3. Mejora del Proceso
3.1. Propuesta del Plan de Mejora de Procesos
The use of technology for educational purposes has a long tradition, including approaches to integrate film, radio, and television in the classroom (Cuban 1986, cited in Koschmann 1996). It was the advent of computer technology that intensified research interest considerably and gave rise to specific research paradigms of
educational technology. Koschmann (1996) discusses four main paradigms, each entailing specific assumptions about learning and teaching, pursuing specific instructional and research objectives, and employing specific technological and research approaches. In particular, these are: computer-assisted instruction (CAI), intelligent tutoring systems (ITS), Logo-as-Latin, and computer-supported collaborative learning (CSCL).
Computer-assisted instruction (CAI) emerged in the 1960s. Applications in this tradition are restricted to presenting teaching materials in a logical sequence to learners, according to instructional goals and didactical considerations (e.g., some piece of content requires another piece as a prerequisite). Practical exercises are realized through programmed drill-and-practice (Stahl et al. 2006), that is, the computer poses a question, the learner inputs an answer (e.g., multiple choice or fill-in-the-blank), and, depending on the correctness of the answer, a more challenging question is presented, and so forth. Developers of CAI programs are often educational practitioners enabled to create contents through courseware authoring programs. Therefore, corresponding CAI software has a strong practical orientation and reflects prevailing traditional views on instruction and learning. In particular, learning is viewed as the (passive) acquisition of knowledge facts and instruction as the transmission of those facts. CAI can be associated with a behavioristic perspective, focusing on stimulus-response relationships: Presenting teaching materials or feedback (the stimulus) leads to improved performance (the response). Internal mental processes are typically not considered. Corresponding research is mainly interested in instructional efficacy, that is, whether knowledge gains superior to some control can be realized through some sort of CAI.
Intelligent tutoring systems (ITS) emerged in the 1970s. ITS research has its origins in artificial intelligence and cognitive science. Its vision and guiding theme is to emulate human tutors by posing challenging, multi-step problems and providing feedback and hints based on the learner’s problem-solving actions, knowledge level, and misconceptions. ITS research puts particular emphasis on the learner’s mental representations of knowledge and problem- solving (cognitivism). Mental structures and processes are modeled using computational student models. For instance, model tracing (Corbett and Anderson 1995) is a computational method to model mental operations during problem solving. The problem-solving process is often conceived of as the traversal through a problem space, which includes an initial state, solutions states, and a limited set of operations to traverse between states. The system
analyzes each problem-solving step (i.e., each user input) to decide whether the step is correct or incorrect, based on a set of rules representing the system’s ideal model of problem solving. More sophisticated approaches utilize buggy
rules to identify misconceptions or mind bugs learners may have. If a step is
incorrect, the system provides feedback and hints to put the learner on the right track again. Model tracing is often used in concert with another technique called knowledge tracing (Corbett and Anderson 1995). Knowledge tracing is an approach to modeling the probabilities of specific skills (or knowledge
components) being mastered by a learner, based on the analysis of his problem-
solving steps. The ITS approach has notable advantage over the CAI approach. First, the ITS-modeling machinery allows the system to make more informed decisions and give support tailored to the specific needs of learners. For instance, ITS systems may select problems based on the learner’s mastery profile and provide feedback tailored to the learner’s misconceptions. Some ITS systems employ domain reasoners to determine, on the fly, whether arbitrary inputs are correct or not. This is a critical advantage over CAI systems, which are typically far less flexible and restricted to checking for predefined answers in a database. Second, ITS systems trace problem-solving on a more fine-grained level and provide feedback for individual problem steps (VanLehn 2006). VanLehn (2011) presents a meta-analysis comparing different kinds of instructions, grouped according to the level of granularity of the interaction involved. He comes to the result that interactions at the step level (the interaction granularity of most ITS systems) are superior to interactions at the answer level (the interaction granularity of most CAI systems) in terms of the achieved learning gains. As a possible explanation, VanLehn (2011) proposes that the interaction granularity in answer-level systems is too coarse. The step from the question to the final answer is simply too big, involves too much reasoning, so students often resort to guessing or quitting. Conversely, ITS systems provide tailored feedback and hints on intermediary steps, which enables learners to accomplish the step themselves while extending or self- repairing their knowledge bases. However, building ITS systems is a complex matter due to the intricacies of student modeling, which requires representing and updating computational representations of mental structures and processes not immediately accessible to researchers and the system. Therefore, most ITS research has focused on relatively narrow, easy-to-formalize, and procedural domains such as arithmetic and calculus. While earlier ITS research was strongly focused on instructional competence, that is, how good computer models can emulate aspects of a real tutor (e.g., diagnosing misconceptions and
assessing knowledge), the question of instructional efficacy gained considerable importance more recently.
