One thing I can't understand clearly is why there is so much focus on the volatility smile. Given my knowledge of the Black and Scholes model, this is what I get:
People use the volatility smile as a forecast of future volatility. They assume that the market is 100% efficient and everyone would engage in dynamic hedging if they could or that everyone agree that the Black and Scholes model gives the 'fair' price for the option, whatever it means to them. Therefore any difference between the market price and the BS price shouldn't be interpreted as an opportunity for arbitrage or a presence of model risk, but what 'the market' is saying the BS parameters should be, in special the volatility the market is predicting for the future time until the maturity of the European option.
Are there many issues with this interpretation? If not, why not to estimate the volatility directly by using a stochastic model like Heston instead of working with so many layers of complexity and assumptions? Could you indicate good papers trying to assess the accuracy of the implied volatility as a predictor of future volatility?