Popular knowledge management platforms such as community-based question answering sites (CQAs) and electronic networks of practice (ENPs) rely on accurate quality assessment of user-contributed content to ensure effective knowledge creation and exchange. However, quality assessment is subjective by nature. Based on the politeness theory, we hypothesize that answers written more politely are more likely to be perceived as high quality answers by the question asker due to the low face threat. We first test our hypotheses through a random coefficient logit model with data obtained from Stack Exchange, a popular CQA platform. We then conduct a randomized experiment where we exogenously manipulate the politeness level of otherwise similar answers. Our analyses, based on both the Stack Exchange dataset and the randomized experiment, lend strong support to the existence of a politeness bias, which affects question askers’ subjective evaluation of answer quality. This study contributes to the literature in knowledge management, cognitive bias, and behavioral issues in information systems.