Research
Many school leaders are being told AI is the answer. Dr. Karen Abrams tested that promise with students in Guyana in a randomized controlled trial. We share what it found so you can make better decisions for your school.
The question
The study took place in STEMGuyana’s Learning Pods, a network of community after-school centers supported by the Inter-American Development Bank’s IDB Lab. The pods serve learners in all ten regions of Guyana, many in communities without internet access at home.
Ten pods were randomly assigned to one of two versions of the same platform. Both groups had the same Ministry-aligned lessons and the same in-person facilitators. The AI version added an AI tutor, AI-generated practice quizzes, mastery-based progression and text-to-speech. The AI features were the only difference between the groups.
What we found
After eight weeks, learners on the AI-enhanced platform did not show statistically significant gains in Mathematics or English over learners on the non-AI version. The result held after accounting for differences between pods.
Total time logged on the platform did not predict learning gains. Where access is interrupted, logged minutes can reflect connectivity and circumstance as much as real learning.
How much students used the AI tutor, AI quizzes or text-to-speech did not predict gains. Heavier use may point to a student who needs more help, not one who is getting ahead.
“The future of this work lies not in asking whether AI works in the abstract, but in asking under what conditions, for which learners, and with what forms of support AI can contribute meaningfully to learning.”
Abrams, K. (2026), Chapter 6
From findings to training
The trial showed that giving students AI features did not, on its own, improve learning. So our program does not start with a tool. It starts with the adults around students. Leaders set purpose and guardrails before anything is adopted. Teachers earn certification by demonstrating understanding on knowledge checks, not by logging hours, because the trial showed that time on a platform is a poor measure of learning. Students meet AI through teacher-led lessons, because the study concluded that AI contributes only when human support surrounds it.
For school leaders
These come directly from the study’s findings. Ask them of any vendor, including us.
AI features added without a change in human support did not raise outcomes. The study concludes that AI works best inside clear routines with human facilitation.
Access to a tool does not mean it gets used well. Facilitators need training in when and how to use AI so it strengthens learning instead of fragmenting it.
Time on the platform did not predict gains. Dashboards full of logins and minutes can look like success without showing whether students learned.
Students who are still building foundational reading and math skills may need stronger support before AI tools help. What suits a high-school student may not suit a third grader.
Ask for independent studies with a comparison group, not only testimonials. Short pilots can miss effects in either direction, so plan to evaluate over time.
How the research shapes our work
Understand AI well enough to set policy, evaluate products with clear eyes and build a realistic roadmap.
Learn what AI can and cannot do, practice responsible classroom use and earn certification before leading lessons.
Teacher-guided, age-appropriate lessons on how AI works, where it goes wrong and when to question it.
Plain-language learning on AI at home, online safety and supporting homework, when your school is ready.
Why the evidence is credible
The intervention ran for eight weeks across a wide grade range, and engagement was measured through platform logs. Longer use, a narrower age group, or measures of confidence and persistence might show effects this study could not detect. We state these limits because schools deserve an honest account of the evidence.
Beyond the study
Through STEM Guyana’s live online AI for Teachers workshops, educators learned how AI works, where it helps and where it falls short, how to use it in lesson planning, and how to handle ethics and age-appropriate use. Certificates were earned by passing an assessment, not by attendance alone.
As Director of STEMGuyana, Dr. Karen Abrams also led the Learning Pods program across all ten regions of Guyana.
“The workshop enlightened me on how to use AI effectively to plan my lessons and create worksheets and posters — while making sure the information is accurate and useful. I encourage other teachers to try it to make their workload lighter. I am so grateful.”
“The STEM Guyana AI for Teachers workshop helped me appreciate the power of using Artificial Intelligence in a meaningful way — in teaching, lesson planning, and innovative pedagogy. It was an enriching experience during my doctoral research on innovative strategies to enhance STEAM education, helping me understand the role technology plays in supporting primary teachers to provide meaningful, engaging, and creative learning opportunities. For teachers who are unsure about AI, I would encourage them to study and adopt it in a competent and responsible manner — and see it as a tool that can help them improve upon effective pedagogy, rather than replace them.”
Cite this work
Abrams, K. (2026). Evaluating the impact of AI-driven tutoring on academic outcomes and engagement in STEMGuyana’s learning pods [Doctoral dissertation, University of Florida].
Dr. Karen Abrams, Ed.D., Educational Technology, University of Florida · Committee chair: Anthony Botelho