Standards for Evaluating Shared Knowledge Representation Interfaces in Teachable Agents

Abstract

Teachable Agents are a unique type of learning technology that aim to harness Learning-ByTeaching theory in order to enhance student learning outcomes. With most Pedagogical Agents, students typically learn passively by receiving information directly from the agent. Alternatively, Teachable Agent systems differ in that students actively teach the agent, which in turn demonstrates its learned knowledge in assessments using artificial intelligence techniques adopted within the program. Over the 20+ years of research and practice evolving Teachable Agents, a consensus has developed among leading researchers that using visual representations to construct TA knowledge and make its thinking visible (i.e. Shared Knowledge Representation), is a critical component in the design of many Teachable Agent interfaces. Although Concept Maps have typically been used as the primary method for Shared Knowledge Representation, research teams have put forth varied proposals of the efficacy of different Shared Knowledge Representation models. However, the field lacks standards for evaluating these various interfaces. In this dissertation, we conduct a Systematic Literature Review in order to identify what researchers believe to be the most important characteristics of effective shared knowledge representation interfaces. With these findings, we propose a framework for evaluating these interfaces in the form of a standards-based assessment rubric. This rubric includes several qualitative standards for assessing current SKR interfaces as well as to be used in the development of new Teachable Agent interfaces.