This paper introduces a novel method for detecting students' confusion in online forums by leveraging self-reported affective states via predefined hashtags. It presents a rule for labeling confusion based on these hashtags, which aligns well with teachers' judgments. An automated classifier for confusion detection is then developed for scenarios without self-reported hashtags, using data from a large-scale Biology course on the Nota Bene platform. This approach aims to enhance tools for educators to identify and address student confusion more effectively.