A topic of recent interest is how to apply crowdsourced information toward producing better software requirements. A research question that has received little attention so far is how to leverage crowdsourced information toward creating better-informed models of requirements. In this paper, we contribute a method following which information in online discussions may be leveraged toward constructing goal models. A salient feature of our method is that it applies high-level queries to draw out potentially relevant information from discussions. We also give a subjective logic-based method for deriving an ordering of the goals based on the amount of supporting and rebutting evidence in the discussions. Such an ordering can potentially be applied toward prioritizing goals for implementation. © 2018 IEEE.