This dissertation highlights the journey of discovery along a research path towards a better understanding of context-aware learning systems. It contains material presented or described in a number of journal papers, book chapters and manuscripts written by the author, and it summarizes their findings. Context-aware learning systems are mobile learning environments used in both formal and non-formal educational settings that adapt based on the device’s—and, thus, the learner’s—ever-changing environmental context. These context-aware learning systems may be standalone devices or may involve a number of serverbased resources, all accessible via a user interface. The creation of these contextaware systems involves various types of technologies. These include, but are not limited to, a myriad of sensor technologies, wireless (IEEE 802.11 XX) technologies, and microprocessors and user interfaces. Thus, the means by which a learner’s context is detected can vary greatly from system to system. The origins of context-aware expert systems can be traced back several decades. One of the precursors of context-aware expert systems was ubiquitous computing. As early as the 1980s, ubiquitous computing was described as computing in which sensors and computational elements are embedded seamlessly into everyday objects (Weiser, Gold, & Brown, 1999). Ubiquitous computing provided the foundation for ubiquitous learning, an educational paradigm that focused on the needs and dynamics of learning (Cope & Kalantzis, 2008). In turn, ubiquitous learning led to context-aware ubiquitous learning, which can be described as a means for integrating context-aware technologies and allowing them to detect and adapt to the varying situations and contexts of learners in the real world (Hwang, Yang, Tsai, & Yang, 2009).

To achieve the aforementioned contextual learning, a learning system may be paired with a knowledge base and an inference engine from an expert system. An expert system is a computer program designed to achieve the same results as actual experts in a particular domain or field (Franklin, Carmody, Keller, Levitt, & Buteau, 1988). In 1989, Levi argued that one of the key benefits of expert systems is that they are potentially more accurate than human experts, since they do not suffer from the types of negative issues that may affect human performance. Levi (1989) further suggested that, given this increased accuracy, expert systems could surpass the performance of both human experts and statistical models. Yet, dealing with expert systems involved certain practical problems (Kusiak, 1989). Kusiak (1989) acknowledged the difficulty of formulating a model (which could be used in an expert system) that relies on easily available data and that, in turn, can be easily solved.