arguments can't be trusted

Most argument that people engage in isn't formally persuasive. It's a back and forth of giving reasons: I give a reason to do a thing, and you give a reason not to, or a reason why my reason might fail. At some point, one of us can't come up with a response and then the other person loses the argument.

In this model, you win if you can come up with many different reasons to support your position, but it also works if you can just find many ways of saying the same reason. It's a sybil attack in reason-space.

Of course, this isn't a perfect model of conversation. But within this model there's no reason to be sure that the person who 'wins' the argument is actually correct, unless you're really able to sit down and formally map out the reasons and weigh the competing concerns against each other.

When you're young, you see arguments as the proper currency of what you should believe and how you should feel: not just in formal domains like mathematics, but in life more generally. In politics, I believed that my side had better arguments than the other side's. But as I've grown older I become less and less convinced by arguments alone.

This is not ideal: ideally I would remember arguments in favor of everything I believe, and could operate in a regime where arguments were dispositive. But I (like most people) am not always that strong.

We might ask: is there a point to informal exchange of arguments, or is it just arbitrary? Maybe the person who's the better debater just wins, even if they don't actually have more insight.

You could tell the story that we're actually playing a game of proving that we've expended computation to think about the issue. Even being able to run the sybil attack is a proof of work in a nontrivial sense: it's difficult to come up with many different coherent reasons without having formed some kind of representation of the domain. You have to have thought about it and abstract it out: even just to be able to form sentences about it that make sense. The ability to do this is an interactive proof that you have a rich representation and can access it flexibly.

This is maybe a proxy for being good at analytical reasoning, logic or being able to weigh out factors within your internal representation---so, there is a kind of evidence---but it's not robust to an adversarial context, because the objective of being able to argue well is not the same as the objective of being able to think well, and neither of those is the objective of actually being right. If you wanted to actually be right you might do less arguing and more actually going around and asking questions.