Intelligent Tutoring System: Challenges in Authoring, Instructional Management, and Evaluation Methods for Adaptive Instructional Systems

• In Event: 34.075 - Symposium on Intelligent Tutoring Systems, BIG DATA–Learning Analytics, and Automated Humanlike Tutoring: Similarities and Differences

Abstract

This talk will focus on significant challenges and emerging solutions related to the adoption of adaptive instructional systems (AISs) as practical and affordable tools for both individual and team learning. To move AISs from state-of-art to state-of-practice, technology (tools and methods) are needed to: reduce the cost and skills needed to author these complex systems; tailor instruction to optimize learning for every learner for every hour invested; engage learners so they will persist in the pursuit of their training and educational goals; apply AIS domains beyond well-defined desktop applications to enhance their relevance and broaden their appeal; and understand impact of AISs and continuously reinforce their effectiveness. This presentation will focus on AIS challenges through the lens of the Generalized Intelligent Framework for Tutoring (GIFT), an open source instructional architecture developed to enhance the efficiency and effectiveness of authoring, instructional management, and evaluation methods to determine the effect of various instructional methods. GIFT has over 800 government, industry, and academic users in over 50 countries.

AIS technologies deliver content, feedback, and support to the learner, and tailor each learning environment (e.g., training simulation) to harmonize the complexity of the task with the capabilities and limitations of the learner. It is also desirable for AISs to optimize performance, retention, and transfer of skills from previous learning environments to new learning environments or work environments for individual learners or teams of learners.

AISs include Intelligent Tutoring Systems (ITSs) which primarily monitor the state of the learner and the learning environment to support decisions about what should be done next (e.g., prompt the learner for information, provide feedback, provide alternate content, change the pace of instruction, or change the challenge level of the learning environment). GIFT attempts manage this process through a set of pedagogical techniques, strategies, and tactics along with data about the condition of the learner and the learning environment in the Learning Effect Model (LEM).

Techniques are domain-independent instructional policies based on best practices of instruction identified in the literature and include: mastery learning, error-sensitive feedback, metacognitive prompting, adaptive spacing and repetition, and fading worked examples. Strategies are learner-centric, domain-independent plans for action by the tutor, and tactics are domain-dependent actions by the tutor based on instructional context and strategy selection. Through a large scale meta-analysis of the literature, ARL has developed the engine for Managing Adaptive Pedagogy (eMAP) to manage strategy selection within GIFT. Course flow in GIFT is managed by techniques framed by David Merrill’s Component Display Theory which simplifies Gagne’s 9 Instructional Events into 4 quadrants of learning (rules, examples, recall, and practice).

The technologies (tools and methods) used to drive AIS capabilities may be grouped into five functional areas with associated intelligent software-based agents: automated authoring agents, adaptive tutoring agents, conversational agents, mobile instructional agents, and effectiveness evaluation agents. Each area will be discussed with respect to current and emerging solutions, their level of technological maturity, how GIFT is seeking to reduce time/cost/skill level requirements, and which challenges remain open problems.

Author

• Robert Sottilare, U.S. Army Research Laboratory