When students use artificial intelligence to learn new subjects, their underlying motivations dictate whether they actually retain information or just acquire superficial facts. A recent experiment found that learners instructed to focus on personal understanding absorbed more knowledge than those instructed to outsmart their peers. The students focused on competing with others also experienced higher anxiety and approached the learning task with shallow strategies. The research was published in Applied Cognitive Psychology.
In the realm of educational psychology, experts rely on a concept called achievement goal theory. This framework defines the internal reasons why an individual puts effort into academic tasks. A student might read a textbook to genuinely understand the workings of the universe, or they might read it solely to secure a passing grade.
Instructors often shape these internal goals by emphasizing different measures of success in the classroom. This environmental influence is known as a goal structure. Teachers can alter this structure through the design of their assignments, the way they grade, or the words they use in their instructions.
A mastery goal structure encourages students to focus on individual progress. In this environment, the primary objective is skill development and deep comprehension. The standards for success are absolute or personal, meaning the student only compares their current knowledge to their past baseline.
Conversely, a performance goal structure pushes students to demonstrate their competence by outperforming others. The focus lies entirely on managing how other people perceive their abilities. Success is based on a normative curve, where doing well means beating classmates and avoiding unfavorable evaluations.
Past educational research shows that mastery environments tend to foster deep learning activities. Students are more likely to exhibit persistence and seek genuine comprehension. Performance environments often encourage shallow study habits, where learners prioritize the appearance of competence over actual understanding.
Today, artificial intelligence tools are rapidly entering mainstream education. Generative chat programs offer a highly personalized way to study, functioning almost like an on-demand private tutor. Yet, early research into these text generators shows wildly varying results.
Some studies indicate that using these programs improves academic achievement. Other studies reveal a drop in performance when students rely on the software.
Educational researcher Laura Schmidt at Ruhr University Bochum in Germany wanted to understand the reasons behind this variation. People might assume that a student’s technical skill with prompt engineering dictates their overall success. But Schmidt suspected that a student’s personal motivation might shape their technical approach to the software.
Along with colleagues Niklas Obergassel and Julian Roelle at the University of Münster, Schmidt designed a study to see if the traditional concept of goal structures still applied in an automated era. Because large language models respond directly to user instructions, a student’s internal motivations could immediately dictate the type of questions they ask during a study session.
To test this idea, the research team recruited 104 university students for a supervised online experiment. The participants were told to use ChatGPT to study four specific social psychology concepts. These included ideas like the mere exposure effect, which describes how people tend to develop a preference for things merely because they are familiar with them.
The students took a test to measure their existing knowledge of the social psychology concepts. Then, they were split into two slightly unequal groups. The first group received instructions designed to create a mastery goal structure.
These participants were told that their objective when using the software was to extend their personal knowledge. The researchers told the students to gain a deep understanding of the concepts, noting that their learning session would be considered a success if they felt they had learned a lot.
The second group was placed in a performance goal structure. The researchers told this group that their true objective was to shine in comparison to the other participants in the study.
The instructions explicitly stated that they should try to gain more and better knowledge than their peers. Their session would be considered a success if they performed better than everyone else on a final test. To make sure the students internalized these specific directions, they were all asked to paraphrase the instructions in writing before beginning.
Both groups then spent 20 minutes interacting with the text generating program to learn about the four concepts. The researchers reminded the students of their specific goal every five minutes. The participants were not allowed to take personal notes.
After the learning phase, the participants reported their intrinsic motivation and emotional state. They then took a final test consisting of 12 open-ended questions designed to measure both their conceptual knowledge and their deeper comprehension. The researchers also collected the chat transcripts and categorized every prompt the students typed into the software.
The results established that the way a task is framed drastically changes the way people interact with artificial intelligence. Students in the mastery group acquired noticeably more conceptual knowledge than those in the performance group. They provided better definitions and explanations of the core psychological concepts.
The chat logs revealed exactly how the performance group’s desire to appear smart sabotaged their learning process. Because these students wanted to outshine their peers, they asked the chatbot for trivial details that might sound impressive in a social setting.
They frequently requested the specific names of researchers who coined certain terms or the exact years that famous studies were published. They asked for the titles of related books and journals. While these facts might help a person sound knowledgeable in a casual conversation, they contributed very little to actual understanding of the core psychological principles.
During the final test, the performance group included these nonessential details in their written answers at an elevated rate. This phenomenon is known as criteria compliance, where learners try to fulfill external evaluation standards in a superficial way that neglects the actual learning objective.
In contrast, the mastery group used their time more efficiently. Rather than searching for obscure trivia, they asked the software for memorization aids to help them properly retain the material. They actively tried to build a structural understanding of the information.
Beyond the difference in knowledge acquisition, the performance goal instruction took an emotional toll on the users. Students in this group reported experiencing elevated levels of pressure and tension during the task.
As they worried about external evaluation and normative testing standards, the fear of potential failure made the learning process highly stressful. They experienced heightened overall anxiety. The mastery group, focusing only on their own intellectual progress, avoided this emotional burden.
The researchers also measured positive facets of intrinsic motivation, such as enjoyment, interest, and feelings of autonomy. Surprisingly, they found that the differences between the two groups on these positive measures were not statistically significant. The researchers noted that both groups displayed very high baseline levels of mastery orientation.
Many university students naturally want to learn, and these internal desires can coexist with external performance pressures. The performance instructions did not stamp out their native curiosity, but they did introduce negative pressures that hindered their progress.
Testing on deeper comprehension yielded similar results. The differences between the groups regarding deep comprehension tasks, which required students to apply the psychological concepts to novel scenarios, were not statistically significant. The researchers suspect that a 20-minute session was simply not enough time to achieve an advanced understanding. Learning how to properly apply complex psychological phenomena likely requires an extended study period.
Future investigations will need to address a few limits of this experiment. The participants likely had varied levels of experience with artificial intelligence. While most students used the program to request basic explanations, very few asked the software to generate practice quizzes or simulate an assessment. A learner’s prior technical skills could easily alter how they pursue different goals.
The researchers also point out that this specific experiment only tested declarative knowledge, which relies on understanding strict facts and concepts. Asking for peripheral details is an easy way to fake competence when studying basic text. However, this strategy might not translate well to learning problem-solving skills or advanced mathematics, where superficial details offer no advantage. Future studies could test different types of learning material to see if the effect persists.
Finally, the study team had no way to track how deeply the students thought about the text generated by the software. Without eye-tracking technology or think-aloud protocols, it is difficult to know if the mastery group actively read the text more thoroughly, or if their superior test scores resulted purely from the difference in their initial prompts.
Despite these unknowns, the takeaway for educators remains straightforward. Pushing students to compete with one another shifts their focus away from mastering the content. When teachers incorporate artificial intelligence into their classrooms, they should avoid framing tasks around peer comparison.
The study, “AIming High: Do Goal Structures Matter in Learning With ChatGPT?,” was authored by Laura Schmidt, Niklas Obergassel, and Julian Roelle.