I've been meaning to write about this for a while. Barry Dahl has a great post analyzing the back-channel comments from our recent panel discussion with Stephen Downes and Robbie Melton. He concludes that only 31% of the posts were productive, by which he means on-topic questions or comments. This issue came up during the panel discussion itself, and Robbie (brilliantly, in my opinion) characterized it as a "teachable moment."
I think a big reason why there was so much off-topic chatter is that we didn't really establish clear patterns or norms for how the back-channel would be encorporated into the larger dialog. The audience treated it like an experiment because we treated it like an experiment. I heard some suggestions from audience members afterward about how the technology could be modified to improve the experience (e.g., disallow anonymous posting, shut off the flow when the panelists are talking, have a moderator filter the comments, etc.), but before I would want to try imposing any of those hard limits on the participants, I would first want to try having a little more preparatory dialog with them about the most productive ways to use the backchannel and how we all would like to interact with it.