Warrington students collaborate through AI Scholars Program

The University of Florida has positioned itself at the forefront of artificial intelligence (AI) in higher education through pioneering AI education and research. With AI tools being increasingly utilized in college classrooms, the question is no longer whether students use them, but whether they know how to use them well, and three Warrington researchers set out to find an answer.

Through the University Scholars Program, specifically the AI Scholars Program, Warrington students Derya Caglayan (BSBA ’27), Parker LaBruyere (BSBA ’26), and Surya Shekar (BSBA ’27) collaborated with Dr. Megan Mocko, a senior lecturer within Warrington and Dr. Beth Johson, a senior lecturer in the College of Liberal Arts and Sciences, on a study titled “Ask Better Questions: Developing Students’ Statistical Critical Thinking Skills Using Prompts to Large Language Models.” Their research explored how structured AI prompting instruction, when embedded into a flipped classroom model, can develop students’ statistical literacy, AI prompting sophistication, metacognitive skills, and sense of agency over their own learning.

The study was conducted across two statistics courses, QMB 5304 (Introduction to Managerial Statistics) and STA 3180 (Statistical Modeling). Through a series of “Ask Better Prompts” activities, students alternated between individual pre-class prompting sessions and collaborative in-class sessions, learning different prompting strategies such as role play and chain-of-thought. Data collected included the prompts themselves, AI-generated outputs, confidence ratings on a 1–5 scale, and written reflections. Researchers analyzed this data through qualitative coding, mosaic plots, contingency tables, and word cloud visualizations to track how students’ prompting sophistication and confidence evolved over time.

Derya Caglayan

Interested in how artificial intelligence can enhance university statistics education, Caglayan focused her end of the research on two key questions: how students use AI before class to learn statistical concepts, and whether their choice of statistical language reflects their level of understanding and prior coursework background. She specifically examined how frequently and accurately students used statistical terminology, such as regression, non-linearity, and heteroscedasticity, in their prompts.

Through qualitative coding and text analysis, Caglayan identified several key findings:

• Students with stronger prior exposure to statistics used more advanced and precise statistical terminology in their AI prompts.
• Engaging with AI before class helped students come better prepared for in-class discussion and application.
• AI prompting supported more active participation and improved the effectiveness of flipped learning environments.

“AI can strengthen flipped learning by encouraging students to engage with material earlier,” Caglayan said. “When students learn how to ask better, more thoughtful questions, they not only improve their understanding but also use AI more effectively as a tool for learning.”

Parker LaBruyere

Aspiring to understand how AI literacy could enhance rather than replace statistical learning, LaBruyere studied how structured prompting instruction and collaborative learning work mutually to develop students’ metacognitive skills. Specifically, he found that:

• Weekly AI users demonstrated the greatest flexibility in evolving their prompting approaches.
• The majority of students improved their prompting confidence ratings, with the strongest gains occurring when students collaborated with peers to refine their prompts.
• Students who actively took ownership of the prompting process used direct language (change, specific, adapt) and viewed AI as supplementary to their learning, while those who struggled used passive language (wish, hope) and expected AI to intuitively understand their needs.

From this research, LaBruyere identified a surprising persistence of passive learning behaviors even among dissatisfied students. Despite receiving instruction on advanced prompting techniques and expressing frustration with their AI outputs, some students remained reluctant to apply the strategies explicitly modeled to them in-class.

“What surprised me most was the stubbornness factor, as some students recognized their prompts weren’t working, felt dissatisfied with the results, and yet still didn’t apply the strategies explicitly taught to them,” LaBruyere said. “This gap between knowing better approaches exist and actually implementing them reveals that building AI literacy requires shifting students’ fundamental relationship with AI technology from passive consumption to active collaboration.”

Surya Shekar

As artificial intelligence becomes more common in education, students are increasingly using tools like ChatGPT, often without guidance on how to use them effectively. Shekar’s end of the research focused on shifting students from passively receiving answers to actively guiding their understanding through structured questioning. He tracked how confidence ratings and prompting structure evolved across the four survey pairs, specifically finding that:

• Prompts evolved from broad, input-output questions to structured instructional and chain-of-thought styles that actively guided AI reasoning.
• Students’ confidence increased across surveys, with a clear shift toward higher confidence ratings, as they gained control over how their prompts shaped AI output.
• Confidence became more meaningful over time, shifting from surface-level correctness to students’ ability to evaluate, refine, and align AI responses with their learning goals.

Word clouds visualizing shifts in student language, alongside survey-to-survey comparisons, highlighted changes in prompting structure, confidence, and metacognitive awareness.

“Artificial intelligence is only as effective as the way students choose to use it,” Shekar said. “What I found is that when students learn how to ask better, more intentional questions, they move from simply getting answers to actually understanding the reasoning behind them. That shift is what turns AI from a shortcut into a tool for learning.”