Ice-sheet mass balance, and thus contribution to global sea-level rise, is largely controlled by the dynamics of fast flowing ice streams that evacuate ice to the coast. Difficult access and short observational records hinder precise attribution of modern ice-stream retreat, let alone confident prediction of future change. Here, I will present ways in which ice-flow models can be used to understand modern changes, better constrain the extent of past ice-flow reorganization, and ultimately place bounds on likely future behavior. In the first part of the talk, I will show models of Smith Glacier, a rapidly changing ice stream in a likely destabilized portion of West Antarctica. The modeling indicates that the glacier was in a state of precarious stability in the 1990s, and that subsequent retreat and speedup are likely a result of increased sub-ice-shelf melt. The model simulations that match remote sensing data suggest that further retreat is likely, perhaps affecting the broader stability of West Antarctica. In the second portion, I will discuss modeling designed to understand changes on timescales longer than the remote sensing record. I will show simple modeling that links ice-penetrating radar and ice-core data near the South Pole to demonstrate that ice-flow history can be usefully preserved in stratigraphy. I will finish with two examples comparing radar and ice-core data to more complex ice-flow models to understand the stability of the interior of Northeast Greenland. This kind of model-data inter-comparison will become increasingly relevant as we continue to acquire more data across the ice sheets.