GIFT

Abstract: This article is intended as a companion document to the more focused report provided by the author at the 2017 American Education Research Association
(AERA) Conference as part of the Technology, Instruction, Cognition & Learning Special Interest Group’s Symposium on Intelligent Tutoring Systems (ITSs). Both the AERA talk and this article focus on adaptive instructional systems (AISs) which are comprised of learners, Intelligent Tutoring Systems (ITSs), and external (non-adaptive) instructional environments. AISs tailor instructional experiences for individual learners and teams of learners based on a model of their learning needs and preferences. An exemplar of an AIS is the Generalized Intelligent Framework for Tutoring (GIFT), an open source architecture for authoring, delivering, managing, and evaluating AIS technologies (tools and methods). This article reviews desired states for AISs in the context of enhancements to GIFT capabilities. This article covers a wide range of desired states for AISs and their affiliated design goals, challenges, and emerging solutions. While we consider the review presented in this paper comprehensive, we acknowledge that it is far from exhaustive. Our primary goal is to present the state of art, potential, and practice in ITS design in order to engage the education and training community in our quest to make AISs ubiquitous.

Keywords: adaptive instruction, Intelligent Tutoring Systems, adaptive instructional systems, Generalized Intelligent Framework for Tutoring (GIFT), authoring tools,
accelerated learning, learner modeling, domain modeling, automated instructional management, distributed learning, mobile learning

Citation: Sottilare, R. (2018). A Comprehensive Review of Design Goals and Emerging Solutions for Adaptive Instructional Systems. Technology, Instruction, Cognition, & Learning (TICL) Journal. Vol. 11, pp. 5–38. Old City Publishing, Philadelphia.

Thoughts on Approaches to Adaptive Learning: An editorial (June 2018)
Robert A. Sottilare, Ph.D.

In examining the three approaches to adaptive learning presented at the AERA conference in 2017 and expanded for publication in the recent issue of the Technology, Instruction, Cognition and Learning (TICL) journal (2018), I found some valuable insights. Fletcher offers a glimpse into the history of adaptive learning through a sampling of technology approaches and instructional domains applied over the last 50 years. Scandura provides a solid viewpoints through his description of structural learning theory (SLT) and the TutorIT and Author IT tools which apply the theory. Scandura provides a nice mix of theory and practice as a model for researchers and developers in the field of adaptive learning. Finally, I provided a review of design goals and challenges for a set of technologies that has been labeled adaptive instructional systems (AISs).

Fletcher’s contribution is important in that it documents the evolution of technologies (tools and methods) for adapting instruction to the learner’s goals, needs and preferences. Fletcher advocates for three categories of adaptation used by the US Department of Education: 1) differentiation: changes to the instructional approach based on learner attributes, but no change to its objectives, 2) personalization: changes to objectives and content based on learner preferences, and 3) individualization: changes to instruction based on the learners’ abilities, prior learning, and learning progress, but no changes to objectives. He also compares and contrasts some of the characteristics and goals of training versus education. These types of formalisms will key to defining and developing standards for AISs moving forward. Finally, a review of the effect size of various technologies round out Fletcher’s discussion.

How to use Fletcher’s article – Fletcher has provided you with a neat package detailing some of the most salient achievements in adaptive learning and particularly in digital tutoring over the last 40+ years. If you consider yourself a student of adaptive learning, follow some of the paths detailed in Fletcher’s paper to help shape your understanding of this fascinating scientific discipline.

Scandura’s contribution is important because it reflects the efforts of someone who has been in the arena – a doer. You simply cannot understand what it takes to develop an intelligent tutoring system without jumping into the arena and building one. Scandura’s approach to developing mathematical tutors is built upon structural learning theory. According Scandura and structural learning theory, “what is learned are rules which consist of a domain, range, and procedure. There may be alternative rule sets for any given class of tasks. Problem solving may be facilitated when higher order rules are used, i.e., rules that generate new rules. Higher order rules account for creative behavior (unanticipated outcomes) as well as the ability to solve complex problems by making it possible to generate (learn) new rules.” Through this learning process, we, as learners, build, modify, and rebuild our mental models.

How to use Scandura’s article – Scandura has provided you with a model for developing an AIS that uses micro-adaptations in a domain of instruction. If you were to use this model and apply it to other domains, it would be essential to understand the idiosyncrasies of that new domain, but the process of developing/vetting content, identifying objectives and measures, modeling misconceptions, and evaluating the learner’s progress would be very similar in other cognitive task domains. Regardless of whether you have created AISs from scratch or used existing tools to develop them, the work by Scandura will help you understand the amount of time and specialized skills required to handcraft AISs. This realization that reuse among and between AISs is rare may lead you to understand and I hope advocate for AIS standards. Recently, an IEEE AIS standards working group (Project 2247) was approved under the IEEE Learning Technologies Standards Committee. If you are interested in this topic, please sign up at instructionalsciences.org.

I believe my contribution is important because sometimes it is not all about what you do, but the vision you provide for others to take on the challenges you define. In 1900, the German mathematician, David Hilbert published 23 unsolved mathematical problems to focus the agenda of the mathematics community for the next hundred years. My intent is to focus adaptive learning research on the seven challenges that I defined for not the next hundred years, but gleefully the next five to ten years. My hope is that readers of my TICL article will feel compelled to help solve some or all of these challenges.

How to use Sottilare’s article - Perhaps one or more of the challenges I listed in my article fall into the scope of your research or even better, perhaps you have already overcome one or more of these challenges. The article provides lots of details about generalized authoring tools, developing automation to make instructional decisions, modeling individuals and teams, using content and instructional strategies to build rapport and engagement, modeling teamwork, applying adaptive instruction to new domains, and evaluating AIS efficiency and effectiveness. It is not that folks have not examined these problems, but that generalized solutions have been slow to develop. If you have a solution or want to apply yourself to one of these problems, please chime in, dialog with us, and show us a better way forward.