Exploring students’ experiences of humanlike AI avatars feedback vs human feedback

Theoretical and empirical research has shown that feedback can effectively support learning and writing processes by guiding learner's cognitive and metacognitive activities.With the growing availability of generative artificial intelligence (AI), AI-based feedback is increasingly used in educational contexts, promising individualized, scalable support and reduced workload for instructors.However, it is still unclear how learners engage with AI feedback and under which conditions it fosters meaningful learning rather than superficial uptake.This symposium addresses the learning process-oriented design of generative AI-based feedback by focusing on learners' engagement, cognitive activities, and regulatory processes as central mechanisms of learning.Across diverse learning and writing formats, including learning journals, argumentative and academic writing, as well as intelligent tutoring systems, the four contributions investigate how specific AI feedback design features influence learners' engagement with feedback.These features include differently pre-prompted AI chat environments, instructional and reflective feedback dialogues, worked examples as additional learning support, and erroneous AIgenerated feedback.The studies examine how engagement patterns translate into cognitive, metacognitive, motivational, and affective processes, and under which conditions motivational benefits or costs emerge.Taken together, the contributions demonstrate that AI-based feedback entails both learning opportunities and risks, depending on feedback design, patterns of use, and learner characteristics.A leading expert on feedback processes will conclude the symposium by synthesizing the findings and discussing implications for designing AI feedback that supports deep learning while balancing engagement, metacognition, and motivation across educational contexts.