Macmillan Learning recently wrapped up EconEd, our annual economics conference which explores the multifaceted ways that instructors teach college economics. This year’s focus was the topic on everyone’s mind: AI’s impact on education, work and life.
There aren't always simple solutions when AI tools can not only quickly finish homework assignments and pass exams, but in some cases, can even outperform students or trick instructors into believing generated essays were written by students. At a time when algorithms and automation are reshaping industries, it’s more important than ever that higher education helps equip students with skills that complement AI. This is especially true in an economics course, where students can develop these critical thinking skills alongside an understanding of how AI intersects with economic theories and practices.
Instructors are considering the best ways to restructure their syllabi and assessment methods and, importantly, seeking out the best ways to support students’ ability to succeed both in class and down the road; they're also exploring how they can best account for the presence of AI tools in the class while at the same time encourage academic integrity.
Here are five things instructors should consider about homework and assessments in their Economics class.
Whether or Not to Use AI is an Economic Decision for Students
Every decision is an economic decision. Thus, it’s also the case for students deciding whether or not to use an AI like ChatGPT when completing homework or other assessments. Using a cost-benefit analysis framework, students have to weigh the benefits of saving time and potentially getting higher grades versus the costs of getting caught, facing punishment, and missing out on learning opportunities.
The Challenge of Academic Integrity
In our digital-first world, the line between legitimate assistance and outright cheating has blurred. With 35% of students admitting to using online tools during remote exams, educators are challenged to reconsider assessment strategies. Wolfers cautions educators about the dangers of relying solely on detection software to catch instances where ChatGPT or other similar models might have been used to complete assignments.
It’s Time to Rethink Traditional Assessments
ChatGPT's performance varies with question type and subject area. For example, it is very proficient in answering introductory economics questions and scored high on multiple standardized tests related to economics, outperforming many students. ChatGPT achieved an A- in microeconomics and an A in macroeconomics in tests conducted at Harvard. It scored 5 out of 5 on AP Microeconomics and AP Macroeconomics exams, placing it among the top students.
Wolfers argues that to discourage cheating, the goal should be to use questions in a way that makes large language models like ChatGPT less reliable. To that end, ChatGPT struggles more with multiple choice than true/false questions, and it’s not much help with the graphical ones. For instance, when presented with a graph depicting economic trends, ChatGPT couldn't interpret the data as effectively. It also struggled with multi-step questions with interdependent information. While its capabilities are advancing quickly, by delving deeper and focusing on application or critical thinking, educators can make these tools less appealing for cheating.
Leveraging Technology's Double-Edged Sword
Ignoring the growing role of AI in education isn't a solution. While the immediate fear is cheating, ChatGPT and similar models can also be used to enhance educational experiences. For example, ChatGPT can act as a learning companion, answering questions, explaining complex concepts, and providing instant feedback. Wolfers demonstrated an AI tutor designed to help students, but without providing direct answers. The tutor follows a Socratic approach, guiding students toward answers rather than giving them outright.
Make the Content Relevant
Making assessments and content in textbooks more relevant and meaningful to students can motivate them to complete their work. Real-world examples and applications can enhance the learning experience not just by presenting information, but by offering context, fostering critical thinking, and facilitating deep understanding. Interactive experiences, real-world case studies, guided exercises, and opportunities for active learning can differentiate a textbook from mere information.
To that end, the role of textbooks should not just be about information delivery but conceptual organization, rich context, and problem-solving guidance. A good textbook doesn't just present data, but crafts a narrative or a framework that makes the data make sense in a way that AI-driven summaries might not -- especially when paired with effective teaching.
According to Wolfers, while tools like ChatGPT might change the landscape of learning, they can't replace the depth, contextual understanding, and human touch that comes from well-crafted educational materials and effective teaching. The challenge is for educators to continually adapt and ensure that the materials and methods they use provide genuine value to students.
A recording of Wolfers’ session alongside content from the last decade of EconEd and fresh content from our teaching community of peer consultants is available at our revamped EconEd page. We will be offering fresh content in this space throughout the year, and giving updates about EconED 2024 in Chicago next fall. For more information from Wolfers, be sure to watch the full webinar: Assigning Homework in a World with ChatGPT.
