The Causal Second Law

This paper argues that under specific physicalist assumptions, the causal regularities of special sciences possess an associated entropy that cannot decrease from robust causes to effects, thereby establishing a "causal second law" analogous to thermodynamics that is compatible with non-metaphysical supervenience and does not necessarily imply a fundamental causal time arrow.

The Gist

The Big Idea: A New Kind of "Entropy"

You probably know the Second Law of Thermodynamics: in a closed room, if you drop a cup of coffee, it spills and makes a mess. It never spontaneously un-spills and jumps back into the cup. Things tend to go from order to disorder. This is often called "entropy."

This paper argues that this rule isn't just for physics. It applies to everything we study in science (and even everyday life), like medicine, psychology, economics, and geology.

The author proposes a "Causal Second Law." It says: When a cause leads to an effect, the "messiness" (or variety) of the effect must be at least as big as the "messiness" of the cause.

Think of it like this: You can't get a bigger, more complex outcome from a tiny, simple input without adding more possibilities along the way.

---

The Two Golden Rules

To make this work, the author assumes two things about how our world works:

1. The "Lego" Rule (State-Supervenience)

Imagine the universe is built out of tiny, invisible Lego bricks (physical particles).

• The Rule: Every big thing we talk about (like "a burning match" or "a stock market crash") is just a specific arrangement of those Lego bricks.
• The Metaphor: If you have a description of a "burning match," there is a specific pile of Legos that represents it. If you have a description of a "burned house," there is a different pile of Legos. The paper assumes that for every "special science" description, there is a matching pile of physical Legos.

2. The "Conservation of Space" Rule (Measure-Preservation)

Imagine the Lego bricks are floating in a giant, magical swimming pool.

• The Rule: As time passes, the bricks move around according to the laws of physics, but the total volume of water they occupy doesn't change. If you have a cluster of 100 bricks, they might spread out, but they will always take up the same amount of "space" in the pool.
• The Metaphor: Think of a drop of ink in water. It spreads out, but the amount of ink stays the same. The "volume" of the cause is preserved as it moves through time.

---

The Main Argument: Why Effects Are "Bigger"

Now, let's put these two rules together to see why the "Causal Second Law" works.

The Scenario:
Imagine you have a Cause (a burning match) and an Effect (a burned house).

1. The Setup: The "burning match" is a specific pile of Legos. Let's say it takes up 10 square feet of space in our magic pool.
2. The Movement: Time passes. The laws of physics (the Conservation of Space rule) say those 10 square feet of Legos must move somewhere. They can't disappear, and they can't shrink.
3. The Result: For the match to reliably burn down the house, those 10 square feet of Legos must end up inside the "burned house" pile.
4. The Conclusion: The "burned house" pile must be at least 10 square feet big to hold all the Legos that came from the match.

Therefore: The "size" (or entropy) of the effect is always greater than or equal to the "size" of the cause.

Why Does the Effect Usually Get Bigger?

The paper explains that in the real world, the effect usually gets strictly larger (more entropy) for two main reasons:

Reason 1: Many Roads Lead to Rome (Multiple Causes)

Imagine a house burns down.

• Cause A: A burning match.
• Cause B: A lit lighter.
• Cause C: A lightning strike.

All three different things can lead to the same result: a burned house.

• The "burned house" pile has to be big enough to hold the Legos from the match PLUS the Legos from the lighter PLUS the Legos from the lightning.
• Because the effect has to accommodate all these different possible causes, the "burned house" pile is much bigger than just the "burning match" pile.
• Result: Entropy increases.

Reason 2: The Map is Smaller than the Territory (Descriptive Mismatch)

This is the trickiest part, but here is a simple way to see it:

• Physics sees the world in extreme detail (every single atom).
• Special Sciences (like Economics or Psychology) see the world in "chunks" or "blurry pictures."

Imagine a map of a city.

• Physics is a satellite photo showing every car, every person, and every crack in the sidewalk.
• Economics is a simplified map that just shows "Traffic Jam" or "No Traffic."

If you have a "Traffic Jam" (the Cause), there are millions of different ways the cars could be arranged to create that jam. But the "No Traffic" state (the Effect) might require a very specific, rare arrangement of cars to happen exactly as the laws of physics dictate.

