The problem with borrowed certainty

The problem with borrowed certainty

A regulator proposes a rule. A respected professor says a technology is dangerous. A commentator says a trend is the defining problem of our time. For me, the label on the messenger is not enough to settle the claim.

Here is a two-question test for weighing a rule, an expert or a trend before it changes a product or business plan.

Question one: who bears the cost, and who gains?

Before I treat a regulation as a safety measure, I want to ask: who can afford to comply, who is excluded, and what did the authors and funders disclose?

The test is to name the rule, estimate the cost for each kind of firm, and check the claimed public benefit.

Who benefits is a question to test, not a verdict; a rule can benefit incumbents and still be right on the evidence, so the benefit question must be followed by the variable question.

The same question applies to experts. My objection to “trust the experts” is not an objection to expertise. It is an objection to treating authority as sufficient when the adviser has a stake in the outcome. S1 I want to examine disclosed financial and institutional interests, then assess the advice on its evidence. In my view, a stake is not proof that someone is wrong. S1 refers to my unpublished notes; no external source has been checked for this article.

The same applies to public discourse. My impression, which I have not checked against data, is that modern discourse rewards status over accuracy. S1 My practical response is to give a claim more scrutiny when I find it flattering to share, not to assume that popularity makes it false.

Question two: what does the evidence say about the underlying variable?

For a debate about energy, fertility, or AI risk, I want to name the outcome at issue before arguing over the policy. What would count as improvement? Over what period? Compared with which alternative? What result would change my mind?

These are my own unverified positions, included to illustrate the test, not as recommendations.

• Energy: position, unchecked. I believe abundant energy unlocks more progress than almost any other single input, and that nuclear deserves a serious place in that picture. S1 The check still needed is a regional comparison of full system costs, reliability, deployment time, emissions, and safety across the available options, not one selected to make nuclear win.
• Fertility: position, unchecked. I think fertility collapse is a bigger near-term problem than most climate scenarios. S1 The check still needed is a defined population, time horizon, and measure of harm, with demographic projections and climate-scenario assumptions laid out side by side. Without a common basis, I should not rank them.
• AI regulation: position, unchecked. I suspect some current AI rules favour firms that already lead. S1 No specific bill is assessed here. A future test would read the bill text, lobbying disclosures and compliance terms of any specific rule I assess. I also need evidence on whether the requirements reduce the specific risk they target. An advantage for incumbents would not, by itself, show that a rule lacks a safety benefit.

None of these positions is a checked conclusion yet; treat them as starting points and test them with the steps below. They are my views, not evidence for your decision.

What to do with this on Monday

Pick one claim shaping a decision you are making this quarter. Write down who would gain if it were acted on, and what evidence would establish those effects. Do not turn a suspected benefit into an assumed motive.

Then name the underlying variable and find two credible data sources that measure it, one of which you did not pick because it agrees with you. Check their methods, scope, and limitations. Write down what result would change your plan before you read the results.

If the evidence changes your assessment, change the plan. If it does not, ask whether it genuinely supports the plan or simply leaves the question unresolved.

Status is a reason to pay attention to someone, not a substitute for checking their claim. I want to build around evidence, not borrowed certainty.