Building an Argument That Holds Up

The skills in Section 2 have been mostly defensive — detecting fallacies, catching statistical manipulation, identifying bias, evaluating sources. This chapter flips the perspective. Critical thinking is also constructive. The same habits that make you a better detector of weak arguments make you a better builder of strong ones.

And in a world where AI can generate fluent, plausible-sounding arguments in seconds, the ability to build one that actually holds up under scrutiny is more valuable than ever — because AI-generated arguments often don't.

The Three-Part Structure of a Sound Argument

Most weak arguments fail not because they're wrong, but because they're incomplete. They state a claim and some evidence and assume the connection between them is obvious. It isn't always.

Claim — what you're asserting. One clear, falsifiable statement. "We should launch in Q3" is a claim. "We should think about timing" is not.

Evidence — what supports the claim. Specific, traceable, and relevant. "Our market research shows 68% of target customers make purchasing decisions between July and September" is evidence. "Research suggests timing matters" is not.

Warrant — the reasoning that connects the evidence to the claim. Why does this evidence support this conclusion? "Because our product is a considered purchase and our target customers are most active in Q3, launching then maximizes the chance of landing in an active decision window" is a warrant. Without it, evidence and claim sit next to each other without actually connecting.

Most AI-generated arguments have strong Claims and plausible-sounding Evidence. They frequently skip or gloss over the Warrant — the part that requires real reasoning rather than pattern completion.

The Strongest Objection Test

Before presenting any argument that matters, do this: spend three minutes constructing the strongest possible case against your own position.

Not a strawman. Not the obvious objection you already have an answer to. The genuinely strongest version of the opposing view — the one a smart, well-informed critic would make.

If you can't articulate that objection clearly, you don't understand your own position well enough yet. If you can articulate it and you don't have a good answer to it, that's important information before you commit to a course of action.

This test also makes your argument more persuasive to others. An argument that addresses the strongest objection head-on is dramatically more convincing than one that ignores it. People trust speakers who acknowledge what's hard.

Using AI to Stress-Test Your Argument

Here's a productive way to use AI in this process: after building your argument, ask the model to steelman the opposing position. Ask it to identify logical gaps in your reasoning. Ask it to find the weakest link in your evidence chain.

Use AI as your adversarial editor, not your validator. The model is very good at generating counterarguments when explicitly asked to do so. That's exactly the task you want it for.

The combination — you build the argument, AI stresses it, you refine — produces better output than either could alone.