In the types of experiments we typically do in class we don't have enough technology to do computer models. So we have always used the students as agents to model a phenomenon. While this is great, it only allows for one run-through so students often miss the importance of multiple trials or fail to see the impact of the "randomness" built into the activity. I think it would be great to start with this type of model and then have the students translate it into a computer model.
Computer models also eliminate human error in the form of cheating, misinterpreting rules and directions, inaccurate measurement, etc.