Intelligent Guided Experiential Learning: Tutoring for Practice
Research has shown that novices learn poorly in unguided environments. To promote learning during the use of immersive, simulation-based learning environments, the IGEL project seeks to integrate techniques from artificial intelligence (AI) and intelligent tutoring systems (ITS) to provide automated guidance to learners. The first form of intelligent guidance IGEL provides is coaching. This provides explicit support during an exercise, in the form of hints and feedback. This requires ongoing assessment of learner actions and algorithms describing when and how to deliver feedback. The second type of guidance comes from a reflective tutor that conducts a post-exercise discussion using the after-action review (AAR) framework of investigating what happened, why it happened, and how to sustain or improve student performance. Our reflective tutoring model, which consists of a large set of production rules, takes as input a log of an entire exercise that includes: the choices of the learner, reactions observed in the simulation, and assessments made by the coach. It then designs an AAR agenda that describes what the tutor needs to address. Finally, the reflective tutor executes its rules to address these agenda items. At its disposal are rules that give positive feedback, ask what was wrong with an action, ask what other action may have been more effective, tell the learner what action would have been best, and explain an observed behavior, among many others.
IGEL software is integrated with the SASO-ST and ELECT BiLAT systems, which both focus on intercultural competence and negotiation training. The IGEL project includes fundamental research on automated assessment, pedagogical authoring, learning science, and natural language processing, all of which play ongoing and critical roles for the IGEL system. In ongoing and future research, we will be investigating advanced approaches to student modeling, modeling of expertise in ill-defined domains, and running studies on learning with serious games.
To set our project goals we first studied the current state of the art in intelligent tutoring systems:
The current state of the art focuses on: * Well-defined domains such as algebra and physics where problems have a clear set of correct solutions * Simple problem solving environments (e.g., writing equations, manipulating graphs) * Custom solutions (i.e., only work in a narrow context) * Bridging the large gap between intelligent tutor performance and expert human tutor performance
Tags: artificial, assessment, automated, coaching, immersive, intelligence, tutoring
Goals
- Intelligent tutoring systems for loosely structured domains where it is difficult to assess student actions, and to generate good alternative courses of action
- Intelligent tutoring systems that interoperate with simulation-based practice environments
- A tutoring architecture facilitating the support of multiple domains and simulation practice environments
- Artificial intelligence technologies (e.g., natural language processing and tutorial planning) to support experiments designed to help build more effective tutoring systems.