A new AI tool designed to guide college students through complex economic concepts, rather than simply providing answers, has shown promising results in improving exam scores, according to research conducted at the University of Wisconsin-La Crosse. The tool, named Macro Buddy, was developed by economists seeking to address concerns that widespread access to generative AI could hinder genuine learning.
The research, conducted in spring 2025, comes as a recent survey revealed that approximately 90% of 1,100 U.S. College students at two- and four-year institutions reported using generative AI for academic purposes, ranging from drafting assignments to clarifying difficult topics. However, the study focused on whether AI could be utilized as a learning aid, rather than a shortcut to answers.
At the University of Wisconsin-La Crosse, 140 undergraduate students enrolled in a macroeconomics course were divided into four sections. All students received identical course materials, assignments, and exams, which were administered in person without access to notes or external resources. After an initial exam, the sections were randomly assigned different study formats. One group continued to study individually without AI assistance, while another worked in groups without AI. A third group studied individually with Macro Buddy, and the final group combined group work with the AI tool.
Macro Buddy was built using ChatGPT’s custom GPT feature, with its web access disabled, ensuring it relied solely on course materials provided by Dr. Saharnaz Babaei-Balderlou, a Teaching Assistant Professor of Economics at the university. The AI was specifically trained using lecture transcripts, slides, and homework questions from the macroeconomics course. Crucially, Macro Buddy was designed to function as a tutor, prompting students with follow-up questions to encourage them to articulate their reasoning, rather than simply providing solutions.
For example, if a student inquired about the relationship between lower prices and consumer spending, Macro Buddy would not offer a direct explanation. Instead, it might ask the student to consider the impact of falling prices on purchasing power, guiding them to arrive at the answer through their own reasoning. Researchers emphasized this distinction, noting that AI tools that simply deliver answers can prevent students from actively engaging with the material and developing a deeper understanding.
Initial results showed a decline in average scores across all four groups on the second exam. However, by the third exam, significant differences emerged. Students who combined the use of Macro Buddy with group discussions achieved the highest average scores. Those who used Macro Buddy individually also outperformed students who worked alone without AI assistance. The group that worked in groups without AI showed the smallest improvement compared to the other groups.
Researchers believe the success of the combined group may be attributed to the increased comfort level students developed with Macro Buddy over time, allowing them to effectively test their understanding while also benefiting from the collaborative learning environment. Explaining concepts to peers and responding to questions further reinforced their understanding.
Dr. Babaei-Balderlou has made Macro Buddy available as a course AI TA for her Global Macroeconomics (ECO120) course, describing it as a “24/7 study partner” designed to reinforce classroom learning. The tool is updated each semester to align with current course content.
The findings suggest that the impact of AI on learning may depend less on the technology itself and more on how We see designed and integrated into the educational process. The research highlights the potential for AI to support, rather than replace, critical thinking skills when structured appropriately.
