In this paper, we introduce a suitable extension of the Black-Litterman Bayesian approach to portfolio construction in the presence of non-trivial preferences about higher moments of asset return distributions. We also present an application to active style allocation decisions in the hedge fund universe. Overall the results in this paper suggest that significant value can be added in a hedge fund portfolio through the systematic implementation of active style allocation decisions provided that a sound investment process is implemented that accounts for both non-normality and parameter uncertainty in hedge fund return distributions.