While the goal of expert advisory committees is to obtain objective policy recommendations, they are not immune to peer effects in situations of group decision-making. By studying sequential voting in FDA advisory committees, I show that even experts have a desire to conform to their groups, which may override their own private signals and prompt them to follow the observed votes of their peers. Using unsupervised machine learning to analyze text data from committee meeting transcripts and constructing a measure of deliberation quality, I find that high-quality deliberation counters conformity, making decisions more reflective of private information. The results shed light on group dynamics among policy-makers related to peer effects, group rationality, and the utility of deliberation.