OER Knowledge CloudConference PaperMcKlin, TomHarmon, S. W.Evans, WilliamJones, M. G.This first phase of a content analysis of online, asynchronous, educational discussions is designed to generate a method for automatically categorizing messages into cognitive categories using neural network software. This phase of research answers two questions regarding the method of automatically analyzing discussion messages: Can a neural network reliably categorize messages under optimum circumstances, and how can the method be improved to generate great reliability? To determine whether neural network software can reliably categorize messages, two trials were conducted. The first, "best fit" trial, a proof of concept trial comprised only of messages which best fit the categorization model, generated strong reliability figures, and the second, systematic sample, a sample much more indicative of the messages generated in an online educational discussion, produced formative reliability figures from which the method of analysis may be optimized. This analysis also provides a distribution based on cognitive presence categories and subcategories of one semester of graduate online educational messages.2001English272-277United Stateshttps://files.eric.ed.gov/fulltext/ED470101.pdfyescognitive processescomputer mediated communicationcomputer networkscomputer software evaluationcomputer uses in educationcontent analysisdiscussiondistance educationgroup discussiononline systemsweb based instructionworld wide webneural networkhttps://files.eric.ed.gov/fulltext/ED470101.pdfAtlanta, Georgia, United States, North America