Was wandering about the comp sci department and picked up “Exposing chat features through analysis of uptake between contributions” by D. Suthers and C. Desiato. It was actually super interesting… Probably one of the first scientific papers I’ve actually consumed with interest. I need to read more published papers, though.
Basically, what Suthers and Desiato were studying was the engagement and productivity of distributed communications (in this paper a chat session about teaching and mentorship) in order to help pinpoint where in large online communities engaging and productive interactions are occurring. They came up with an algorithm to find engagement levels and major contributors in the chat and compared the graph created from that data to a graph created from data compiled by a human analysis of the same chat. The graphs ended up looking very similar and it turns out that the rule-based algorithm actually captured subsets of what the human analysis did (which is to be expected, as the rule analysis was based on syntax and limited time-frames and the human analysis had the benefit of semantic understanding and longer time-frames). It was kind of cool to see how you could turn a chat session into data on engagement and principal contributors. They’re still working on expanding the usefulness of the rule-based analysis (not to mimic human analysis, but to provide useful insight into chat engagement and, again, to see where important discussions were taking place and who were the main contributors).