De-Idealizing De-Idealization: Beyond Full Reversal

This paper critiques the overly rigid philosophical account of de-idealization and proposes a more expansive, practice-driven framework—categorized into intra-model, inter-model, and measurement procedures—that demonstrates how physicists successfully justify idealizations and scrutinize models without requiring a full reversal to perfect realism.

The Gist

The Big Problem: Why Do We Trust "Fake" Science?

Imagine you are trying to navigate a city using a map. But this isn't a normal map; it's a cartoon map.

• It draws the buildings as simple squares.
• It ignores the hills and valleys.
• It pretends the traffic lights never turn red.
• It assumes everyone drives at exactly 60 mph.

This is what scientists call an idealization. They strip away the messy, complicated details of the real world to make the math solvable. But here's the scary question: If the map is so fake, how can we trust it to get us to the grocery store?

For a long time, philosophers of science said: "You can only trust this map if you can eventually draw a perfect map that includes every pothole, every tree, and every traffic light. Until you have that perfect map, the cartoon map is just a guess."

The authors of this paper say: "That's a silly demand."

They argue that asking for a "perfect map" is impossible. We will never have a map that captures the entire universe perfectly. If we wait for that, we'll never do any science. Instead, we need to change how we check our maps. We don't need to reverse the cartoon back to reality all at once; we just need to show that the cartoon is close enough in the ways that matter.

The authors call this process "De-Idealizing De-Idealization." It's a mouthful, but it just means: Let's stop pretending that checking our models requires a perfect, impossible standard. Let's look at how scientists actually do it.

They found three main ways scientists "check" their cartoon maps to make sure they are reliable.

---

1. The "Tweak the Dials" Method (Intra-Model De-Idealization)

The Analogy: Imagine your cartoon map has a slider that says "Realism." Right now, it's set to "Cartoon Mode" (0% realism).

• The Old View: You can't trust the map until you slide it all the way to "Perfect Reality" (100%).
• The New View: You don't need to go to 100%. You just need to slide it a little bit to "5% Realism" and see if the map still looks like the cartoon. If the roads are still in the right place when you add a tiny bit of detail, you know the cartoon is a good approximation.

In Science:
Scientists take a simplified model (like the Ideal Gas Law, which assumes gas particles are tiny, non-touching dots) and add a tiny bit of complexity (like the Van der Waals equation, which admits particles have size and stickiness).
If the complex model gives almost the same answer as the simple one under certain conditions (like low pressure), then the simple model is justified. You don't need to fix everything; you just need to prove that adding a little bit of reality doesn't break the model.

2. The "Different Lenses" Method (Inter-Model De-Idealization)

The Analogy: Imagine you are trying to describe a mysterious creature.

• Lens A (The Cartoon): You draw it as a stick figure.
• Lens B (The Sketch): You draw it with more muscle definition.
• Lens C (The Photo): You take a blurry photo.

None of these are "perfect." But if the stick figure, the sketch, and the blurry photo all agree that the creature has three legs, you can be pretty sure the creature actually has three legs. You don't need a perfect photo to know that fact is true.

In Science:
Sometimes, you can't just "tweak the dials" of a model because the math is too hard. So, scientists use a completely different mathematical approach to study the same thing.

• Example: In Black Hole physics, one method uses perfect, symmetrical equations (the cartoon). Another method uses messy, topological logic (the sketch).
• If both methods agree that "Black Holes have a point of no return (an event horizon)," then that feature is real, even if the math is messy. The fact that different tools point to the same conclusion proves the idealized model is doing something right.

3. The "Reality Check" Method (Measurement De-Idealization)

The Analogy: This is the most practical one. Imagine you are baking a cake using a recipe that says "Add a pinch of salt."

• The Old View: "Is a pinch the exact right amount? If we don't know the precise chemical weight of the salt, the recipe is fake."
• The New View: You bake the cake. You taste it. It tastes good. You try it again with slightly more salt, and it tastes better. You try it with less, and it's bland.
• Even if you never know the exact chemical weight of the salt, the fact that your cake keeps tasting good and getting better as you tweak the recipe proves the recipe works.

In Science:
This is about comparing the model's predictions to real-world data over and over again.

