How can the MIT community leverage generative AI to support learning and work on and off campus?
During the 2024 Science Festival at MIT, faculty and instructors, students, staff and alumni exchanged views on the digital tools and innovations they are experimenting with in the classroom. The panelists agreed that generative AI should be used to enhance, not replace, learning experiences.
Co-sponsored by MIT Open Learning and the Office of the Vice Chancellor, this annual event celebrates innovations in teaching and learning. Introducing up-to-date teaching and learning technologies, panelists emphasized the importance of iteration and teaching students how to develop critical thinking skills while leveraging technologies such as generative artificial intelligence.
“The Science Festival brings the MIT community together to explore and celebrate what we do every day in the classroom,” said Christopher Capozzola, senior associate dean for open learning. “This year’s deep dive into generative AI was reflective and practical – another extraordinary example of ‘mind and hand’ here at the Institute.”
Incorporating generative artificial intelligence into learning experiences
MIT staff and faculty aren’t just keen to experiment with generative AI – some believe it’s an crucial tool for preparing students to be competitive in the job market. “In the future, we will know how to teach skills using generative AI, but to get there we need to take next steps rather than wait,” said Melissa Webster, a lecturer in executive communications at the MIT Sloan School of Management.
Some educators are rethinking the learning goals of their courses and redesigning assignments so that students can achieve desired outcomes in an AI-enabled world. For example, Webster previously combined written and oral assignments to facilitate students develop ways of thinking. However, she saw an opportunity to teach experiments with generative artificial intelligence. If students operate tools like ChatGPT to facilitate with their writing, Webster asked, “How can we still put the thinking part in there?”
One of the up-to-date assignments developed by Webster required students to generate cover letters via ChatGPT and critique the results from the perspective of future hiring managers. In addition to learning how to refine AI’s generative prompts for better results, Webster said “students are thinking more about their thinking.” Reviewing the cover letter generated by ChatGPT helped students determine what to say and how to say it, supporting their development of higher-level strategic skills such as persuasion and audience understanding.
Takako Aikawa, a senior lecturer in MIT’s Global and Language Studies Section, redesigned the vocabulary exercise to give students a deeper understanding of the Japanese language, not just right or wrong answers. Students compared tiny sentences written by themselves and ChatGPT and developed broader vocabulary and grammar patterns beyond the textbook. “This type of class develops not only language skills, but also stimulates metacognitive and analytical thinking,” Aikawa said. “They have to think in Japanese during these exercises.”
While these panelists and other Institute faculty and instructors are redesigning their tasks, many MIT undergraduate and graduate students from various academic departments are using generative artificial intelligence to boost productivity: creating presentations, summarizing notes, and quickly finding specific ideas from long documents. But this technology can also creatively personalize learning experiences. Its ability to convey information in a variety of ways allows students with different backgrounds and abilities to tailor course material in ways specific to their particular context.
For example, generative AI can facilitate with student-centered learning at the K-12 level. Joe Diaz, program manager and STEAM educator for MIT pK-12 at Open Learning, encouraged teachers to support learning experiences in which the student can take ownership. “Take something that kids are interested in and passionate about and they can recognize where [generative AI] may be incorrect or trustworthy,” Diaz said.
Panelists encouraged educators to think about generative AI beyond a course policy statement. When incorporating generative AI into tasks, the key is to clearly define learning goals and be open to sharing examples of how generative AI can be used in ways that are consistent with those goals.
The importance of critical thinking
While generative AI can have a positive impact on learning experiences, users need to understand why huge language models may produce incorrect or biased results. Lecturers, instructors and student panelists emphasized that it is extremely vital to contextualize how generative artificial intelligence works. “[Instructors] try explaining what’s going on on the back end and it will really facilitate me understand the responses I get from ChatGPT or Copilot,” said Joyce Yuan, a senior computer science major.
Jesse Thaler, professor of physics and director of the National Science Foundation’s Institute for Artificial Intelligence and Fundamental Interactions, cautioned that a probabilistic tool should be trusted to give definitive answers without bands of uncertainty. “The interface and output must be in a form that contains elements that can be verified or cross-checked,” Thaler said.
Introducing tools such as calculators and generative artificial intelligence, faculty and instructors on the panel said it is vital for students to develop critical thinking skills in these specific academic and professional contexts. For example, computer science courses could allow students to operate ChatGPT to facilitate with homework if the problem sets are so broad that generative AI tools would not capture the full answer. However, introductory students who have not developed an understanding of programming concepts must be able to recognize whether the information generated by ChatGPT was true or not.
Ana Bell, a senior lecturer in the Department of Electrical Engineering and Computer Science and a digital learning researcher, devoted one class at the end of the semester to course 6.100L (Introduction to Computer Science and Programming Using Python) to teach students how to operate ChatGPT for programming questions . She wanted students to understand why setting up generative AI tools in the context of programming problems and entering as much detail as possible would facilitate achieve the best possible results. “Even if you get an answer, you have to be critical of it,” Bell said. By waiting to introduce ChatGPT until this stage, students were able to critically examine generative AI responses as they spent the semester developing skills to determine whether problem sets were incorrect or might not work in every case.
Scaffolding for gaining experiences
The conclusion of the panelists at the Science Festival was that generative AI should provide a scaffold for engaging learning experiences where students can continue to achieve their desired learning goals. MIT undergraduate and graduate student panelists found it invaluable for teachers to set course expectations regarding when and how to operate AI tools. Informing students about their learning goals allows them to understand whether generative AI will facilitate or hinder their learning. Student panelists asked for confidence that they would operate generative AI as a starting point or treat it as a brainstorming session with a friend for a group project. Faculty and instructor panelists stated that they will continue to modify their lesson plans to best support student learning and critical thinking.
Panelists from both sides of the class discussed how vital it is for users of generative AI to be responsible for the content they create and avoid automation bias – implicitly trusting the technology’s response without thinking critically about why it generated the response and whether it is true. But because generative AI is created by humans making design decisions, Thaler told the students, “You have the power to change the behavior of these tools.”