A review of automated feedback systems for learners: Classification framework, challenges and opportunities

Teacher feedback provided to learners in real-time is a crucial factor for their knowledge and skills acquisition. However, providing real-time feedback at an individual level is often infeasible, considering limited teaching resources. Fortunately, recent technological advancements have allowed for developing of various computer tutoring systems, which can support learners at any place and time by generating personalized feedback automatically. Such systems have emerged in various domains, tackle different educational tasks, and often are designed in very distinctive ways. Consequently, the knowledge in the field of automated feedback systems is rather scattered across different domains and applications, as we illustrate in this study. This paper aims to outline the state-of-the-art of recently developed systems for delivering automated feedback, and thus serves as a source of systematic information for educators, researchers in the educational domain and system developers. More specifically, the contribution of this study is twofold. Firstly, we offer an extensive literature review of the field. As a result of a rigorous selection process, consisting of 4 phases, a total of 109 automated feedback systems is selected for a detailed review and thoroughly classified against numerous dimensions. Secondly, based on a detailed analysis of the recent literature sources and following the design science research approach, a classification framework for automated feedback systems is developed. This framework is used to classify the selected systems, and thus give a detailed overview of the predominantly available educational technologies, the educational settings in which they are applied, the properties of automated feedback they deliver, and the approaches for their design and evaluation. Based on this analysis, several important observations and recommendations are put forward as an outcome of the study. In particular, our study outlines the current fragmentation of the field, discusses a need for a common reference framework, and calls for more personalized, data-driven and student-centered solutions to exploit a larger set of opportunities offered in this age of data.