Hume
Habit That Looks Like Reason
The classic idea
Habit looks like reason. We never observe cause and effect directly, Hume argued, only repeated patterns. Belief in causation is custom wearing the costume of logic.
The AI inversion
AI makes habit look like reason at an industrial scale. Pattern repetition, run through enough data, produces outputs confident enough to be mistaken for inference.
"AI makes habit look like reason."
Hume's most uncomfortable claim was that "because A, therefore B" was never something we actually observed. We observed A followed by B, repeatedly, and our minds built a feeling of necessity out of repetition. Reason, in the strict sense, played almost no role.
A statistical model does something close to a pure version of this. It has no concept of cause, only the regularities buried in its training data. When it produces a fluent, confident, structured answer, it is doing exactly what Hume said human minds do: converting habit into the appearance of necessity.
The unsettling part isn't that AI lacks reasoning. It's that Hume already suspected we did too, dressed up more convincingly. AI just makes the costume easier to see, because the machine has no story to tell about why the pattern holds, only that it has, so far, repeated.
Why it matters
Systems built on pattern recognition will keep producing reasonable-sounding answers that are habit, not inference, at exactly the moments habit fails.
The question AI brings back
If a fluent answer is just well-dressed repetition, what would actual reasoning need to add?
Sources for further reading
- David Hume, An Enquiry Concerning Human Understanding
- Judea Pearl, The Book of Why
