Reasoning Well: Deduction, Induction, Abduction & Steelmanning

Reasoning is the work your mind does to get from what you know to what you should believe next. There are three main shapes this work takes, and most people only ever use one of them well. This chapter teaches all three in plain words, plus two habits — steelmanning and changing your mind by degrees — that separate clear thinkers from stubborn ones. Master these and you will argue better, read deeper, and spot opportunities other people walk right past.

The three ways to reason

Let me define each one with a plain sentence first, then we'll use them.

Deduction

You start with a general rule and apply it to a specific case. If the rule is true, the conclusion must be true. It is certain.

Induction

You look at many specific cases and guess the general pattern behind them. The conclusion is probable, not certain — more examples make it stronger.

Abduction

You see a surprising fact and ask, "What's the best explanation for this?" You pick the most likely cause. This is the everyday detective move.

DEDUCTION   rule  -> case   "all -> this one"   certain
INDUCTION   cases -> rule   "these -> all"      probable
ABDUCTION   clue  -> cause  "this -> why?"      best guess

Deduction: certainty, if your premises hold

Deduction is the safest reasoning — but only as safe as the rule you start with. If the general rule is wrong, the airtight logic just delivers a wrong answer with total confidence. So the skill is not the logic; it's checking the premise.

Use deduction to apply what you already trust: policies, definitions, math. When you write or speak, deduction is how you show something follows — "Because X is true, Y must be true." That word "must" is the signal you're being deductive.

Induction: finding the pattern

Induction is how you learn from experience. You notice repeats and form a general belief: "Mornings are our busiest hours." This is most of what reading gives you — authors show examples, your mind extracts the rule. Connecting information across what you read is induction in action.

But induction is a guess, and it gets stronger or weaker with evidence. Two cases is a hunch. Two hundred is a reliable pattern. The classic trap is the hasty generalization — building a big rule on a tiny sample.

Abduction: the best explanation — and the opportunity-spotter

Abduction is the most underused and most useful for a curious mind. The philosopher Charles Peirce named it; doctors, detectives, and good analysts run on it. The move: you see something that doesn't fully explain itself, and you reach for the most likely cause among several possible ones.

It's called inference to the best explanation. "Best" matters — there are always many explanations for any fact. Abduction picks the one that is simplest, fits the most evidence, and requires the fewest lucky coincidences.

Notice what just happened: abduction turned a confusing data point into a specific, testable lead. This is exactly how opportunities hide in data. A weird spike, a stubborn drop-off, a question customers keep asking — each is a "surprising fact" begging for an explanation. The person who asks "what would have to be true for this to happen?" finds the gap, the unmet need, the bug, the new product. Spotting opportunities is abduction.

Steelmanning vs strawmanning

Now to argument. When you disagree with someone, you can represent their view two ways.

Strawmanning

You attack a weak, dumbed-down version of their view — one easy to knock down. It feels like winning. It teaches you nothing and convinces no one who actually holds the view.

Steelmanning

You build the strongest version of their argument — even stronger than they stated it — and respond to that. If you still disagree, your position is genuinely strong.

Steelmanning is also a reading and thinking tool, not just an arguing one. When you read something you disagree with, force yourself to write the single best sentence in its favor. This breaks the "curse of knowledge" of your own opinion and is one of the fastest ways to generate new ideas from what you read — you're forced to think with the author, not just against them.

Changing your mind by degrees (simple Bayesian thinking)

Most people treat belief like a light switch — fully on or fully off — and then resist flipping it because flipping feels like losing. Better thinkers treat belief like a dimmer knob. You start with a confidence level, and each piece of evidence nudges the knob up or down. This is the everyday spirit of Bayesian reasoning, named for Thomas Bayes.

You don't need math. You need the habit:

1. State your starting confidence. "I'm about 60% sure the new layout will sell more."
2. Ask of new evidence: which world is this more likely in? A small test shows a 5% lift — that's more expected if the layout works than if it doesn't.
3. Nudge the knob. Move to maybe 75%. Not 100% — one small test isn't proof.
4. Repeat as evidence arrives.

Putting it together

A clear thinker fluidly switches modes: abduce a likely explanation for what they see, deduce what it would imply if true, induce a pattern from testing it across cases, steelman the rival explanation, and update their confidence as results come in. That loop — guess, predict, test, fairly consider alternatives, adjust — is critical thinking in one sentence.

Practice

1. Label the mode. For these three statements, write D, I, or A: (a) "Every order over $50 ships free; this order is $60, so it ships free." (b) "The last ten Mondays were slow, so Mondays are slow." (c) "The lights flickered and the fridge stopped — probably a power surge." (Answers: a=D, b=I, c=A.)
2. Abduction on data. Take one surprising fact from your week, list three explanations, pick the best by fit-and-simplicity, and name a test.
3. Steelman a stranger. Find an online opinion you disagree with. Write the single strongest sentence in its favor before you write a word against it.
4. Dimmer drill. Take one current belief, assign it a percent, and write what evidence would move it 15 points each way.