Evidence of Students Resorting to GenAI in an Adaptive Learning and Assessment System

• In Event: Virtuous or Viceful? How K-12 and Undergraduate Students Use and Perceive Generative AI

Abstract

This study examines how student learning behaviors and performance within the ALEKS adaptive learning platform have evolved following the widespread adoption of Generative Artificial Intelligence (GenAI) tools, with a particular focus on changes in learning efficiency and knowledge retention patterns. ALEKS, an acronym for Assessment and LEarning in Knowledge Spaces, is grounded in knowledge space theory. It personalizes students’ educational progress by identifying their mastery at a granular level and guiding them toward concepts they are ready to learn.

Using a mixed-methods approach, the study analyzes longitudinal data by comparing pre- and post-GenAI adoption periods through quantitative and comparative analysis of learning metrics. The research draws from a comprehensive 10-year dataset encompassing several million learning occasions from ALEKS math courses spanning Middle School through College Algebra levels, supplemented by ALEKS Math Placement test data comparing proctored versus non-proctored exam performance in higher education settings.

Key findings reveal that, coinciding with the widespread adoption of GenAI tools, students demonstrate significantly faster learning times for text-based math problem types. For High School and College Algebra courses, the average reduction in learning duration exceeds 20% compared to the pre-GenAI adoption baseline. Conversely, problem types requiring graphing and plotting tools show minimal variation in learning duration and retention, suggesting differential impacts of GenAI across problem formats. When problem types are classified based on changes in learning time, results indicate that the acceleration in learning time is accompanied by a decline in retention rates during subsequent knowledge checks. Problems in the top 10th percentile for learning time acceleration exhibit an average retention rate drop exceeding 5%, while the retention rate change for the remaining problems is under 0.5%.

Additionally, comparing students' response times on ALEKS Placement exams in proctored and non-proctored environments shows that proctored environments mitigate these effects. In non-proctored environments, response times for correct answers are lower by an average of 5% in the post-GenAI era across all problem types, while there is a negligible increase in response times for correct answers in proctored environments.

This research provides empirical evidence of GenAI’s growing influence on digital learning platforms and identifies concerning patterns in student learning outcomes. The findings indicate that while students complete learning tasks more quickly, they retain less knowledge, suggesting that widespread GenAI use may create a disconnect between apparent progress and actual learning. As GenAI tools become increasingly integrated into students’ academic practices, this study highlights the need for educational institutions to reconsider their GenAI access policies, promote responsible AI usage, and adopt instructional approaches that align with the realities of rapidly advancing technology. These findings inform educational policy and practice, particularly regarding the interplay of ethics, technological integration, and authentic learning.

Authors

• Hasan Uzun, McGraw-Hill

• Jeffrey Matayoshi, McGraw-Hill

• Eric Cosyn, McGraw-Hill

• Eyad Kurd-Misto, McGraw-Hill

• Sina Rismanchian, University of California - Irvine