Source-linked AI summary
Generative AI Expands the Intellectual Reach of Course Based Undergraduate Research Experiences (CUREs)
Aditi Babar, Kristin J. Davin, Alex Dornburg
TL;DR
CUREs need responsive support as students undertake increasingly complex authentic research, but whether GenAI strengthens inquiry, collaboration, and scientific reasoning remains unresolved. Using longitudinal qualitative data from a bioinformatics and genomics CURE across three semesters, the study finds that GenAI expanded novice students’ research reach, coordination, and responsibility while retaining human judgment. The findings are limited by the study’s small, self-selected honors-CURE sample and setting.
Problem
The study addresses how embedding GenAI across a CURE affects student inquiry, collaboration, and scientific reasoning while CUREs face growing demands for responsive instructional support.
Method
The authors analyze longitudinal qualitative data from student research in a bioinformatics CURE conducted across three semesters.
Results
GenAI expanded novice students’ research activity by providing individualized support, coordinating differentiated expertise, and enabling questions beyond instructor domains.
Takeaways & Limitations
GenAI can extend CURE support while students retain intellectual responsibility for evaluating and making research decisions.
Takeaways & Limitations
The findings’ generalizability and causal interpretation are constrained by a relatively small, self-selected sample from an honors CURE at one university.
Abstract
from arXiv · showhide
Course-based undergraduate research experiences (CUREs) broaden access to authentic scientific inquiry through responsive instructor support as research problems become increasingly complex. Generative artificial intelligence (GenAI) may extend this support by providing individualized assistance that can adapt as student needs change. However, how embedding GenAI within a CURE to provide support across the research process impacts student inquiry, collaboration, and scientific reasoning remains unresolved. Here we use longitudinal qualitative data collected across three semesters of a bioinformatics and genomics CURE to show that GenAI expanded the intellectual reach of the research experience in three distinct ways. First, personalized, on-demand scaffolding allowed students to move beyond the boundaries of instructor expertise and transform their own interests into researchable inquiry, with all teams developing distinct self-directed projects rather than selecting instructor-provided topics. Second, GenAI became part of the distributed cognitive system of research teams, helping novice researchers communicate and coordinate across differentiated expertise without eliminating specialization. Third, expanded capability did not replace the need for disciplinary judgment. Students increasingly validated, revised, or rejected AI-generated contributions, such that research independence emerged through retained intellectual responsibility. Together, these findings suggest that GenAI can extend the reach of CUREs by expanding what novice researchers can investigate, how they can collaborate, and the level of responsibility they can assume while preserving human judgment central to authentic scientific inquiry.
1. Introduction
CUREs broaden access to authentic scientific inquiry but require substantial, responsive instructional support as research problems and student populations grow. The paper examines whether GenAI can extend this support while preserving active intellectual engagement and research responsibility.
- Motivation: CUREs expand participation in authentic research but place substantial demands on instructor preparation, mentoring capacity, and institutional support.These demands become more consequential as CUREs reach larger student populations.
- Motivation: GenAI can provide dynamic, individualized support through immediate feedback, alternative explanations, and help refining ideas as students’ understanding develops.These capabilities may support question formulation, evidence interpretation, and progressively greater responsibility for scientific decisions.
- Motivation: Unstructured GenAI use can relocate intended cognitive work, encourage cognitive offloading and overreliance, and diminish retention and critical engagement.Unrestricted outsourcing of learning tasks may also impede later learning when access to GenAI is restricted.
- Research questions: The study asks how students use GenAI as personalized support, collaborative scaffolding, and guidance while developing increasingly independent research competencies.It specifically examines personalization, collaboration, and research competency development in CURE environments.
- Approach: Using longitudinal student data from a Bioinformatics CURE, the authors trace how GenAI use develops alongside participation in authentic research.Student artifacts, reflections, and interviews support a sociocultural analysis of GenAI as a mediational tool.
- Contribution: The findings propose that GenAI can extend CURE support and create alternative pathways for learning, collaboration, and research practice while preserving active intellectual engagement.The paper presents a learner-centered framework for understanding when and how GenAI supports research-skill development.
