Can selectively forgetting training data improve or harm model fairness? This thesis investigates the interplay between machine unlearning and fairness metrics across multiple domains using the ERASURE framework.