Because our "chunky" descriptions (Special Sciences) are too simple to capture every single detail of the physical world, the "Effect" description usually ends up covering a much larger, vaguer area of possibilities than the "Cause" description. The "blurry" effect is bigger than the "blurry" cause.

---

Addressing the "Time Travel" Objection

You might ask: "If physics is reversible (like a movie played backward), why can't we just run the movie backward and have the house un-burn?"

The author says: You can't.

• If you play the movie backward, the "burned house" (which is a huge, messy pile of Legos) would have to magically shrink down into a tiny "burning match."
• But the "Conservation of Space" rule says the pile can't shrink!
• Therefore, the backward movie is physically impossible for a robust cause. The "burned house" is too big to fit back into the "match" box.

This explains why we have a Time Arrow. We know the future is the "burned house" because that's the only direction where the "size" of the possibilities grows.

Why This Matters

1. It Unifies Science: It suggests that the same logic that explains why ice melts also explains why a bad mood spreads, why inflation happens, or why a disease spreads. All of these follow the "Causal Second Law."
2. It's Not Just Physics: You don't need to be a physicist to understand causality. If you have a reliable cause-and-effect relationship, the "variety" of the outcome is always greater than the variety of the input.
3. It Solves a Puzzle: It explains why causes and effects are asymmetric (why the cause comes first and the effect comes later) without needing to invent new laws of physics. It's just a matter of counting possibilities.

The Bottom Line

Imagine you are pouring water from a small cup (the Cause) into a bucket (the Effect).

• The water (the physical states) never disappears.
• If you have many different cups that can pour into the same bucket, the bucket must be huge.
• If your bucket is "fuzzy" and covers a wide area, it's even bigger.

The Causal Second Law simply states: You can't pour a small cup of water into a smaller cup and expect it to fit. The effect is always at least as big as the cause, and usually much bigger.

Technical Summary

1. Problem Statement

The paper addresses a foundational gap in the philosophy of science and physics: the lack of a unified, non-thermodynamic principle explaining why causal regularities in "special sciences" (e.g., biology, economics, psychology, folk physics) exhibit an asymmetry where effects are "larger" or "more probable" than their causes.

While the Second Law of Thermodynamics dictates that entropy increases in isolated systems, it is unclear if a similar principle governs causal relationships in non-thermodynamic domains. Furthermore, standard objections regarding the reversibility of physical laws (time-reversal invariance) suggest that if microscopic dynamics are reversible, macroscopic causal asymmetries should not be exceptionless. The author seeks to determine if a "Causal Second Law" exists that is robust against these objections and independent of specific thermodynamic assumptions.

2. Methodology

The author employs a dynamical systems approach to causation, formalizing causal claims within the framework of measure theory and deterministic (or stochastic) dynamical systems.

• Framework: The paper models the underlying physical reality as a dynamical system $(P, \Sigma, m, U_t)$, where $P$ is the phase space, $\Sigma$ is a $\sigma$-algebra of measurable sets, $m$ is a measure, and $U_t$ is a time-evolution operator.
• Special Science Descriptions: Descriptions of states of affairs in special sciences (causes and effects) are treated as subsets of the physical phase space.
• Key Assumptions:
  1. State-Supervenience: Every special science description corresponds to a set of physical states (a measurable subset of the phase space).
  2. Measure-Preservation: The underlying physical dynamics preserves the measure (phase volume) of sets over time. This is justified by Liouville's theorem for Hamiltonian systems (classical mechanics, electrodynamics, etc.).
• Definitions:
  • Robust Causal Regularity: A cause $C$ robustly leads to an effect $E$ if almost all (measure-theoretically, a set of measure zero exception) physical states instantiating $C$ evolve into states instantiating $E$ within a characteristic time $t^*$.
  • Causal Entropy: Defined as the phase volume (measure) of the set of physical states instantiating a cause or effect.

3. Key Contributions and Results

A. The Causal Second Law (Main Theorem)

The central result is the Causal Second Law for Robust Causal Regularities:

If a special science satisfies state-supervenience and the underlying physics is measure-preserving, then the causal entropy of a robust cause cannot be larger than the causal entropy of its effect ($m(C) \leq m(E)$).

• Proof Logic: Since the dynamics are measure-preserving, the phase volume of the set of states evolving from $C$ remains constant. If almost all of $C$ evolves into $E$, the volume of $E$ must be at least as large as the volume of $C$ to accommodate them.

