Differential engagement with generative artificial intelligence in higher education: Gender, motivation, and achievement trajectories
Дифференцированное взаимодействие с генеративным искусственным интеллектом в высшем образовании: гендер, мотивация и траектории учебных достижений
2026-07-30
SCID: 54.1/jz2g9zs8
Discuss with AI
Autonomous motivationGenerative artificial intelligenceHigher educationRetrieval-Augmented GenerationZero-inflated negative binomial
Figures from the paper
Abstract (AI)
Abstract This study moves beyond asking whether generative AI (GenAI) improves learning to examine who engages, how they engage, and what kinds of questions they ask. We investigated how graduate students used a GenAI chatbot designed based on Retrieval-Augmented Generation (RAG) architecture to support statistics learning and how engagement patterns associated with learner characteristics, motivation, and academic growth. Drawing on 2,819 chatbot queries across a semester, motivational measures, and performance data from 97 students, we modeled engagement using a zero-inflated negative binomial (ZINB) framework and identified performance gain profiles through k -means clustering. More than one-quarter of students did not use the chatbot at all, highlighting meaningful non-participation. Engagement frequency was shaped by learner differences: female students, those with stronger autonomous motivation, lower prior knowledge, and higher course performance interacted more frequently, while higher-performing students were less likely to abstain from use. Importantly, engagement was not merely a matter of frequency but of inquiry quality and diversity. Growing Achievers (students who began with weaker achievement but demonstrated substantial performance gains) posed more diverse and conceptually oriented questions. In contrast, Declining Performers engaged minimally and focused on narrower, procedural inquiries despite stronger initial knowledge. These findings suggest that GenAI tools function differently across learner subpopulations and can either amplify growth or remain underutilized depending on motivational and achievement trajectories. The observed variability in student engagement profiles indicates that the efficacy of AI integration depends on moving beyond uniform technological access toward strategies that prioritize autonomous motivation and scaffold concept-focused AI support tailored to individual learner trajectories. Such supports are essential to ensure shared human-AI agency so that AI integration expands learning opportunities and promotes upward academic mobility instead of concentrating benefits among already advantaged learners.
Key Findings
1
Female students, autonomously motivated students, students with lower prior knowledge, and higher-performing students engaged more frequently with the chatbot.
2
GenAI effectiveness varied across learner subpopulations, supporting motivation-sensitive and trajectory-tailored scaffolding rather than uniform technology provision.
3
Growing Achievers asked more diverse and conceptually oriented questions, whereas Declining Performers used the chatbot minimally for narrower, procedural inquiries.
4
More than one-quarter of students did not use the GenAI chatbot, demonstrating substantial non-participation despite access.
5
The study analyzed 2,819 chatbot queries from 97 graduate students using a Retrieval-Augmented Generation chatbot for statistics learning.
Research Object
graduate students’ use of a Retrieval-Augmented Generation (RAG)-based generative AI chatbot for statistics learning
Research Subject
engagement patterns, inquiry quality and diversity, and their associations with gender, motivation, prior knowledge, and academic achievement trajectories
Publication Details
Publication Date
2026-07-30
Journal
Publisher
ISSN
Cited by
3
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai6
Three Approaches to Qualitative Content Analysis2005
The "What" and "Why" of Goal Pursuits: Human Needs and the Self-Determination of Behavior2000
Retrieval-Augmented Generation for Large Language Models: A Survey2023
Students' questions: a potential resource for teaching and learning science2008
Gender Differences in the Use of Generative Artificial Intelligence Chatbots in Higher Education: Characteristics and Consequences2024
Evaluating Retrieval Quality in Retrieval-Augmented Generation2024