What Makes AI Responsible? For iTELL, It Starts With the Science

By Wesley Morris, iTELL Chief Executive Officer

If you've spent any time in Career and Technical Education (CTE) circles recently, you've heard the unease about AI. Instructors are afraid of being replaced. Program directors worry that AI-generated content will be shallow, inaccurate, or simply wrong in ways that matter. Students worry that the feedback they are getting is generic and impersonal. Underneath all of it is a broader cultural anxiety, that the push toward AI is really a push toward cutting costs and cutting people, dressed up as innovation.

These concerns are not irrational. There are plenty of examples of AI being deployed in education in ways that deserve skepticism. There are tools that generate plausible-sounding content with no grounding in evidence, products that promise to transform learning while making the experience feel more impersonal, and tools that quietly shift decision-making away from instructors. When people push back on AI in education, they are pushing back on something real.

I want to make the case in this blog post that iTELL is different. From day one, we built iTELL to promote learning by centering teachers and learners. These are the specific design choices we've made that set iTELL apart.

The science comes first, not the technology

iTELL was built by education researchers at a university, not big tech. Co-founder Prof. Scott Crossley has spent his entire career studying how people learn from text, what helps, what doesn't, and why.

The core of the platform is grounded in specific, peer-reviewed findings from our own research. In a recent randomized controlled trial, we found that students who received personalized AI feedback on what they wrote learned significantly more than students who used a static digital text. More importantly, we found that simple interactivity without feedback wasn't enough. The thing that made the difference was the feedback itself: a response that engaged with what the specific learner actually wrote, not a generic hint or a pre-scripted answer.

That finding shapes everything about how iTELL is built. We are not adding AI to an existing product because AI is popular. We are building around formative feedback on learner-authored responses because the evidence says that this specific mechanism works. When we expand into new areas, we test whether the mechanism holds. When we're not sure, we say so.

The human is always in the driver's seat

Students

One of the most common objections to AI in education is that it makes learning passive, that students interact with a machine rather than engaging with ideas. iTELL is built on the opposite premise.

Every interaction in iTELL requires the learner to produce something. Not click a button, not select a multiple-choice answer, but write in their own words, from their own understanding, a response to the material they just read. The learning science term for this is the Generation Effect, the well-documented finding that information we generate ourselves is retained better than information we passively receive.

iTELL AI reads what the student wrote and responds to it specifically, providing personalized feedback at every stage of the reading experience at a scale no human instructor could sustain alone. This is not a machine testing a student. It is a machine creating a learning event where the student has to think, produce, and engage with feedback on their own thinking. The student is the author of their own understanding. iTELL's job is to create those opportunities and then get out of the way.

Instructors

In more than 250 interviews with CTE program directors that we conducted during our NSF I-Corps customer discovery, we heard the same problem again and again. The problem is not that instructors don't know their material, but that they can't be everywhere at once. A biotech instructor with thirty students can't give each of them individualized feedback on what they read last night before coming to class. That's not a failure of the instructor, it's a structural constraint on what any human being can do.

iTELL is designed to operate in that gap. When a learner reads a chapter of technical content and writes a summary, iTELL responds to what they wrote. When a student gets a question wrong, iTELL explains why, specifically, based on their answer. That feedback happens at two in the morning, on a phone, between shifts, at the moments when no instructor is available.

What iTELL does not do is tell an instructor how to teach, it does not replace the instructor's judgment about what a student needs, or make decisions about learner progress that belong to a human. The program director still designs the curriculum. The instructor still knows the students. iTELL handles the part that doesn't scale: individualized written feedback on every learner's own words, every time they engage with the material.

More than that, iTELL is a tool for curriculum designers to scale what they already know. A skilled instructor who has developed excellent training materials can now reach far more students with those materials, without those materials becoming less effective as enrollment grows. The expertise stays in the content. iTELL extends its reach.

Being honest about what we don't know

AI systems can be wrong. In education, wrong feedback is not a minor inconvenience. It can entrench a misconception, undermine a student's confidence, or, in safety-critical domains like electrical work or laboratory procedure, point someone in a genuinely dangerous direction.

We take this seriously and we are building iTELL accordingly. In partnership with Vanderbilt University, we are actively investigating where our AI scoring and feedback models perform reliably and where they don't, particularly as we expand from university content into the narrow, high-stakes technical domains of CTE. We are not assuming that our model generalizes correctly to every domain; we are testing it, measuring it, and being transparent about the results.

This is not a selling point we invented. It is a constraint we imposed on ourselves because deploying AI feedback in safety-critical vocational training without knowing whether it works, would be irresponsible. The research program exists because we don't yet have all the answers and we think the field deserves to know that.

What we are trying to build

While people are right to be skeptical about some of the wild and irresponsible promises from AI hype-people, we believe that AI can support learners and educators when it is packaged in a tool that centers the human at every stage. At Edumonia, we are building iTELL on a solid evidence base, being honest about its limitations, and designing a tool that supports learners all while remaining accountable to the educators who know better than we do (and better than AI does) how to help their students.

If that sounds like the kind of tool your program could use, we'd like to talk.