Representation bias occurs when a dataset or model disproportionately reflects certain groups or perspectives, leading to skewed outcomes or decisions. This can perpetuate stereotypes and inequalities, making it crucial to identify and mitigate such biases in data collection and algorithm design.