This study of the flow of online discussion examined how students' recent (current and earlier) messages affected the likelihood of a current message's correct, new idea (correct contribution or CC), positive social cue (PSC), or negative social cue (NSC). A new framework characterized an online discussion message's properties along six dimensions: (1) evaluation of the previous idea (agreement, disagreement, or neutral); (2) knowledge content (CC, wrong contribution, uncertain contribution, justification, repetition, or null content); (3) invitational form (question, command, or statement); (4) social cues (PSC, NSC, or no social cue); (5) personal information (online gender, past posts, and topic initiator or not); and (6) elicitation (eliciting response or not). Dynamic multi-level analysis was used to model statistically 894 messages of 183 participants on 60 high school mathematics topics from one of the world's largest mathematics problem solving website (http://www.artofproblemsolving.com/Forum/index.php?f=149).;Results showed that during online discussion, correct evaluations of others' ideas, CCs, justifications, and questions increased the likelihood of a CC. Agreements increased the likelihood of a PSC, whereas justifications reduced it. Disagreements increased the likelihood of a NSC; in contrast, CCs, wrong contributions, and uncertain contributions reduced it. Meanwhile, earlier PSCs, NSCs, and personal information did not affect the likelihood of a current message's CC, PSC, or NSC. Together, these results support the claim that online discussion processes can promote content-focused discussion. Applied to practice, teachers can help students engage in online content-focused discussions and encourage them to evaluate others' ideas carefully, express and justify their own ideas, post more CCs, and ask more questions during the discussion. |