We propose (and test experimentally) a model of observational learning in which players hold social preferences. To this aim, we design an experiment -based on a classic parlor game known as the Chinos Game- in which we vary (by way of an exogenous iid stochastic process) the probability of getting the prize in the event of a correct guess. By this design, we are able to estimate more efficiently players' sensitivity to difference in payoffs (and how this sensitivity affects information decoding process along the sequence). We also condition our estimates upon additional information on subjects' socio-demographics, risk attitudes and cognitive reflection by way of a questionnaire we collect at the end of each session