQ and cognitive tests than prior generations, reversing a century-long “Flynn Effect.” Average IQ scores have stopped rising and begun declining in Norway, Denmark, Finland, the UK, and Australia. Today’s 20-year-olds in Norway score approximately 5 to 7 points lower than the prior generation. The timeline generally aligns with the global expansion of smartphones, social media, and digital education platforms in the early 2010s. Unlike IQ tests, metrics such as reaction time and working memory cannot be improved through learning or technique; the fact that both are simultaneously declining makes the reverse Flynn Effect argument particularly compelling.

X The Collapse of Diversity: A Thermodynamic Process of Civilization 다양성 붕괴: 문명의 열역학적 과정

The convergence effects of globalization and scientific civilization are manifested not only at the cognitive level, but more broadly in the full-dimensional collapse of human civilizational diversity. Of approximately 6,000 currently existing languages, only 10% are considered safe from extinction; it is estimated that every two weeks a language loses its last native speaker. Between 1950 and 2010, over 230 languages went extinct.

Each language is an independent abstraction system — a unique mode of classification, ordering logic, analogical structure, and concretization pathway. When a language dies, a channel is permanently closed. Globalization found a local optimum (English + Western consumer culture + quantitative scientific methodology), then used economies of scale to crush all other solutions.

The ultimate microcosm of this process is the morphological evolution of the smartphone. Before 2007, phones were products of a variable world — flip, slide, rotating, circular, triangular, dual-screen, physical keyboards, each the concretized output of a different design path. After the iPhone appeared, the entire industry converged on the touchscreen rectangular slab. Design space collapsed from multi-dimensional to fine-tuning a few parameters: screen size, bezel width, camera count. A reasonable alternative explanation must be acknowledged: smartphone form-factor convergence may partly derive from physical ergonomic optimization — the human hand’s dimensions, visually comfortable screen ratios, and pocket space constraints jointly define the physical boundaries of design space, and form-factor convergence may be the result of functional efficiency and physical limits approaching each other, rather than purely the collapse of civilizational creativity. But even granting this point, software-level convergence (identical app ecosystems, identical interaction patterns, identical attention-capture mechanisms) remains a convergence dimension independent of hardware form factor, and one that more directly affects cognitive diversity.

An even more extreme case is the 2025 Netflix animated film K-POP: Demon Hunters. The virtual girl group HUNTR/X’s song “Golden” reached the top three of daily charts in over 90 countries, and topped the Spotify U.S. chart with 1.48 million single-day plays. Children in 90 countries simultaneously sang the same song, danced the same dance, and participated in the same challenge — and the song’s performers are not even real people. Virtual idols have no personal experiences, no unpredictability — they are products of pure quantitative optimization. Global cultural products can synchronize attention, emotion, and mimetic behavior within a short timeframe, constituting cultural-level high-frequency synchronization.

XI Methodological Self-Reference: The Epistemological Position of This Paper 방법론적 자기 지시: 본 논문의 인식론적 위치

This paper contains a methodological paradox that requires a direct response: it critiques the epistemological limitations of quantitative science throughout, yet extensively uses quantitative science’s data outputs — the percentage decline in disruptiveness scores, IQ test scores, the rate of sperm concentration decline, language extinction statistics — to support its own arguments. Does using quantitative data to argue for the limitations of quantitative methods constitute self-contradiction?

This paper’s response is: this is not self-contradiction, but precisely an instance of this paper’s thesis. What this paper critiques is not quantitative methods themselves (Chapter I already clearly distinguished scientific method, scientism, and the research institution), but the epistemological contraction of scientism that treats quantitative methods as the only legitimate knowledge source. Using quantitative data to reveal the applicability boundaries of quantitative methods is logically legitimate — this is equivalent to using a ruler to measure how far this ruler can measure. There is nothing wrong with the ruler itself; the problem is believing that what the ruler cannot measure does not exist.

At a deeper level: the very fact that this paper has no choice but to use quantitative data is itself proof of scientism’s dominance in the knowledge ecology — in contemporary discourse, if one does not cite statistics, the argument will not be “taken seriously.” This situation, where argumentation must don quantitative garb to gain legitimacy, is itself a manifestation of scientism’s compression effect on the knowledge space. This paper consciously uses this paradox rather than attempting to evade it.

XII Falsifiable Hypotheses 반증 가능한 가설

XIII Conclusion: The Paradox of Scientific Civilization 결론: 과학 문명의 역설

Scientific civilization faces a self-manufactured paradox: the scientific method itself is the most powerful knowledge verification tool in human cognitive history; but when scientism treats quantifiable, reproducible knowledge as the only legitimate knowledge, it systematically suppresses unpredictable exploration, non-standard paths, and high-risk creation. When the research institution replaces real-world impact with publication counts and citation loops as evaluation criteria, knowledge production degrades from an innovation system to a paper factory.

Scientific civilization simultaneously possesses both expansive and compressive effects. In breakthrough stages, new tools open new variable spaces — the microscope, the internet, and gene sequencing each opened new cognitive dimensions. In institutionalized stages, standardization, efficiency optimization, and the demand for predictability begin to close the exploration space that had been opened. The current problem is not a defect of the scientific method, but that the institutional compression effect is overtaking the expansion effect — basic research as a share of R&D has fallen to historical lows, paper disruptiveness has declined over 90% in sixty years, and principal investigators under 35 have dropped from 18% to 2%.

The metric of civilizational progress is redefined within this framework: not the increase of quantitative efficiency, but whether variable space is expanding. A qualification is necessary: expansion of variable space is not disordered inflation, but the increase of effective diversity that can be integrated, inherited, and generate new paths. Variable space contraction may bring short-term stability and efficiency accumulation, but the long-term cost is the system’s loss of adaptability and innovative capacity.

This paper hypothesizes the candidate time window when the institutional compression effect began to overtake the expansion effect as the 1970s–1990s — the publication-count-driven evaluation system globalized during this period, basic research’s share of funding began its sustained decline, and the disruptiveness decline curve also accelerated during this period. This temporal hypothesis requires direct testing with scientometric data.

What this paper proposes is a candidate integration framework; it does not exclude the possibility that each phenomenon has independent mechanisms and localized counter-trends. The framework’s value lies in revealing possibly shared structure, not in asserting single causation. This paper’s positioning is to propose a discussable, revisable, testable critical framework, not a final verdict. The five hypotheses proposed in Chapter XII establish the experimental conditions required to falsify this framework. This paper, together with Cognitive Architecture Theory and the remaining papers in this series, constitutes a cross-scale analytical system.