B. Conditions for Strict Entropy Increase

The paper identifies two sufficient conditions under which causal entropy strictly increases ($m(C) < m(E)$), creating a causal asymmetry:

1. Multiplicity of Possible Causes: If a single effect $E$ can be robustly caused by multiple distinct causes ($C_1, C_2, \dots$), the phase volume of $E$ must be larger than that of any individual cause.
2. Mismatch of Descriptive Capabilities: Special sciences often lack the descriptive granularity to define the "pull-back" of an effect (the exact set of all physical states leading to it). Consequently, the actual cause described by the special science is a proper subset of the physical states leading to the effect, forcing $m(E) > m(C)$.

C. Response to the Reversibility Objection

The paper rigorously refutes the argument that time-reversal invariance invalidates the Causal Second Law:

• Argument: Even if physical laws are time-reversible, the reverse of a robust causal regularity is generally not a robust causal regularity.
• Reasoning: If $C \to E$ is robust and $m(E) > m(C)$, then the time-reversed set $TE$ (states of $E$ reversed) has a larger phase volume than $TC$. Due to measure preservation, $TE$ cannot evolve almost all its states into $TC$. Thus, the reverse process fails the definition of a "robust cause," preserving the exceptionless nature of the law in the forward direction.
• Crucial Distinction: The paper argues that special science descriptions are not necessarily closed under time reversal; a time-reversed physical state may not correspond to any valid special science description.

D. Relation to Thermodynamics and Jaynes' Explication

The author reconstructs E.T. Jaynes' (1965) argument to show that Thermodynamic Entropy is a specific instance of Causal Entropy.

• By reversing Jaynes' logic, the paper demonstrates that the phenomenological Second Law (experimental entropy increase) can be derived from the Causal Second Law applied to the special science of thermodynamics, provided one assumes reproducible experimental processes and state-supervenience.
• This justifies using the term "entropy" outside of thermodynamics, associating it generally with the relationship between causes and effects.

E. Relaxation of Assumptions

The paper explores how the law holds under relaxed conditions:

• Portional Causal Regularities: If a cause only leads to an effect with a certain efficacy $\alpha < 1$ (rather than "almost all"), entropy can theoretically decrease. However, strict increase is maintained if multiple distinct causes exist or if the "loss" is compensated by other factors.
• Theory-Supervenience: The metaphysical requirement of a single fundamental theory is relaxed to "theory-supervenience," where special science descriptions are empirically adequate relative to an underlying theory (which could be another special science).
• Open Systems: The measure-preservation assumption is compatible with open systems if the phenomenon is embedded in a larger, dynamically isolated composite system (e.g., the solar system).

F. Causal vs. Temporal Asymmetry

A significant philosophical contribution is the distinction between:

1. Entropy from Cause to Effect: The Causal Second Law guarantees $m(Cause) \leq m(Effect)$.
2. Entropy in Time: The law does not guarantee that entropy increases over time ($t_1 < t_2 \implies S(t_1) \leq S(t_2)$).

• The paper argues that "causes precede effects" is an independent postulate. If retro-causal regularities existed (where the effect precedes the cause in time), the Causal Second Law would still hold (entropy would not decrease from cause to effect), but entropy could decrease in time. Thus, the Causal Second Law does not inherently generate a temporal arrow without the additional assumption that causes precede effects.

4. Significance

• Unification: It provides a unified physicalist account of causality across all special sciences, grounding causal asymmetry in the geometry of phase space (measure preservation) rather than specific thermodynamic properties.
• Defense of Exceptionlessness: It offers a robust defense of the Second Law against reversibility objections by shifting the focus from "entropy always increases in time" to "entropy cannot decrease from a robust cause to its effect."
• Methodological Shift: It promotes a "dynamical systems" approach to causation that avoids counterfactuals and mechanisms, relying instead on the mathematical properties of state evolution and measure theory.
• Practical Application: It suggests that the success of thermodynamics in predicting heat engine behavior is a specific case of a broader principle applicable to any special science with robust causal regularities, potentially allowing for the derivation of "special science second laws" in fields like economics or biology.

In summary, Gyenis establishes that causal entropy is a fundamental, non-decreasing quantity from cause to effect in any special science that supervenes on measure-preserving physics, providing a rigorous mathematical foundation for causal asymmetry that is distinct from, yet compatible with, thermodynamic entropy.