Meeting the needs of specific learners can lead to advantages over traditional “one-style-fits-all” teaching practices. In most cases, adaptive learning systems are based on the identification of learning-styles. The main purpose of such adaptive learning systems is the adjustment of the learning process to suit the individual learner. Individuals are presented with appropriate learning objects according to their fit within the scales of a learner-style model. Such systems can increase learners' motivation and thus improve learning outcomes.
Adaptive systems have mostly relied on the completion of learner-style and personality questionnaires to profile individual learners. However, this method of assessment can be time consuming, and thus off-putting for students, resulting in reduced motivation to learn. Research has indicated that it is possible to detect a user’s learning-style through their behavior patterns when interacting with a learning system. This research has mainly focused on the use of such methods within eLearning environments. However, most existing approaches to the automatic detection of learning style (e.g., Spada et al., 2008) cannot be applied in mobile environments because of the user interaction techniques involved, for example, analysis of mouse movement patterns. It is therefore necessary to develop a method of detection suitable for use with mobile environments.
Background
Research has indicated that it is possible to identify a learner’s learning-style within an eLearning system. In most cases, the research has involved the categorization of learning-style in accordance with the Felder-Silverman Learning Style Model (FSLSM). In this model, “learners are characterized by values on the four dimensions. These dimensions are based on major dimensions in the field of learning-styles and can be viewed independently from each other” (Graf, 2007). Most practical work has focused on the Global/Sequential dimension of the model.
Mouse movement patterns were examined by Spada et al., (2008) as a means of gathering data for the detection of learning-styles. This research found a high degree of correlation between the way in which an individual uses a mouse and their learning-style, as determined using the Felder-Solomon Index of Learning-Styles (FSILS) questionnaire. The researchers were thus able to predict scores on the Global/Sequential dimension of the FSLSM with a high level of confidence, based on measurements of mouse acceleration. This approach was used to determine the Global/ Sequential dimension of an individual's learning-style solely through the way in which they used a mouse whilst interacting with an educational Website. Changes in mouse coordinate as the user moved the input device were recorded. Students also completed the FSILS Questionnaire, which was incorporated into the Website. The results indicate a strong correlation between maximum vertical speed of mouse movement (y axis) and Global/Sequential dimension score. The correlation coefficient was found to be r=-0.8, indicating that students with a lower maximum vertical speed tend to be more Sequential, while students with a higher maximum vertical speed tend to be more Global.
Other researchers have considered user behavior patterns in LMS (Learning Management Systems), including scrolling and time spent on pages, as a means of gathering user interaction data (Graf et al., 2008). Bayesian Networks (Garcia et al., 2005) and Feed Forward Neural Networks (Villivarde et al., 2006) have also been considered in the research.
Popescu (2008) and Cha et al., (2006) have focused on the Visual/Verbal dimension of the FSLSM. Both set out to identify visual and verbal learners by comparing the time spent on visual and verbal learning objects, and use the results to provide personalized content in eLearning environments.
These methods have shown positive outcomes for eLearning environments, in particular the work of Spada et al., (2008). However, in most cases the methods employed in traditional eLearning environments cannot be employed in mobile environments, for example, because mouse-based interaction is excluded. Other biometric interaction-based technologies, including eye tracking and accelerometer technologies, can be used to overcome interaction limitations presented by mobile devices, facilitating the creation of adaptive mobile learning environments. These technologies have been successfully employed within learning environments to determine user's scores on the Global/Sequential dimension (Mehigan et al., 2009; Mehigan & Pitt, 2010; Mehigan & Pitt, 2012) and the Visual/Verbal dimension (Mehigan et al., 2009; Mehigan & Pitt, 2011; Mehigan & Pitt, 2012) of the FSLSM. Therefore, these methods make it possible to identify learning-style automatically in mobile environments.