We investigate implementation of social choice functions that map states to lotteries, where agents have preferences not only for consequences but also for 'honesty'. We show that in the complete information environments with three or more agents, every social choice function is implementable in Nash equilibrium. This is in contrast with the standard implementation models where agents have preferences only for consequences and no social choice function depending on factors other than agents' preferences is implementable. We show also that in the incomplete information environments with two or more agents, every Bayesian incentive compatible social choice function can be implemented in Bayesian Nash equilibrium by a mechanism that is universal in the sense that it does not depend on the detail of the private signal structure.