Logo-as-Latin describes a class of approaches that emerged in the 1980s, inspired by constructivist learning theories. ITS systems may be designed with the constructivist ideal in mind too, e.g., the system may prompt learners to reflect on their errors or to self-explain a solution step—as noted by VanLehn (2011), system-generated scaffolding in ITS systems may help students to do most of the reasoning themselves. However, the flow of interactions is still largely controlled by the system along predefined pathways. ITSs typically classify student actions as right or wrong, based on the body of domain knowledge encoded in the system, then try to remediate behaviors not conformant with their model, an approach criticized as the “arrogant ‘tutor knows best’ style of ITSs” (Self 1990). Approaches under the Logo-as-Latin paradigm give learners much more freedom and control over their activities and learning. Corresponding instructional technologies (e.g., micro-worlds and simulations) take a passive role in providing a playground or environment for learners to freely explore concepts of interest, carry out experiments, and test out ideas. Many approaches are based on the Logo programming language, which enables young children to experiment with programming concepts, such as loops and variables. The main goal, however, is not to promote programming skills but rather to foster self-regulated learning and problem-solving skills more generally. To emphasize these more general objectives of the approach, Koschmann (1996) proposes the term Logo-as-Latin in analogy to Latin, whose learning was in former times assumed to improve general intellectual abilities. Since the main goal is to support general skills, evaluation studies often focus on questions of transfer.
Computer-Supported Collaborative Learning (CSCL) was established in the 1990s as an independent field of research. Two developments contributed to the emergence of CSCL (Stahl et al. 2006). First, rapid technological advances and the widespread adoption of personal computers and the Internet raised the question of how these new technologies could be employed to improve education and prepare children for the digital age. Second, as discussed, the learning sciences recognized the great potential of collaborative learning approaches to promote deep content learning and general collaboration and thinking skills. In contrast to CAI and ITS research, CSCL sees the interactions between peers—not the instructions and support provided by the system—as the primary source of learning (Stahl et al. 2006). Yet, the role of technology is
typically not reduced to a pure medium of discourse. Rather, a main portion of CSCL research investigates ways how technology can facilitate, guide, and scaffold high-quality interactions between learners. Two prominent ideas how such guidance can be realized—specific knowledge representation formats and collaboration scripts—will be discussed next. CSCL essentially expands on the theories of learning and instruction adopted by collaborative learning researchers, as described in section 2.6. CSCL research puts particular emphasis on peer interactions, group-level processes, such as knowledge co- construction, and true group tasks involving ill-structured problems. Therefore it is not surprising that argumentation is one of the “flash themes” in CSCL (Stahl 2007). CSCL research employs and partly mixes methodologies from different traditions, including experimental, descriptive, and iterative design approaches (Stahl et al. 2006).
It can be observed that, over the years, the different fields extended their scope and imported questions, ideas, and methods from one another. For instance, a number of more recent ITS approaches do not try to emulate an expert tutor anymore but simulate, for instance, learning companions (Goodman et al. 1998) and tutees to be taught by a learner (Walker et al. 2011). ITS systems nowadays do not solely focus on specific, narrowly focused skills but also try to promote self-regulation and metacognition (Azevedo et al. 2010), help-seeking behaviors (Aleven et al. 2006), and collaboration (McManus and Aiken 1995). ITS researchers transcended the boundaries of formal knowledge domains to target more ill-defined and open ones, which are notoriously hard to tackle using traditional ITS methods and approaches (Lynch et al. 2009). Researchers of the CSCL community increasingly realize the potentials of adaptation technologies—the classical province of ITS research—to support collaborative learning processes (Fischer et al. 2013). There is new research in the cross-section between ITS and Logo-as-Latin. For instance, the MiGen project uses ITS methods to (unobtrusively) support exploratory learning (Noss et al. 2012). The Metafora project takes the idea even one step further by researching ways to automatically support collaborative exploratory learning, combining ideas and techniques from the ITS, Logo-as-Latin, and CSCL traditions (Dragon et al. 2013).