AI use cases in teaching at MIT

Updated 10/11/23

How are MIT faculty/instructors leveraging generative AI in the MIT classroom?

Sheryl Barnes, Director of Digital Learning in Residential Education, sought out examples across MIT over a brief one-week period to answer that question.

Barnes, along with Elizabeth Reilley (Arizona State University), and Ryan Lufkin (Instructure) shared their use cases in a webinar - The trailblazers in higher ed: How the best use AI and other emerging technologies to innovate.

Key takeaways for MIT

• Faculty/instructors decide if/how to use AI in their classes. Helpful links:
  • Generative AI Resources from Teaching + Learning Lab
  • Example syllabus/policy statements
  • AI Resource Hub from Sloan Technology Services
  • Generative AI in your teaching: Ways to begin engaging
• AI is most effective for students when helping them to learn (not avoid learning).
  • Alternative assessments can shift the focus to the learning process
  • Student reflection can be paired with AI use for psets
• A common approach is to inform students that they can use AI much like any other tool or human collaborator (with proper citation).
• Students recognize tools such as generative AI as important for their future, and want to understand both how to use the tools and its impact on the world.
• Issues of course exist, from equity and accessibility, to privacy concerns. Projects and courses are beginning to explore best practices.

Use cases at MIT

Adapting course approach

• CMS/W
  • Several instructors re-working assignments to have students synthesize more; use generative AI for first drafts followed by critique or activity; and include data, visualizations, or other interactive forms
• 15.279: Management Communication & 15.276: Communicating with Data
  • Repurposing assignments and sessions to focus more on evaluating communication, and analyzing situations and audiences
  • Instructor: Melissa Webster

Using Chat-GPT or LLMs as a tutor

• 8.01: Physics I - Chat-GPT
  • Developing user interface on top of Chat-GPT for students to synthesize physics concepts, check accuracy of LLM responses, generate practice problems, and discuss acceptable use of LLMs
• 8.01: Physics I - LLM tutor
  • Implementing a LLM physics tutor, with MIT physics content as the data set
• Collaborative research - Generative AI tutor
  • MIT collaboration with Georgia State University and Quinsigamond Community College to design, implement, and evaluate a generative AI tutor for introduction to computer science
  • Lead: Cynthia Breazeal

Linking research and practice

• MAS.S68: Generative AI for Constructive Communication Evaluation and New Research Methods
  • Bringing students into the conversation about the future of communication technology
• 6.S062/MAS.S10/MAS.s60: Generative Artificial Intelligence in K12 Education
  • Introducing generative AI and its opportunities, with project-based work to deploy and test with K-12 students and teachers

There are certainly other use cases at MIT that we’ve inadvertently missed. Want to share your approach to using generative AI in teaching? Fill out this short form.

AI resources at MIT

• MIT President and Provost Initiative to explore the impact of AI
• MIT Ignite: Generative AI Entrepreneurship Competition
• Generative AI Resources from Teaching + Learning Lab
• AI Resource Hub from Sloan Technology Services
• The impact of chatGPT and other large language models on physics research and education

Want more information or a consultation? Email ol-residential@mit.edu to get help from the Residential Education team.