Human and Artificial Agent Interaction to Provide Generative Feedback at University

Recent scholarly investigations have highlighted the critical impor- tance of feedback in enhancing students’ educational outcomes, autonomy and motivation. Nevertheless, despite its acknowledged importance, the practical implementation of feedback processes in everyday teaching is often hampered by large class sizes and time constraints. Recent technological advancements have led AQ1 to the development of diverse computer tutoring systems designed to support the feedback process across various educational domains and tasks. Notably, plenty of tools investigate multiple-choice questions and relatively few on open-ended questions. In order to meet these challenges, we initiated an investigation explor- ing the use of Artificial agents in the evaluation of open texts. The research design investigates three key phases of the assessment process: test preparation and exe- cution, test evaluation and analysis, and recursive feedback delivery. Specifically, this paper explores how the interaction between Artificial Agent and Human can be organized in order to assess open-ended tasks in large classes, defining a recur- sive pathway. The research is carried out under the PRIN AI&F project and the overall goal is to build a system that can support effective and sustainable learn- ing, favoring a formative assessment using generative feedback. First results from the analysis of two tasks completed by 263 university students in the academic year 2023/24 seem to indicate that the Artificial Agent can support the feedback process by suggesting a useful classification of the students’ answers.