/
2026-4-27 Educator's WG: AI-Powered Learning Design

2026-4-27 Educator's WG: AI-Powered Learning Design

Meeting Notes: Educator's Working Group

Date: Apr 27, 2026

Topic: Generative AI in Course Development and Grading

AI-Created, Human Reviewed

Link to Recording, transcript, and chat log.

At a Glance

The session explored two primary applications of generative AI within the Open edX ecosystem: MIT's Ask Tim, a system-level AI designed for student support and tutoring, and Grady, an AI-assisted grading and feedback tool intended to amplify educator impact and improve student learning outcomes through detailed, automated feedback.

MIT Learn: Leveraging AI for Systems-Level Change

Mary Ellen Wiltrout detailed how MIT uses generative AI through the "Ask Tim" interface to enhance the learner experience at [link removed].

  • Ask Tim Functionalities:

    • Recommendation AI: Assists learners in finding relevant courses based on their specific interests and goals (e.g., preparing for graduate school).

    • Syllabus AI: Allows users to ask specific questions about a course's content, such as whether a particular topic like hemoglobin is covered.

  • In-Course AI Tutoring [00:10:00]:

    • A pilot program featuring an AI tutor that mimics pedagogical behaviors.

    • Rather than providing direct answers, the tutor uses Socratic methods to guide students through problem-solving logic.

    • Video Integration: Includes features for autogenerated video summaries, key definitions, and flashcards based on transcripts.

  • Key Insight: Usability testing showed 90% of learners would use the Recommendation AI again, while 30% expressed interest in the Syllabus AI.

Grady: AI-Assisted Teaching and Grading

Anastasios Sidiropoulos and Liz Bottomy introduced Grady, a tool designed to automate grading while keeping the human educator "in the loop."

  • Core Objectives [00:25:00]:

    • Feedback at Scale: Provides rigorous, timely, and specific feedback for any class size, moving beyond the limitations of multiple-choice questions.

    • Human Amplification: Acts as an "exoskeleton" for professors, automating repetitive tasks (like grading) so they can focus on research, direct student interaction, and reduce grading errors.

  • Key Features & Workflow [00:30:00]:

    • Context Awareness: Ingests syllabi, lecture videos, and external resources (e.g., Khan Academy) to provide students with direct links for remediation.

    • Interactive Rubrics [00:35:00]: Professors can provide high-level instructions (e.g., "be more lenient on numerical errors"), and the AI updates the rubric and regrades all submissions fairly.

    • Analytics: Surfaces "common error clusters" and class-level comprehension patterns to inform potential course redesigns.

  • Scalability [00:45:00]: The system is designed to handle cohorts of 2,000+ learners and is already integrated with Open edX, Canvas, and Blackboard.

Q&A and Discussion Points

  • Technical Implementation: Sarina Canelake noted that MIT's tools likely rely on XBlock Asides, a feature that requires better documentation for broader community use.

  • Academic Integrity: Discussion touched on using AI to monitor "curiosity-driven learning" versus "hacking" a course, with Mary Ellen noting that AI chat histories can help identify "hot spots" where course material may be unclear.

  • Cost: Grady's cost for online settings is approximately an order of magnitude cheaper than the $10–$20 per student/course fee for traditional university settings involving handwritten material.

Next Steps & Announcements

  • Conference Presence: Both MIT and Grady representatives will be at the Open edX conference in Utah.

  • Presentations: Ildi Morris will deliver two talks at the conference regarding Content Libraries and virtual instructor-led training.

  • Next Meeting: No meeting in May due to the conference; the group will reconvene in June.

Contact Information: