Stephen Downes has posted slides and audio from his recent presentation in San Jose on E-Learning 2.0 in Development (audio improves after 2 minutes of introductory background talk). He begins by delivering a well-rounded critique of why education modeling language (EML) and other learning design approaches are ineffective (primarily due to complexity of concepts and different perspectives - i.e. the example he has used over the last few years of Paris (i.e. the person, the substance (plaster of paris), the city). To counter the shortcomings of traditional design, Stephen proposes use of connectivist and connectionist orientations as the basis for seeing learning in today's world. Slightly off topic, connectionism has a long history - Thorndike is one of the first researchers I'm aware. Thorndike's work found its roots in behaviourism...and fell from prominence. Connectionism was then revived by neuroscience and artificial intelligence as a means of explaining thought/thinking/knowing at a neural level. Connectionism focuses on micro-level networks. Connectivism, while incorporating micro-level activities, also incorporates macro (connecting people, databases, social networks) and conceptual (connecting ideas or artifacts expressing ideas) levels.
Anyway, following the connectionism/connectivism discussion, Stephen moves into defining groups and networks...and why education (and today's educational tools) need to be based on network models.