How does dimensionality reduction affect bias-variance tradeoff?

It reduces variance but may increase bias depending on information loss.

Reducing features simplifies model, lowering variance and overfitting risk, but excessive reduction removes predictive information, increasing bias.