Why Did the Window Break? The Search for a Good Explanation

The Broken Window and the Blueprint

A friend throws a rock at a window. The glass shatters into a thousand pieces. You ask, “Why did the window break?” Your friend says, “Because the rock hit it.” That’s an explanation, but is it a good one? Philosophers of science have asked this for centuries.

One popular answer comes from a group called the new mechanists, active since the 1990s. They say a good explanation is like a machine’s blueprint. It shows all the parts, how they are arranged, and how they work together step by step to produce the outcome.

A mechanism is a system of real things — molecules, cells, pistons — that interact in an organized way to cause something. The parts have causal relationships among themselves. But the relationship between the parts and the whole mechanism is not causal; it is constitutive. That means the parts simply make up the whole, like ingredients make a cake. The blueprint shows the spatial layout, the timing, and the step-by-step sequence.

Consider how a neuron fires an electrical signal (an action potential). A mechanistic explanation describes the cell membrane, the tiny ion channels, where they sit, and the precise order in which they open and close. It’s a movie script for the cell’s machinery.

But how much detail should the blueprint include? Some mechanists, like Carl Craver (a 21st‑century philosopher and neuroscientist), argue that more relevant detail is always better. In the 1950s, Hodgkin and Huxley modeled the action potential but left out the molecular details of how the channels open. Craver calls that a mechanism sketch — not a fully satisfying explanation. Ideally, you’d include the molecules, too.

Not everyone agrees. Imagine explaining lung cancer. The high‑level cause is smoking. But “smoking” contains dozens of cancer‑causing chemicals that differ from cigarette to cigarette. Listing every molecule in every patient would bury the one factor that made the difference for nearly everyone: smoking itself. Sometimes leaving out detail gives a clearer, more useful explanation. Craver and David Kaplan (2020) tried to find a middle ground: include all the detail that is “relevant” for the purpose at hand, but stop before you reach every quantum wiggle.

Wiggling Causes and What‑If Questions

Now picture a dimmer switch in a kitchen. Slide it up, the light brightens; slide it down, the light dims. You have a simple explanation: the switch position controls the brightness. If someone claimed the bulb’s color mattered, you’d object — painting the bulb a new color doesn’t change the brightness. That factor is irrelevant.

This intuition powers the interventionist account of explanation, developed by James Woodward (born 1946) and others. For an interventionist, a good explanation shows that if you were to perform an intervention on the cause — a clean, experimental change — the effect would change. You don’t need a full blueprint. You just need a true generalization that tells how the effect depends on the cause, and that generalization must be invariant over a wide range of changes.

Invariance means the relationship holds under many different interventions. Hooke’s law for springs (force = –k × extension) is invariant as long as you don’t stretch the spring so far that it breaks — a narrow range. Newton’s law of gravity is invariant under a much wider range, until you get close to the speed of light. The more what‑if‑things‑had‑been‑different questions (w‑questions) a generalization can answer, the deeper the explanation.

Take the electric field around a long, straight, charged wire. Physics tells us the field strength depends on the charge density and distance, but not on the wire’s color. A good interventionist explanation uses Coulomb’s law plus the wire’s shape to answer many w‑questions: what if the wire were twisted into a loop? What if it were a sphere? The color is left out because changing it doesn’t alter the field.

Interventionism doesn’t demand the step‑by‑step machinery the mechanists love. It is happy with abstract causes. For example, biologists explain the sex ratio in some species by pointing to “parental investment” — the energy each sex requires. That factor alone makes a difference to which equilibrium emerges, even though the microscopic details differ in every individual. As long as changes in investment are systematically linked to changes in the ratio, you have a causal explanation. That’s a sharp departure from the blueprint ideal.

The Physics Behind the Curtain

Another approach goes even further. The kairetic account, created by Michael Strevens (a 21st‑century philosopher), says the best explanation starts from fundamental physics and then strips away everything that isn’t absolutely necessary. Strevens calls it a two‑factor view. First, identify all the causal influences that physics reveals for an event — say, a window shattering. Second, run an abstraction procedure that removes every detail that is not a difference‑maker for whether the event occurs. What remains is a stand‑alone explanation: the minimal story that still guarantees the shattering.

Most details, like the exact vibration of each glass molecule, affect only how the window breaks, not whether it breaks. So the procedure drops them. But there’s a strict extra rule: the factors that remain must be cohesive. That means the different physical processes that could lead to the outcome must be neighbors in the space of fundamental physics. A rock hitting the window and a sonic boom shattering it are not cohesive; the underlying physics is too different. So a kairetic explanation won’t accept “rock or sonic boom” as a valid cause, even though both break windows.

This creates a problem. Scientists often use equations that apply to wildly different physical systems. The Lotka‑Volterra equations describe predator‑prey cycles whether the predator is a lion chasing a zebra or a spider catching a fly. Strevens handles this by saying those explanations are frameworked — we black‑box the low‑level physics and accept the explanation as useful, but not as deep in an absolute sense. Critics like Angela Potochnik (21st century) point out that practicing scientists treat such explanations as deep, so Strevens’s model doesn’t match how science actually works. The kairetic view pushes the idea that the ultimate measure of a good explanation is its fit with fundamental physics, a demand many find too strict.

Your Life, Your Explanations

You face the same puzzle every day. When you’re late and your teacher asks why, you could deliver a minute‑by‑minute recap: “The alarm battery was 0.02 volts low, so the buzzer circuit couldn’t oscillate…” Or you could say, “My alarm didn’t go off, so I missed the bus.” The second story leaves out irrelevant details and still explains why you’re late. That’s exactly the judgment scientists make.

The philosophical debates about mechanisms, interventions, and abstraction aren’t just academic. They guide how researchers design experiments, what they look for in a new theory, and how we evaluate everyday reasoning. Should a doctor explain a disease by listing every molecule involved, or by pointing to the one lifestyle factor that matters most? Both can be true, but they serve different purposes. Recognizing that explanation comes in degrees of depth and relevance makes you a sharper thinker.

Next time you ask “why?” about anything, remember the broken window, the blueprint, and the dimmer switch. A good explanation might be a machine with all its parts on display. It might be the one cause you can wiggle to see a difference. Or, in some cases, it might be a stripped‑down core that physics guarantees. The real skill is choosing the right tool for the job.

Think about it

1. If a friend tells you, “I failed the test because I didn’t study,” do you need to know which exact brain cells formed the memories? When might a more detailed explanation actually be worse?
2. Suppose you have two explanations for why a plant grew tall: one lists every gene and chemical reaction; the other says, “It got plenty of sunlight and water.” Which is more useful for a gardener? Which is more scientific? Could both be right in different ways?
3. In a video game, pressing a button makes a character jump. Inside the code, the jump is a complicated chain of instructions. Which is a better explanation for a player: “the character jumps because you pressed A,” or the full code? Why might a game designer prefer one, and a programmer the other?