Dynamic Cognitive Modeling for Adaptive Serious Games
Cognitive modeling can be a viable tool to assess the cognitive state of the users and to determine their current learning needs. For instance, adaptive educational systems must match the learning needs by estimating the level of memorization or forgetting. The research question is, how to model latent cognitive variables such as memory degradation and how to make use of it for adaptivity scenarios in the e-learning context. Tools like cognitive architectures with established psychological underpinnings can help here. However, development of cognitive architecture models is often complex, domain- and application-specific and its transfer or general applicability is not evident. We present an innovative dynamic modeling approach which automatically creates declarative rules from interoperable activity stream observations to form models for the cognitive architecture ACT-R. The developed framework uses those models to analyze user actions according to their frequency, temporal occurrence and memory activation levels. An adaptive e-learning system can use the chunks' activation levels to assess which concepts need repeated user attention. A prototype implementation for a serious game for process training demonstrates the feasibility of the approach.