Negative edge weights in a graph can represent scenarios such as cost savings or credits, and they significantly impact algorithms like Dijkstra's, which cannot handle them, leading to incorrect shortest path calculations. Instead, algorithms like Bellman-Ford are used, as they can accommodate negative weights and detect negative cycles, which indicate the possibility of infinitely reducing path costs.