It is found that there is no significant differential effect of GenAI availability on grades overall or among previously lower-performing students, and the findings temper concerns that GenAI inflates grades and reduces students's satisfaction.
Abstract
The spread of generative AI (GenAI) in higher education has raised concerns that students offload cognitive effort to AI, earning high grades without learning. If this"GenAI substitution hypothesis"is true, grades should rise disproportionately in GenAI-susceptible courses--those relying more on assessments like take-home problem sets and essays rather than in-class exams. Substitution could also affect student satisfaction, measured here as self-reported understanding and interest in the subject, which prior research links to assessments. We test the substitution hypothesis using syllabus and administrative data from a large U.S. university (2016-2025; 138,386 students; 72,730 course offerings). We measure courses'GenAI susceptibility using a human-validated LLM pipeline to extract assessment types from syllabi, and use a differences-in-differences design comparing outcomes across courses before and after ChatGPT's release, while modeling COVID-19 pandemic effects as either persistent or transient. We find no significant differential effect of GenAI availability on grades overall or among previously lower-performing students. Effects on self-reported understanding are likewise insignificant; effects on interest are significant only assuming transient pandemic effects. Our findings temper concerns that GenAI inflates grades and reduces students'satisfaction.
We study how generative AI affects student learning in a randomized experiment. In proctored, in-person sessions, undergraduates learn about an unfamiliar topic and write an analytical essay with or without access to off-the-shelf generative AI, then complete unaided assessments immediately and one week later. We measure learning with knowledge tests (factual and conceptual understanding) and open-ended essays (higher-order skills). AI access raises immediate test scores by 0.27 standard deviations. These gains persist one week later. Essay quality, by contrast, changes little while students have AI access but improves in style and relevance one week later, when students write unaided. These delayed gains are larger among augmentation users-who use AI to explain concepts rather than generate text-whereas automation users'short-run quality gains vanish once AI is removed. We find evidence for two mechanisms behind the learning gains: students shift time away from drafting text and toward reading and searching for information, and they report greater learning enjoyment.
Students may learn at different paces due to differences in prior programming experience (PPE), physical or mental health challenges, and economic or lifestyle barriers (e.g., employment or caregiving responsibilities). Additionally, increasing use of AI tools for homework can contribute to inaccurate self-assessment and poor preparation for supervised tests. Motivated by these challenges, we introduced bi-weekly low-stakes checkpoint quizzes in a large (~500-student) introductory CS course. Inspired by alternative grading paradigms such as mastery learning (ML), each quiz could be attempted multiple times without penalty, offering students frequent feedback and opportunities for iterative improvement. Our mixed-methods study investigated the impact of these quizzes on performance, self-assessment, stress levels, and overall experience, using performance and survey data (N=456). Results showed that though retake opportunities allowed students to improve quiz performance, frequent retake attempts were associated with lower final exam outcomes, suggesting continued struggle on novel problems. Despite limited performance benefits, survey data revealed strong affective outcomes based on overwhelmingly positive student sentiment: students reported high Likert-scale ratings for learning/engagement and stress reduction value (though subgroup differences by gender, PPE, English fluency, and retake frequency suggest room to improve equity outcomes), and the majority of open-ended responses described the quizzes as helpful for improving self-assessment, reducing stress, and supporting meaningful learning. Overall, our implementation allowed students to experience some benefits of ML while retaining enough structure to prevent procrastination, illustrating how ML?inspired assessment can be incorporated into courses without a full course redesign.
Sadia Sharmin, Paul He· Annual Conference on Innovat...· 0 citations
The rapid diffusion of artificial intelligence (AI) tools is fundamentally reshaping learning practices in higher education, particularly in mathematics and statistics courses. This study investigates how undergraduate students across multiple programs at IU International University engage with AI-based mathematics tools, which applications they prefer, and how they evaluate their performance. Drawing on survey data from 174 students enrolled in mathematics lectures during the spring term of 2025, we analyze awareness, usage intensity, and perceived usability, understandability, correctness, and value for money of leading AI tools. The results reveal a highly concentrated market structure: ChatGPT, Photomath, and Gemini dominate student awareness and usage. While ChatGPT is perceived as the most user-friendly and offers strong value for money, Photomath receives the highest ratings for correctness of results. Gemini, in contrast, is evaluated more cautiously across dimensions. Differences in awareness between math- intensive and non-math-intensive programs are small and statistically insignificant, suggesting that AI adoption in mathematics is broadly distributed across disciplines. The findings indicate that students integrate AI tools primarily as complementary learning aids rather than replacements for traditional materials. Overall, the study provides an empirical baseline for understanding how AI functions as a study partner in mathematics education and highlights the growing importance of evaluating not only usage frequency but also perceived reliability and pedagogical value.
M. Altin, Kirsten Jäger, Silke Jütte et al.· IU Discussion Papers Busines...· 0 citations
Maths support has long been present in Australian tertiary education, primarily as a physical space staffed by tutors and complemented by asynchronous resources and accessible via integration into course curricula. This format has proved effective in improving student retention and transition (Horrocks et al., 2026) and will only increase in salience as student cohorts diversify. Artificial intelligence (AI) has shown potential for providing low-stakes mathematics and numeracy support that gives immediate, personalised feedback and clarification to students. A trial of generative AI (GenAI) agents as a first point of contact for mathematics support for nursing and education students at Federation University Australia found that students accessed support outside of office hours and during semester periods of highest need, indicating the potential of AI as a just-in-time learning tool. Furthermore, the qualitative data from the agent conversations showed that the agents were being utilised overwhelmingly for learning rather than cognitive offloading.
Luis Camacho, Christopher Bridge, Birgit Loch· Student Success· 0 citations
Generative Artificial Intelligence (GenAI) tools are increasingly used by computing students, yet their effects on learning outcomes remain mixed. Prior work found that while GenAI use may improve performance on assignments, it can negatively relate to overall course performance. We aim to replicate and extend this work across four computing courses. Using self-reported GenAI usage from assignments and study preferences alongside course performance data, we examine how these relationships vary by course, and compared to the previous study. Our results show that students who used GenAI tools to solve the assignment performed equally or better than those who did not report using it, however, they received lower final grades in the course. We observe no major difference between students who used GenAI to study for the midterm test compared to those who did not. These findings suggest that the impact of GenAI use is present in various contexts, highlighting the need for instructional guidance on how students should use GenAI as a learning aid, and insights for other instructors that wish to integrate GenAI tools into computing curricula.
Valeria Ramirez Osorio, Ido Ben Haim, Ahmed Ashraf et al.· Annual Conference on Innovat...· 1 citation