• The Bohr Model (Atoms): Early on, this model was a huge simplification. But it predicted the color of light hydrogen emits with amazing accuracy. When scientists found small errors (residuals), they didn't throw the model away. They tweaked the model (adding relativity, etc.) to fix the errors.
• The Ising Model (Magnets): This model was originally thought to be useless because it was too simple. But when real experiments on magnets came out, the model's predictions matched the data better than any other model.
• The Lesson: If a model keeps getting closer to reality as you refine it, and if it predicts things that other models miss, it is justified. You don't need a perfect theory; you just need a model that "converges" (gets closer and closer) to the truth.

---

The Takeaway: "Check, Please!"

The authors are telling us to stop waiting for a "Theory of Everything" that is perfectly true. That will never happen.

Instead, science is a process of continuous improvement.

• We start with a cartoon map.
• We check it against reality.
• We see where it fails.
• We tweak it, compare it with other maps, and see if it gets better.

As long as the model is responsive to the world—meaning it changes when the data changes, and it predicts things accurately enough for our needs—it is a valid tool.

The paper concludes with a playful twist on the old idea that idealizations are "unchecked." The authors say: No, we don't leave them unchecked. We check them constantly! We just check them in a messy, practical, "good enough" way, rather than waiting for a magical, perfect standard.

In short: Don't demand a perfect map. Just make sure the map you have gets you to the grocery store without getting you lost. If it does, and you can explain why it works, that's good enough science.

Technical Summary

1. Problem Statement

The paper addresses a central epistemological problem in the philosophy of science: How can scientists justify the use of idealized models (which contain known distortions or false assumptions) to make reliable inferences about real-world target systems?

• The Standard View (McMullin's De-idealization): Traditionally, de-idealization is viewed as a process of "reversing" idealizations by removing simplifying assumptions to arrive at a "full representation" of the target system. Justification is granted only if the idealized model can be shown to converge to this full representation (often via asymptotic reasoning).
• The Critique: Recent philosophers (e.g., Knuuttila, Morgan, Potochnik, Weisberg) argue that this standard view is flawed because:
  1. Impossibility: Scientists never possess a "full representation" of the world; they only have effective theories.
  2. Misalignment: The strategy assumes scientists start with reality and simplify, whereas practice often involves starting with simple models and adding complexity.
  3. Permanence: Some idealizations are permanent and ineliminable due to cognitive or computational limits.
• The Gap: Critics often conclude that idealizations need not be "checked" or that de-idealization is a misguided philosophical construct. The authors argue that this conclusion is too strong because it rejects the entire project of justification based on an overly strict, "idealized" definition of de-idealization.

2. Methodology

The authors employ a practice-driven, case-study approach rooted in the history and philosophy of physics.

• Conceptual Analysis: They deconstruct the standard philosophical definition of de-idealization (specifically the requirement for "full representation" and strict mathematical convergence) and propose a relaxed, attenuated schema.
• Schema Relaxation: They redefine the two core components of de-idealization:
  1. "More Realistic": Does not require a non-idealized "full" model, but rather a more realistic theoretical item (which may still be idealized in other ways).
  2. "Closeness": Does not require strict asymptotic convergence to a final truth, but rather a context-dependent proximity that satisfies the specific goals of the modeling practice.
• Case Studies: They analyze specific historical and contemporary examples from physics (Bohr model, Ideal Gas, Black Holes, Gravitational Waves, Ising Model, QED) to demonstrate how scientists actually justify idealizations without "full reversal."

3. Key Contributions

The paper proposes a tripartite framework for De-idealization, categorizing justificatory practices into three distinct types. This moves the field from a binary view (justified vs. unjustified based on full reversal) to a spectrum of epistemic strategies.

A. Intra-Model De-idealization

• Definition: Justifying an idealized model by showing its convergence to more realistic models within the same parameter family via asymptotic reasoning.
• Mechanism: Identifying hidden parameters in an idealized model (e.g., setting interaction strength to zero) and showing that as these parameters approach realistic values (non-zero), the model converges to the idealized limit.
• Example: The Ideal Gas Law ($PV=nRT$) is justified by the van der Waals equation. By treating intermolecular forces ($a$) and particle volume ($b$) as parameters, one can show that as $a \to 0$ and $b \to 0$ (low pressure, high temperature), the van der Waals model converges to the Ideal Gas Law.
• Limitation: Not always available (e.g., in non-linear systems like General Relativity where parameters cannot be easily isolated or where singularities prevent convergence).