2. Theoretical Framework
The theoretical framework treats cognition as socially situated and distributed across people, tools, and environments. It positions GenAI as both individualized scaffolding and a mediational resource embedded in collaborative research.
- Sociocultural theory: Sociocultural theory conceptualizes cognition as socially situated and shaped through interaction with cultural tools and other participants.This framework emphasizes guided interaction rather than cognition occurring independently of social activity.
- Zone of proximal development: The zone of proximal development describes the difference between what learners accomplish independently and what becomes possible with appropriate mediation.Mediation is temporary and adaptive support that enables reasoning or problem-solving beyond current independent capabilities.
- GenAI as mediation: The framework extends this logic to GenAI as a mediator that can provide cognitive scaffolding when students remain actively engaged.GenAI is therefore analyzed as an individualized scaffold and a mediational resource within collaborative research activity.
- Distributed cognition: Cognition in CUREs is distributed across students, instructors, and other resources as research knowledge develops collaboratively.GenAI may reshape how ideas are generated, communicated, evaluated, and acted upon within research groups.
3. Literature Review
The literature frames CURE learning as dependent on sustained participation, high-quality guidance, collaboration, and students’ continued engagement with uncertain research. It identifies an unresolved tension between GenAI as scaffolding and GenAI as a source of cognitive offloading.
- Research learning: CURE participation can support scientific self-efficacy, scientific identity, research performance, and problem solving when students remain engaged with research challenges.Productive learning is not guaranteed and depends partly on guidance quality during unfamiliar research practices.
- Instructional support: CUREs face a trade-off between expanding access and providing the individualized professional guidance characteristic of apprenticeship research.Teaching assistants and peer mentors have been incorporated to broaden support for technical tasks, writing, feedback, and increasingly independent activity.
- Collaboration: Collaborative research requires teams to coordinate specialized work, exchange information, and contribute jointly to a shared scientific objective.For novice researchers, specialization may limit practice of unfamiliar skills and concentrate decision-making among perceived experts.
- Collaboration: GenAI may help address the tension between efficient team research and broad participation needed for research competency development.The extent to which students use GenAI to support both goals remains unknown.
- GenAI scaffolding: GenAI can support higher-order thinking, problem-solving, autonomy, formative feedback, and collaboration when students question, evaluate, and revise generated outputs.These benefits differ from use in which learners delegate the underlying cognitive work to the technology.
- GenAI risks: Overreliance on GenAI is associated with cognitive offloading, reduced critical engagement, and weaker independent reasoning when learners accept generated outputs without interrogation.This distinction is consequential in CUREs because independent research decision-making and scientific reasoning are target outcomes.
- Open question: The literature leaves unresolved how students move between supportive and outsourcing forms of GenAI engagement during authentic research experiences.The paper therefore focuses on how technological support can expand capability while preserving responsibility for interpretation and decision-making.
4. Methods
The study follows an honors bioinformatics and genomics CURE across three semesters using longitudinal qualitative data from student research activity. GenAI was embedded throughout an open-ended, gradually released research process while students retained discretion over its use.
- Study overview: The CURE engaged students in the complete research process, from question formulation and experimental design through data collection, analysis, interpretation, and dissemination.The course centered on the intersection of GenAI, bioinformatics, and genomics.
- Study overview: Students worked in small teams to develop original bioinformatics and genomics questions about GenAI and progressively develop their projects.Teams could choose instructor-developed topics or formulate projects of their own design aligned with bioinformatics or genomics.
- Instructional sequence: The instructional sequence gradually shifted responsibility for research decisions from instructors to student teams.Weekly instructor-guided discussions introduced practices for the next project stage, followed by independent team advancement through literature review, data work, analysis, interpretation, and presentation.
- GenAI implementation: GenAI was embedded across the research process, and students were not prescribed a required frequency, amount, or stage-specific restriction for its use.Students retained discretion over whether, when, and how they incorporated AI support.
- GenAI implementation: Instruction emphasized critical engagement by asking students to examine the strengths, limitations, and usefulness of AI outputs for their research goals.This created repeated opportunities to use GenAI while treating its use as an evolving component of research activity rather than a standardized treatment.
