Desktop version

Home arrow Philosophy

  • Increase font
  • Decrease font

<<   CONTENTS   >>

Promising research within technology-mediated learning environments

While surveying the literature on the use and application of EDM/LA techniques, what became clear is that while there is a growing pool of work that invokes the sentiments and the terminology of the field, very little of it embraces the full complement of affordances EDM/LA has to offer. In order to provide some insight on how EDM/LA can inform our understanding of the strategic process learners invoke when interacting with digital environments as well as how these actions can be captured and then used to dynamically automate scaffolds to improve learning, the following section reviews findings across a diverse set of studies on technology-mediated teaching and learning. The studies reviewed were extracted from the extant pool of research stemming from an EDM/LA lens but also draw on literature that uses in situ data to examine strategic processes outside of EDM/LA.

Intelligent Tutoring Systems

Well prior to the current EDM/LA movement, research examining the development and utility of intelligence tutoring systems (ITS) began to harness the utility of data for the purpose of developing student models and adapting instructional supports to dynamically meet the needs of learners (Sottilare, Graesser, Hu, & Holden, 2013). A very broad and historical definition of ITS is any computer supported application that contains some “intelligence” that can be used for the purpose of teaching and learning (Freedman, Ali, & McRoy, 2000). This definition has evolved as our technologies have become more sophisticated, and now incorporates a focus on automating and aligning three areas of intelligence: complex modeling of a domain, a students’ knowledge base and strategic processing, and the pedagogical strategies of teachers and tutors (Arroyo et al., 2014; Khachatryan et al., 2014).

There are a number of well-known ITS applications (e.g., SQL-Tutor, Mitrovic & Ohlsson, 1999; ALEKS, Craig et al., 2013; ASSISTments, Koedinger, McLaughlin, & Heffernan, 2010) that have large user bases that function to ascertain the systems effectiveness, as well as the refinement of our understanding of how to personalize instruction through data interrogation. Multiple literature syntheses and meta analyses have concluded that, under certain circumstances, ITS can produce, and often even outperform, traditional classroom instruction practices (Kulik & Fletcher, 2016; VanLehn, 2011). From a strategic processing perspective, ITS research has also been useful for exploring the strategic processes necessary to successfully solve presented problems (domain model), collect data on the strategies (or misconceptions) that learners use when engaging with an ITS (student model), and then present feedback in order to reinforce successful strategies or redirect a learner to a more successful strategy when implementing a faulty one (pedagogical model, Anderson & Koedinger; Koedinger

& Corbett, 2006). In ASSISTments, for example, Koedinger and colleagues (2013) used data from engaged learners to successfully ascertain problems/errors within the domain or student models by examining the problem-solving strategies used by engaged learners which led to an iterative, offline redesign of the system to be more sensitive to these types of interactions.

In a slightly different vein, researchers have also isolated a subset of strategies that users engage in an unexpected way. In a strategy generally known as “gaming the system,” certain users systematically exploit affordances of an ITS to simply get through a task with minimal effort rather than engage in deep thinking and learning the presented material (Muldner, Burleson, Van de Sande, & VanLehn, 2010). There are two common strategies that learners employ when gaming the system: trial and error, and help abuse (Baker, Corbett, & Koedinger, 2004a). Not surprisingly, learner use of these strategies has a negative correlation with learning (Baker, Corbett, Koedinger, & Wagner, 2004b). What is interesting, however, is that research investigating gaming-the-system behaviors has found that use of these strategies is best predicted by characteristics such as motivation, boredom, and interest rather than a faulty instructional practice (Rodrigo et al., 2008). While some of these characteristics can be captured by an ITS through crude system analytics such as reaction time, other attributes are best captured through external data streams that feed into an ITS, which can supplement system interactions to develop a more robust student model. To these ends, this is an area where EDM/LA approaches can push forward the development of robust machine-learning ITS systems.

<<   CONTENTS   >>

Related topics