We often consider preferences to be given and unchanging. I reported earlier about risk aversion to be found to be changing according to external circumstances. But this can still be rationalized as invariant preferences if one increases the state space over which preferences are formed to include these circumstances. But could preferences be learned? There is no doubt about this. Look around you and see for example how gastronomic taste runs in families. There is also evidence that intertemporal substitution and savings behavior more generally is inherited (a reference escapes me now).
Pushing the envelope further, Bruno Moreira, Raul Matsushita and Sergio Da Silva find an intriguing result: toddlers are risk lovers, yet older children and adults are risk averse, so it must be that risk aversion is learned. I cannot reject this conclusion, but so is an alternative explanation, that there are hormonal changes or changes in the brain as toddlers age that would alter their preferences. I guess we need to wait for neuroeconomics to study children as well to convince me of the authors' conclusion.