The Stavros conjecture: Coining terms for problems makes them easier to solve

Uncharacteristically for me, this article isn’t about technology, it’s about a special kind of logic error (or, rather, suboptimality) that I’ve often seen people commit. So, don’t worry, your HTTP proxies are as efficient as ever. However, I have not seen this fallacy described anywhere, let alone named, so I have named it the “inefficient proxy” fallacy and will describe it here.

It’s not really my place to be writing about errors of reasoning, as I know very little about the subject, but I’m doing it for two reasons. Firstly, I have seen many people commit the very error I am about to describe, so raising awareness helps people to not make the mistake in the first place. Two, I believe that defining and detailing errors of reasoning helps communicate and avoid them more easily. It’s much simpler to be able to say “this is a No True Scotsman” and have people immediately know what you mean, than to try to both point out the fallacy in the argument and simultaneously explain what it is. The argument loses its conciseness and, with it, a lot of its force.

By the way, the “No True Scotsman” fallacy is when someone tries to defend a generalization from counterexamples by changing the definition as the examples come along. For example, when someone says “all immigrants are lazy and stupid”, you point out that their hard-working coworker is an immigrant, and they say “well they’re one of the good ones, I meant the others”, they’re committing a No True Scotsman. Wikipedia has a better example, but you get what you pay for.

The “inefficient proxy” fallacy

So, what is it, exactly? Nothing complicated:

The “inefficient proxy” fallacy is when one tries to measure or optimize an indirect metric (or works towards an indirect goal) to solve a problem, when measuring or optimizing for the solution directly would have been just as easy, or easier.

A concrete example of someone committing the inefficient proxy fallacy would be a government banning the consumption of alcohol (a complicated and difficult task) to decrease drunk-driving accidents, when it would be much simpler and easier to just ban drunk driving.

TODO: Add another example here about measuring something by measuring a proxy instead of the thing directly.

This baby wouldn't be so cavalier with its water usage if it knew it's about to be thrown out.

The inefficient proxy fallacy is related to the idiom “throwing the baby out with the bathwater“, but it’s not identical. The latter is more about inadvertently eliminating something good while trying to eliminate something bad, whereas the former is about trying to change something by influencing something related and hoping it carries over rather than just changing the thing itself.