Intro economics courses allow AI use for problem sets, change grading – Yale Daily News
The professors of “Introductory Microeconomics” and “Introductory Macroeconomics” told the News that permitting AI help for problem sets and grading only for completion lets students use the assignments for practice.
Jaeha Jang
Staff Reporter
Katya Agrawal, Contributing Photographer
Since ChatGPT launched in November 2022, professors have adopted varying methods to address the use of artificial intelligence in the classroom.
Neither of the two introductory lectures in Yale’s most popular major — Economics 1115, or “Introductory Microeconomics,” and Economics 1116, or “Introductory Macroeconomics” — has a policy prohibiting AI use. Instead, the current professors of both courses told the News that they have shifted their classroom policies in accordance with advances in AI technology: They allow AI use in problem sets and grade them for completion rather than correctness.
“In ECON 1116, we explicitly embrace the use of AI,” economics professor Michael Peters, who teaches “Introductory Macroeconomics,” wrote in an email to the News. “I think using AI can be quite effective. Being able to discuss question with a chatbot and/or getting a starting point into a new literature and topic is useful.”
Professors of both introductory courses told the News that recent developments in AI technology prompted them to change their problem set grading systems.
Considering the progress of AI technology, Peters wrote, AI can answer problem sets “mostly in a flawless way.” If anything, grading for correctness would punish students who attempt to solve problem sets on their own, he added.
“We now grade for completion and I explicitly tell my students that it is all about trying the problem set by yourself,” Peters wrote. “So far, this seems to be working ok, but this is the first time I am doing that and I will need to experiment more.”
Economics professor Cormac O’Dea, who teaches “Introductory Microeconomics,” also wrote that since he does not ban AI use in class, some students will “understandably use LLMs to check steps,” referring to large language models that power chatbots like ChatGPT.
Grading based on completion, O’Dea wrote, allows him to use problem sets as “low-stakes practice” where students can struggle and learn. Seeking any assistance too quickly — whether from him, a teaching assistant or a large language model — can “short-circuit the slow work where real problem-solving skills form,” he added.
“You won’t have AI in the exam room, and if you lean on it too soon or too often, you’ll be lost when it counts,” this semester’s “Introductory Microeconomics” syllabus notes.
The syllabus also advises students that AI will be part of their college and professional careers and that they should learn how to use AI to “complement your own skills, not replace them.”
Both O’Dea and Peters emphasized that overreliance on AI can hinder learning.
“I think AI can be genuinely helpful for certain tasks, but it cannot replace learning how to reason with economics,” O’Dea wrote in an email to the News.
Peters also wrote that he worries AI technology might generate disparities between students who use it to complement their learning and those who “take the easy route and learn less than they used to.” It requires considerable effort from students to use AI technology in a productive manner, he added.
The “Introductory Macroeconomics” syllabus for this semester warns that AI chatbots can “hallucinate” and provide inaccurate answers, adding that the ability to judge a chatbot’s answer is only acquired with expertise in both chatbot technology and the relevant class material. It also recommends that, when stuck on a problem, students ask AI for a hint instead of the answer.
O’Dea and Peters also emphasized the importance of flexibility when it comes to AI policy in economics courses.
Giovanni Maggi, the department’s director of undergraduate studies, wrote in an email to the News that AI policies in economics courses are “left entirely to the instructors.”
“My approach currently suits the course I am teaching, and it may not suit other courses,” O’Dea wrote. “I am glad our department allows instructors to tailor policies to their courses.”
Economics professor Ray Fair, who teaches “Applied Econometrics: Politics, Sports, Microeconomics,” told the News that he gives students “an open ticket to use Chat as much as they want,” referring to ChatGPT.
While ChatGPT can help students develop ideas for their projects, Fair added, it cannot do the exact empirical research and regression-based analysis that he looks for in their papers.
“Chat is now very good at answering problem sets, making up essays, and things of that sort,” Fair said in a phone interview with the News. “That’s not something that I need to worry about.”
This fall, 528 students are enrolled in “Introductory Microeconomics” and 81 are enrolled in “Introductory Macroeconomics.”
