In a discrete choice study, I challenge the common hypothesis of Bayesian updating in the face of new information. To this end, a large-scale on-line experiment is designed to elicit the social desirability of mining for gold, uranium and rare earths in the Canadian province of Quebec. In this experiment, I compare how individuals with opposite priors update their risk beliefs in response to more useful information. As a result, I find mixed evidences of Bayesian updating in varying valuation contexts. Despite of useful information, the good in question matters for valuation purpose in preference analysis.