overfitting

consider something like a polynomial interpolation:

Interpolating polynomial (or most ML models in general) are smooth, and so interpolating between points will result in “overshooting” regional points and “bouncing around”