Random effects are components of a statistical model that capture variability across different levels of a hierarchical structure, allowing for inferences about populations rather than just the observed data. They are crucial in mixed-effects models, where they account for correlations within groups and improve the generalizability of the model's predictions.