- Participants and setting: The dataset covered one honors laboratory section at a large public university, with 18 consenting students representing approximately 82% of honors-CURE enrollment.Participants ranged from freshmen through upper-level students and included STEM and non-STEM majors.
- Data and analysis: The qualitative study used team presentation recordings, student artifacts, reflections, interviews, and collaboratively developed interpretations across complementary researcher positions.Participation was voluntary, and course materials from nonparticipants were excluded from the research dataset.
5. Findings
Across three semesters, GenAI provided individualized support that made unfamiliar disciplinary and computational material more accessible and helped students develop self-directed research questions. It also supported team coordination across differentiated expertise while students retained responsibility for evaluating information and making research decisions.
- Personalized scaffolding: GenAI offered individualized, adaptable support that helped students understand unfamiliar disciplinary material and overcome computational barriers.Students described reshaping explanations to their knowledge level and using GenAI to implement analyses in unfamiliar coding environments.
- Student-directed inquiry: GenAI expanded students’ perceived research possibilities by generating alternatives and helping them investigate topics beyond their prior backgrounds.Students connected interests in exercise, disparities, epigenetics, and medical AI to projects they felt capable of pursuing.
- Student-directed inquiry: Students used GenAI to transform personal interests into researchable questions and developed distinct projects across diverse domains.Projects addressed areas including human health, exercise physiology, medical decision-making, forensics, environmental science, and emerging AI technologies.
- Research independence: Students increasingly treated GenAI as a tool requiring human evaluation rather than a replacement for learning or disciplinary judgment.Students reported framing questions, evaluating answers, deciding what to investigate next, and reconsidering how AI should be used.
- Collaborative research: Teams distributed tasks around differentiated expertise while using GenAI to mediate communication and bridge disciplinary gaps.Students described complementary roles such as software developer, researcher, and data collector, without specialization isolating members from the shared project.
6. Discussion
Across three semesters, GenAI expanded what novice researchers could investigate, how teams coordinated, and the responsibility students assumed, while disciplinary judgment remained essential. These benefits depended on students developing the knowledge and critical judgment needed to evaluate AI contributions, and their generalizability and causal interpretation remain constrained.
- Personalization: GenAI expanded novice students’ research activity through individualized, on-demand support that enabled conceptual and technical work beyond instructor expertise.Students used GenAI to address problems arising from their own projects, allowing instructors to focus on feedback and methodological or interpretive decisions.
- Personalization: All research teams used GenAI to pursue personally meaningful questions beyond the instructor’s research domains, creating learner-directed personalization.Students linked this freedom to greater investment and ownership as they transformed personal interests into researchable inquiry.
- Collaboration: GenAI supported interdisciplinary participation while students retained differentiated roles, so expanded access did not ensure that every learner practiced every research component.Some students entered unfamiliar areas or assisted peers, while others remained primarily within established roles.
- Collaboration: GenAI became part of teams’ distributed cognitive systems by helping members communicate across differentiated expertise without replacing individual specialization.Students divided work according to disciplinary backgrounds and skills while using GenAI to explain unfamiliar concepts and provide technical assistance where expertise was absent.
- Research competency: Research competency increased as students developed disciplinary knowledge and critical judgment to evaluate, revise, or reject AI-generated contributions.Later use included checking outputs against primary literature and testing AI-generated interpretations with independent statistical analyses.
- Research competency: Research independence emerged through intellectual responsibility: students retained authority over questions, evidence validation, and whether AI contributions entered the research process.Authentic disciplinary use was associated with students describing GenAI conditionally as an assistant, collaborator, peer, or accelerator while setting boundaries around those roles.
- Implications: The findings support pairing substantial GenAI access with instructional structures that preserve learner responsibility for explanation, validation, reflection, and decision-making.Expanded capability alone should not be treated as evidence of actual learning, because surrounding instructional design remains critical.
- Limitations: Interpretation is limited by the small, self-selected sample from one honors CURE and by students’ varying, uncontrolled patterns of GenAI use.The study cannot isolate which AI-engagement patterns produced particular learning outcomes or determine whether similar effects would occur under tighter controls.