CARE-AI: Context-Aware AI Study Coaches
Background Information
In large university courses, a significant portion of student learning and assignment preparation occurs online outside the physical classroom, leaving teachers with limited insight into individual study behaviors. Consequently, educators often remain unaware of how students regulate their learning, experience motivation issues, or navigate pedagogical barriers during their coursework. While active study engagement and effective self-regulation strategies are directly linked to improved academic performance, students exhibit substantial individual differences in their ability to maintain these habits consistently. In blended or large-scale courses, instructors simply lack the necessary time and resources to provide individualised, one-on-one self-regulatory coaching to cohorts consisting of hundreds of students. Although general artificial intelligence tools offer promising opportunities to assist learners, traditional generative chatbots frequently risk encouraging student overreliance, effectively outsourcing critical thinking rather than fostering true intellectual autonomy. Existing educational software often fails to deliver pedagogically responsible, context-aware support that systematically strengthens a student's self-regulated learning skills over time. To address this challenge, there is a critical need for intelligent systems that can unobtrusively monitor learning behaviors using virtual learning platform clickstream data. By making hidden study processes visible, universities can deliver timely, personalized scaffolding that guides students through complex assignments while actively building foundational study skills that remain transferable to future academic endeavors.
Aim of the Project
The primary objective of the CARE-AI project is to develop and pilot a family of Context-Aware, pedagogically Responsible AI study coaches integrated directly into the Canvas learning environment. These AI coaches are designed to serve a dual objective: providing immediate scaffolding to help students complete specific assignments while structured to build transferable self-regulation skills. Rather than acting as generic Socratic agents or providing direct content answers, each coach targets specific pedagogical mechanisms grounded in established self-regulated learning theory, such as metacognitive planning or motivational persistence. Personalization is achieved by combining learning analytics derived from Canvas clickstream data with embedded in-chat micro-surveys to unobtrusively assess a student's current self-regulation skills and emotional state. By absorbing routine self-regulatory prompting and monitoring study progress, the automated coaches free valuable instructor time, allowing human educators to focus on higher-level academic guidance that demands human judgment. Teachers retain full oversight, choosing the appropriate coach variant for their assignments while ensuring that assessment and substantive content feedback remain firmly in human hands. Initial development will focus on two foundational coach exemplars tested across four courses within the Industrial Engineering and Innovation Sciences department. Ultimately, the project aims to establish a modular, scalable framework complete with implementation guidelines and training materials to facilitate seamless adoption across other university faculties.
This project is still ongoing.
For more information, please contact:
