How do scientists measure the temperature of Earth a thousand years ago, when no one was holding a thermometer? How do they gauge intelligence, happiness, or the health of an entire ecosystem? These things can't be observed directly, yet science routinely produces reliable answers about them.

The trick is something called proxy measurement: finding something visible that reliably reflects something invisible. Tree rings stand in for ancient climate. Test scores stand in for cognitive ability. Stock prices stand in for economic confidence. Done well, proxies unlock knowledge we'd otherwise never reach. Done poorly, they create false certainty about the wrong things entirely.

Imagine you want to know how warm a summer was five hundred years ago. You can't travel back in time, but trees were there, growing wider rings in good years and narrower ones in harsh years. The ring becomes your proxy: a visible record of an invisible past.

Choosing a good proxy starts with a causal story. There must be a believable reason why your indicator responds to the thing you actually care about. Tree rings work because temperature and rainfall genuinely affect growth. Ice core bubbles work because they trap actual atmospheric gases. The connection isn't a guess; it's a mechanism.

The best proxies are sensitive enough to change when the hidden phenomenon changes, but stable enough not to wobble for unrelated reasons. A thermometer that also responds to humidity, sunlight, and your mood would be useless. Scientists spend enormous effort identifying indicators that move because of the target variable, and not much else.

Takeaway

A proxy is only as good as the causal story behind it. If you can't explain why your indicator should track the hidden thing, you're not measuring, you're guessing.

A proposed proxy is just a hypothesis until it's tested. Scientists validate proxies by checking them against known cases where the answer is already established. If tree rings claim to record temperature, do they match thermometer readings during the years thermometers existed? If they do, confidence in the older readings grows.

This calibration step is crucial. Researchers often combine multiple independent proxies, asking whether ice cores, coral layers, and lake sediments all tell the same story. When different indicators, each with different sources of error, point to the same conclusion, the conclusion gets stronger. Scientists call this convergence.

Validation also means probing the limits. Does the proxy work in cold climates but fail in warm ones? Does it saturate above certain values? Honest scientists publish these limitations rather than hide them. A trustworthy proxy comes with an instruction manual: here is what it can tell you, and here is where it stops being reliable.

Takeaway

Trust grows when independent measurements agree. One proxy is a clue; several proxies pointing the same direction start to look like knowledge.

Proxies fail spectacularly when we forget they are proxies and treat them as the thing itself. Goodhart's observation captures the danger: when a measure becomes a target, it stops being a good measure. A school that teaches to the test no longer produces students whose scores reflect learning, only students who reflect testing.

Failures also happen when the link between proxy and reality quietly breaks. For decades, body mass index served as a proxy for health, until researchers noticed it ignored muscle, bone density, and fat distribution. The number stayed precise; its meaning drifted. A measurement can look authoritative long after it has stopped representing anything useful.

The deepest failures come from circular reasoning. If you define intelligence as what IQ tests measure, you can never discover that the test is wrong. Scientific honesty requires keeping the proxy and the target conceptually separate, so that evidence can challenge the link rather than confirm it by definition.

Takeaway

Every proxy is a translation, and translations can quietly become mistranslations. The number is not the thing; treating it as such is how science becomes superstition with decimal places.

Proxy measurement is one of science's most powerful tools and one of its most fragile. It lets us see into deep time, distant systems, and abstract human qualities that would otherwise remain invisible.

But the power comes with responsibility. Every proxy needs a causal story, ongoing validation, and honest acknowledgment of its limits. The next time someone cites a number, ask what it actually measures, and how anyone knows. That single question separates curiosity from credulity.