For users in information intensive environments, it is becoming less likely that a single document will provide an answer to a question. Our goal is to develop techniques that unify redundant and contradictory answers, which are inevitable when multiple documents contain an answer to a question. When users make consequential decisions, such as users in medicine and public health, it is important that they play an active role in the process, and that the answer reflects all of the available evidence. Our study of scientists as they answered research questions based on existing biomedical literature motivated the hierarchical question answering approach that we outline in this position paper. The proposed approach, which we call Information Synthesis (IS), is suitable for the class of questions that have a quantitative answer. We are currently developing a system that supports our proposed solution.