Weak Arguments Only Count If Someone Tried

⊕ Figure 1: Satire on False Perspective, by William Hogarth, 1754

A bad argument for a claim is evidence against the claim. That sounds like a license to dismiss anything poorly argued, and it mostly isn’t — the inference only works when you had reason to expect a good argument and didn’t get one. Weak arguments count against a position only to the extent that someone capable actually tried to make the strong case. Stated flatly that seems obvious; in practice the heuristic is easily used as if the “actually tried” part were free, and it isn’t.

The underlying principle is a theorem — Conservation of Expected Evidence:

The expectation of the posterior probability, after viewing the evidence, must equal the prior probability.

P(H) = P(H,E) + P(H,¬E)

P(H) = P(H|E) × P(E) + P(H|¬E) × P(¬E)

The version you actually reach for when reading something is something like Zvi’s gloss of it:

When you read something, you should expect it to change your mind as much in one direction as the other. If there is an essay that is entitled Against Widgets, you should update on the fact that the essay exists, but then reading the essay should often update you in favor of Widgets, if it turns out the arguments against Widgets are unconvincing.

That’s a totally valid Bayesian fact and not contrarian posturing.

But a better form would be using “less convincing than expected” instead of “unconvincing”. “Than expected” means you’re assuming the author made the best case available. And that’s where I’m cautious about this kind of quick-to-use heuristic, especially for contested or under-determined topics.

Capable people can produce weak cases for reasons orthogonal to truth, and those reasons are symmetric to whether they’re right — so in opinion-heavy venues the signal washes out. Competence raises the baseline expectation, low-effort ⊕ low-effort here means either literally low-effort but also cases where the argument is degraded because of audience pull, media constraint, etc… lowers it.

In the legal domain’s adverse inference, the low-effort risk is close to nil. A competent lawyer, absent some form of horrible ethical issue, will / has to make her best argument and thus, in that case, absence of evidence is evidence of absence.

But on controversial, opinionated, ill-defined topic, the opportunity for a low-effort argument rises.

So when you encounter a weak argument, apply the heuristic only if you think that was truly a best effort by the author. And if you don’t, the move isn’t to update or to ignore the argument, it’s to go find the stronger version yourself.

Deciding to not update without doing that work is a trap, the mirror image of the first one: “they didn’t really try” is a fully general excuse. You can attach it to every argument that threatens a view you hold and never update against anything: immunization from reputation (Popper/Albert).