Networked computing allows new ways of learning and working in groups. For example, shared workspace systems with structured graphical representations allow free user interaction and collaborative construction of problem solutions for potentially open-ended tasks. In traditional intelligent tutoring paradigms, student modeling provides the basis for intelligent support. However, in collaborative environments, system control is limited, and hence both group modeling and the provision of intelligent feedback is more difficult. To cope with this limitation, action-based collaboration analysis takes on a process-oriented perspective by monitoring the users' actions and by abstracting and interpreting these raw data in the context of the group collaboration and the problem representation. The system uses an operational hierarchy to model and dynamically identify group activities and provides visualizations of the online analysis results to stimulate self-reflection and tutoring interventions.