Hume's Guillotine and the Metric That Lies
David Hume noticed something so simple that for centuries nobody had said it out loud. Writers would reason along comfortably about what is — how the world works, what happens, what causes what — and then, no warning, slip into what ought to be. One sentence about facts, the next about duty, no bridge between. You can’t derive an ought from an is. No pile of facts, however tall, tells you on its own what you should do. This is Hume’s guillotine, and it’s separated more heads from more bad arguments than any blade in history.
I think about it most while staring at a dashboard.
A metric is an is. Engagement is up, latency is down, this number moved. Fact about the world, facts are good. The error — the one everyone makes, the one whole companies are built on — is sliding silently from the is of the metric to the ought of the decision. “Engagement is up” becomes “so we should do more of this,” and nobody notices a value judgment got smuggled in between two sentences. The metric measured what is. It said nothing about what ought. The ought was your goal, your bet on what matters — and the metric quietly took credit for a decision it never made.
This is how you get optimization disasters. You pick a proxy metric because it’s measurable, the number goes up, and you treat the rising number as a moral verdict. But the number is just an is. The ad revenue is up and the product is getting worse, both facts, and the dashboard can’t adjudicate between them because adjudicating needs an ought that lives outside the data. Hume’s guillotine is just remembering that metrics describe the world and values decide what to do about it, and the second can never be outsourced to the first.
The deists and the skeptics around them were obsessed with this boundary because they were trying to build ethics on reason instead of revelation. Hume, the most ruthless of them, basically said: careful — reason tells you what is, beautifully, but the ought comes from somewhere reason alone doesn’t reach. Not nihilism. Jurisdiction. Facts are one court, values another. Stop letting the first pretend to rule on the second.
For a builder this is enormously practical. Choose your ought deliberately, before you choose your metric, because the metric will never choose it for you and will happily pretend it did. What are we actually trying to make true about the world. That’s a values question, answered in words by humans willing to be wrong out loud. Then go find the is — the metric that tells you whether you’re getting there. Pick the metric first and you’ll spend years optimizing a number that was never wired to anything you cared about, wondering why winning feels like losing.
The dashboard is full of is. Your job is the ought. Hume drew the line. Stop letting your metrics step over it.
One of a series of essays. I’m Prajjwal Chittori. prajjwalchittori.com.