There is a quiet assumption underlying much of modern science, medicine, and policy: if something can be measured, it can be known—and if it can be expressed numerically, it becomes more trustworthy.
This assumption now feels almost self-evident. It shapes clinical guidelines, research funding, institutional evaluation systems, and public decision-making. Statistical evidence is treated as the gold standard. Indicators replace deliberation. Even highly complex phenomena—health, behavior, psychological states—are increasingly translated into scores, metrics, and data points.
Numbers appear neutral, precise, and objective. Yet behind this apparent neutrality lies a deeper transformation: quantification is no longer merely a methodological tool. It increasingly functions as an implicit criterion for what counts as reality itself.
This is the shift from quantification to datafication.
And it raises a fundamental question:
What happens when measurement no longer only describes reality, but begins to shape what we recognize as real?
1. The authority of numbers
Numbers occupy a distinctive epistemic position in modern societies. They appear non-negotiable. While qualitative claims require interpretation, numbers seem to speak for themselves.
This perceived immediacy grants them authority. In medicine, randomized controlled trials are considered the gold standard of evidence. In policy, indicators often replace argumentation. In scientific research, statistical significance is frequently equated with truth.
As a result, numbers do not merely inform decisions—they legitimate them.
However, this legitimacy rests on a tacit assumption: that what is measurable is what is sufficiently real to matter.
This is where the shift begins.
2. Quantification as productive reduction
Quantification, in its original scientific sense, is not problematic. It is a method of deliberate reduction.
It translates complex phenomena into numerical form by selecting certain features and excluding others. This selective abstraction is precisely what makes it powerful: it enables comparison, modeling, and generalization.
In physics, this works extremely well. In epidemiology, statistical modeling can reveal population-level patterns that would otherwise remain invisible. In parts of biology, individual variability stabilizes at aggregate levels.
Importantly, this reduction is usually understood as such: the model is not the world, and the number is not the phenomenon.
The hierarchy remains clear:
Reality → measurement → number → interpretation
Numbers function as tools for orientation, not substitutes for reality itself.
The problem begins when this hierarchy becomes blurred.
3. When quantification becomes the default logic
Over time, quantitative methods have shifted from being one approach among many to becoming the default epistemic framework in many scientific fields.
This is not primarily a theoretical development, but an institutional one. Research funding favors measurable outcomes. Journals prioritize statistically significant results. Software makes statistical analysis widely accessible. Evaluation systems require standardized indicators.
As a result, what is easily quantifiable becomes more visible, more fundable, and more publishable.
And what resists quantification—context, meaning, lived experience, ambiguity—becomes structurally less visible within scientific legitimacy.
Importantly, this rarely happens explicitly. No system declares that only numbers matter. Instead, practices evolve in that direction.
Quantification becomes the default.
4. From quantification to datafication
At this point, a deeper transformation emerges.
Quantification means translating aspects of reality into numbers.
Datafication means treating reality itself as fundamentally accessible through data structures.
Phenomena are no longer primarily understood as complex, meaning-laden processes, but as potential data points.
Behavior becomes data. Health becomes metrics. Social interaction becomes traceable signals. Psychological states become standardized scores.
This shift is subtle but profound.
Because once reality is primarily accessed through data, what cannot be translated into data begins to lose epistemic visibility—not because it ceases to exist, but because it becomes increasingly difficult to integrate into dominant knowledge systems.
Thus, the shift is not necessarily ontological in a strict sense. It is epistemic: what becomes visible as “real” is reorganized.
5. Statistical reality: when models begin to overtake phenomena
Modern statistical practice intensifies this development.
Tools such as p-values, effect sizes, regression models, and meta-analyses are powerful instruments for identifying patterns in complex datasets.
However, they also create a subtle epistemic risk: statistical outputs can appear more concrete than the phenomena they represent.
A statistically significant effect can feel like a discovered fact. A large effect size can feel like a stable property of the world. A model can appear to mirror reality itself.
Yet what these outputs actually represent is more limited:
They are results within formal systems defined by assumptions, not direct access to reality.
 |
| Pic: Evgeny Ozerov on Unsplash |
This distinction is often obscured in practice, especially when statistical outputs are treated as endpoints rather than starting points for interpretation.
6. Effect sizes, averages, and the problem of heterogeneity
One of the clearest examples of this tension is the use of effect sizes.
Effect sizes are methodologically important. They move beyond binary significance testing and allow estimation of magnitude. They are essential for meta-analysis and evidence synthesis.
However, they also rely on aggregation.
When data are averaged across individuals, contexts, and time points, variability is compressed into a single number. This produces clarity—but at a cost: heterogeneity is reduced.
In medicine, this is particularly significant. A treatment may show a modest average effect while being highly effective for some groups and ineffective for others.
The aggregated effect remains statistically valid—but it may not reflect the structure of lived clinical reality.
The model is correct, but incomplete.
7. Datafication in practice: medicine and research systems
Nowhere is this tension more visible than in medicine.
Evidence-based medicine relies heavily on statistical aggregation. Guidelines are built on randomized controlled trials and meta-analyses. This is scientifically indispensable.
At the same time, clinicians frequently report a gap between statistical evidence and individual patient experience.
Patients are not averages. They are complex, context-dependent cases shaped by comorbidities, histories, and individual variability that often disappears in aggregated data.
This is not a failure of statistics. It is a limitation of abstraction.
8. Prediction instead of understanding
A central feature of datafication is a shift in epistemic orientation:
From explanation → toward prediction.
Modern data-driven systems are often evaluated not by how well they explain phenomena, but by how accurately they predict outcomes.
This leads to an inversion: predictive success becomes more important than interpretive depth.
Yet prediction and understanding are not identical. A system can predict without explaining. It can identify patterns without interpreting meaning. It can optimize without comprehension.
This distinction becomes especially visible in machine learning systems, where predictive performance can be high even when interpretability is minimal.
9. What is lost in the process
The expansion of quantitative and data-driven methods does not eliminate other forms of knowledge. But it can marginalize them.
Particularly affected are:
- contextual sensitivity
- subjective meaning
- qualitative differences
- temporal dynamics
- individual variability
These are not errors. They are features of complex phenomena that resist numerical compression.
The issue is not that data are wrong. The issue is that they can become too dominant in defining what becomes visible and relevant.
10. Numbers as a limited form of truth
Quantification is not the problem. It is one of the most powerful epistemic tools of modern science.
Without it, large parts of medicine, physics, and the social sciences would not function.
The problem arises when its limits are forgotten.
Numbers are not reality itself. They are structured representations of selected aspects of reality under specific assumptions.
Data do not replace phenomena. They translate them into formats that allow analysis, comparison, and computation.
The epistemic mistake begins when this translation is no longer recognized as such.
Conclusion: partial truth, not total reality
We are not living in a world where numbers are false. We are living in a world where numbers increasingly define what becomes visible as true.
This is the central tension of contemporary science: not a conflict between quantitative and qualitative knowledge, but between representation and substitution.
Quantification reduces complexity to make it intelligible.
Datafication risks turning this reduction into a default ontology of reality itself.
The challenge is not to abandon numbers, but to properly situate them.
They are not reality itself, but one specific and powerful way of making reality legible—while necessarily leaving something else out.