B. Inter-Model De-idealization

This strategy justifies idealizations by establishing conceptual continuity between an idealized model and more realistic models that are not related by simple parameter variation.

• Type 1: Within a Domain:
  • Mechanism: Using different mathematical techniques to model the same target system to show that core features persist across methods.
  • Example: Black Hole Singularities. The Exact Solution approach (Schwarzschild/Kerr metrics) assumes symmetries to find singularities (curvature blow-ups). The Topological-Causal approach (Penrose-Hawking theorems) uses generic conditions (without symmetry) to prove singularities exist via inextendible geodesics. The Initial-Value approach (numerical relativity) shows trapped surfaces form dynamically.
  • Justification: The feature "singularity" is robust across these distinct frameworks, justifying the use of the idealized symmetric models despite their lack of realism.
• Type 2: Across Distinct Domains:
  • Mechanism: Establishing formal and physical similarities between an idealized model in one domain and a more realistic model in a different domain.
  • Example: Gravitational Waves vs. Electromagnetic Waves. Justifying that gravitational waves carry energy (in highly idealized asymptotically flat spacetimes) by drawing an analogy to electromagnetic waves. The argument relies on the fact that the de-idealization strategies (e.g., handling near-field vs. far-field zones) work similarly for both, providing physical, not just formal, similarity.

C. Measurement De-idealization

• Definition: Justifying an idealized model through the robust, diachronic convergence between the model's predictions (and their refinements) and a history of experimental measurements.
• Mechanism: Analyzing residuals (structured discrepancies between model and data).
  • If residuals shrink or become systematically attributable to known errors upon refinement, the model is justified.
  • If residuals persist or diverge, the justification fails.
• Examples:
  • Bohr Model: Initially justified by the close fit to the Rydberg constant. Justification eroded when residuals (fine structure) could not be eliminated by further intra-model refinements, leading to its replacement by Quantum Mechanics.
  • Ising Model: Initially rejected as irrelevant. Gained justification in the 1960s when its predictions for critical exponents (e.g., $\delta$) showed better convergence with experimental data than competing classical models, despite the model itself remaining highly idealized.
  • QED (Dyson Series): Justified initially purely by its ability to "save the appearances" (reduce residuals in predictions) before the theoretical justification (Wilsonian renormalization) was later established.

4. Results

• Rejection of "Full Reversal": The authors demonstrate that the demand for a "full representation" is a philosopher's construct not found in scientific practice.
• Validation of Partial Justification: Idealized models can be epistemically justified even if they are never fully de-idealized, provided they maintain "closeness" to more realistic theoretical items or measurement outcomes in specific, context-dependent ways.
• The "True-ing" Process: De-idealization is redefined not as a one-time event of reaching truth, but as an ongoing process of "true-ing"—continuously tuning models to remain responsive to the world.
• Context-Dependence: What counts as "close enough" or "more realistic" is determined by the specific goals of the modeling (e.g., coarse-grained motion vs. fine-grained muscle movement) and the error tolerances of the measurement context.

5. Significance

• Philosophical Impact: The paper resolves the tension between the "rampant idealization" of science and the need for epistemic justification. It saves the concept of de-idealization from its critics by stripping it of its "idealized" baggage (the requirement for full truth).
• Methodological Shift: It shifts the focus from synchronic analysis (does the model match the world now?) to diachronic analysis (how does the model evolve and converge with data over time?).
• Practical Application: It provides a framework for scientists and philosophers to evaluate models not by whether they are "true" in a literal sense, but by whether they are constrained by the world through specific justificatory practices (intra-model, inter-model, or measurement).
• Defense of Pragmatism: It argues that "saving the appearances" (fitting data) is not merely pragmatic but epistemic, provided it involves robust, refinement-sensitive convergence. This validates the use of effective field theories and other approximate models that are known to be incomplete.

In conclusion, the authors argue that we should not abandon the project of checking idealizations. Instead, we must adopt a "messier," practice-based notion of de-idealization that recognizes multiple pathways to justification, allowing idealized models to remain central to scientific inquiry without requiring impossible standards of total realism.