Agent-based models (ABMs) have their origins in considerations of complexity science stipulating that many phenomena can be ‘grown from the bottom up’. Explicitly, this was expressed in Epstein & Axtell’s (1996) Growing Artificial Societies as the change from ‘Can you explain it?’ to ‘Can you grow it?’. In 2008, Epstein published an article entitled Why Model? in which he discussed his exasperation with people asking for predictions from ABM, pointing out that many other purposes to which it might be applied are more worthy of consideration than prediction, including explanation, improving data collection, testing theories and suggesting analogies. Fourteen years later, the debate about the predictive powers of ABM is still unresolved. This special issue presents the range of positions on ABM and prediction, tackling methodological, epistemological and pragmatic issues.