Definition
Plain language
An extra allowance in a calculation for cases where two ingredients affect each other instead of just adding up.
As stated in the literature
A regression component capturing non-additive dependence between factors; omitting all interaction terms yields a purely additive model, and unexplained residual variance is the natural place higher-order interactions would live.
Also called: interaction terms, higher-order interactions
Why it matters: Leaving them out keeps a model simple but hides exactly the combined effects that make some prompt tricks unexpectedly powerful.
For example, a polite tone and a technical framing might each shift an answer a little on their own, but together shift it far more than the sum of the two.
Heard on the show
“… Assume the outcome is a baseline plus one fixed contribution per fragment, with no interaction terms at all, and solve for those contributions — one number per animal per position, one number …”Episode 243 — How a Hundred Meaningless Word Choices Add Up to Flip a Model's Answer