In this [Social Science Research Network] paper, Ethan Mollick and Lilach Mollick (Wharton School, University of Pennsylvania) say that in the current debate on the uses and abuses of ChatGPT and other large language models (LLMs), educators haven’t paid enough attention to some important classroom applications. The authors identify five pedagogical strategies that are not used enough in classrooms because they are time-consuming and hard to implement – and show how the new bots can be helpful:
• Generating examples to help students understand difficult and abstract concepts – The best way to explain new and challenging material is to give students a number of examples. “If students are presented with only one example,” say Mollick and Mollick, “they may focus on the superficial details of that example and not get at the deeper concept. Multiple examples of a single concept can help students decontextualize the idea from the example, leading to better recall and understanding.”
Ideally, examples provide a real-world context, anchor abstract ideas in an analogy or story, ground concepts in engaging details, reveal complexity, highlight nuances, help students think critically, and support the transfer of learning to new situations. These demanding criteria show how difficult it is for teachers to generate enough high-quality examples. That’s where the bots come in. All a teacher needs to do is specify the concept, ask for varied examples, and describe the grade level of students and the style of writing required. Click on the full article below for examples on the concept of opportunity costs.
Of course the teacher needs to evaluate the examples generated: Are they factually correct? Are they relevant? Do they have enough detail? Will students find them interesting? Do they connect the abstract to the concrete? Having narrowed down to a good list of examples and presented them to a class, the teacher might then ask students what the examples have in common, have them compare and contrast several, and ask which different aspects of the concept each example highlights.
• Providing varied explanations and analogies to address student misconceptions – Clear explanations are central to good teaching, helping students build mental maps and achieve deeper understanding. But good explanations must be built on students’ prior knowledge, take into account likely misconceptions, plan a step-by-step approach with organizational cues so students can follow along, and provide concrete details and analogies. LLMs can tackle these exacting demands, quickly generating explanations and analogies for a specific grade level and level of understanding. See the article link for a suggested explanation of the concept of photosynthesis for elementary students.
• Producing low-stakes tests so students can practice retrieving information – Checking for understanding is a proven method of cementing material in long-term memory. But generating high-quality tests, quizzes, and mid-lesson “hinge” questions (to see if students are ready to move on to a new topic) is “an effortful task,” say Mollick and Mollick. LLMs can quickly generate diagnostic retrieval exercises. See the article link for examples of quizzes on U.S. history and high-school biology.
• Assessing students’ knowledge gaps to guide instructors’ next steps – The best way for teachers to know what to do next is asking students questions like these:
– What is the most important idea or concept covered in class today?
– Why do you think this idea is important?
– What is the most difficult class concept so far?
– What did you struggle to understand?
– What concept or problem would you like to see explored in more detail?
LLMs can be asked to digest students’ responses to questions like these (perhaps in the middle of a class) and quickly generate an analysis of responses. See the article below for key points and areas of confusion on a lesson on BATNA (Best Alternative to a Negotiated Agreement).
• Creating distributed practice exercises to reinforce learning – “Students need to practice retrieving information not just once but multiple times during a course,” say Mollick and Mollick. It’s also important for students to continuously make connections among the different concepts and skills they’ve learned. But even when students know about the value of distributed practice, they continue to “cram” for tests at the last minute, which means teachers must be intentional about distributing practice. To do so, teachers need to know:
– What are the most important topics for students to remember?
– Which connections between topics are critical and should be practiced often?
– How often and when should students retrieve previously learned material?
– What is the best spacing of assessments to allow just the right amount of forgetting?
– When have students had enough practice?
LLMs can be very helpful designing and scheduling quick quizzes spread out over days, weeks, and months, providing an effective way to lodge concepts and skills in students’ long-term memory. See the article link for examples of distributed practice during a unit on the Enlightenment and the American Revolution.
*This piece appeared in Kim Marshall’s Marshall Memo 978, The full citation for the Mollicks’ article is:
“Using AI to Implement Effective Teaching Strategies in Classrooms: Five Strategies, Including Prompts” by Ethan Mollick and Lilach Mollick in SSRN, March 17, 2023; the authors can be reached at email@example.com and firstname.lastname@example.org.