The Problem of Predictability: Unstable Systems, Traffic Dividers and Arms Races

Social scientists seem to do a good job predicting some major events in history and spectacularly poor job with others. This paper is concerned with the puzzle that this poses. It investigates systems of interaction and suggests that, in principle, some criteria can be established to aid our intuitions about what might and might not be predictable. It differs from the more usual approaches which tend to treat history in monolithic terms. Instead it examines a simple concrete model of processes like economic growth and coalition formation.