Designed for education and grounded in ethics, one AI tool shows how thoughtful design can improve academic outcomes without cutting corners.
GUEST COLUMN | by Kavitta Ghai
The world’s education systems are at a critical crossroads, with AI bringing both the biggest technological opportunities and risks to schools since the invention of the computer.
On one hand, it promises to transform how we teach and learn. On the other hand, if mishandled, we risk creating a generation of students who lose out on curiosity, independence, and the ability to think for themselves.
‘On one hand, it promises to transform how we teach and learn. On the other hand, if mishandled, we risk creating a generation of students who lose out on curiosity, independence, and the ability to think for themselves.’
These are stakes too high for us to sit idly. So we put it to the test.
Putting Purpose-Built AI to the Test
My company partnered with Los Angeles Pacific University (LAPU)—a fully online, asynchronous institution serving nearly 1,900 full-time working adults—to find real answers. Together, we ran one of the largest peer-reviewed field studies of AI in college courses to date.
Unlike other chatbots, our solution was designed as an AI tool strictly confined to course materials and engineered to guide, not complete, assignments. Here’s what that looks like in action.
First, we confine its knowledge to materials uploaded by faculty, no external internet access, and no off-platform data collection. Everything our AI assistant “knew” came straight from instructor-uploaded lectures, syllabi, and course materials. Second, our natural-language prompts are engineered to be Socratic, asking questions like “What formula applies here?” or “How might you approach this problem differently?” This nudges students to think through concepts step by step. This ensures we avoid spoon-feeding, encouraging students to use our AI assistant as a study aid, not a homework machine.
Behind the scenes, we layered in robust measurement capabilities.
Access Logs: Every student-AI assistant interaction is timestamped and visible to instructors, enabling early identification of misuse patterns.
Configurable Guardrails: Faculty can tighten or relax our AI’s hint levels or pause access entirely.
Privacy-First Architecture: Our tool is FERPA and SOC II compliant; no student data ever leaves the institution or trains third-party models.
Results That Matter—for Grades and Integrity
So did it work? We applied four distinct statistical methods across 99 class sections and nearly 2,100 course enrollments and found that, over a full term, students who used our AI-powered assistant three or more times saw their average GPA increased by 7.5%, even after controlling for age, gender, and ethnicity.
Instructors were now able to record and track more insights. Students reported deeper engagement thanks to our AI’s prompting style, helping students move from passive reading to active problem-solving.
Following LAPU’s results, community colleges and four-year institutions are integrating this tool into freshman success programs, remedial courses, and STEM pathways—where personalized, 24/7 support can make the greatest impact.
The Importance of Incentives and Recommendations to Stakeholders
The secret in edtech is that most consumer AI tools are built for user growth, not learning. Companies that want students as “users” instead of graduates build chatbots that pose risks around hallucinations, misinformation, and plagiarism—the kryptonite to effective learning.
If an edtech company exists to serve schools, its incentives are crystal clear—improve student outcomes, or disappear.
‘If an edtech company exists to serve schools, its incentives are crystal clear—improve student outcomes, or disappear.’
To ensure AI fulfills its educational promise, institutions should:
1. Vet for Pedagogical Design: Choose tools that confine AI to course content, prompt rather than provide, and align with learning objectives.
2. Implement Usage Safeguards: Use platforms offering detailed logs, configurable settings, and exam-mode controls to prevent misuse.
3. Prioritize Seamless Access: Integrate AI directly into the LMS, no extra apps or logins, to remove barriers for all students.
4. Monitor and Evaluate: Track both quantitative metrics (GPA, pass rates) and qualitative feedback (student and faculty surveys) to guide ongoing improvements.
The Bottom Line
AI is poised to become as ubiquitous as the textbook. But its value hinges on thoughtful design and rigorous evaluation. Our LAPU study proves that AI can boost academic achievement without sacrificing academic rigor. As AI tools proliferate, let’s insist on solutions built for classrooms, not clicks.
Let’s reinvent the way students learn—but never shortcut what makes real learning possible.
—
Kavitta Ghai is CEO & Co-Founder of Nectir, an AI-powered course assistant platform designed to provide real-time, personalized academic support within learning management systems. Connect with Kavitta through nectir.io.
0 Comments