Every university improves courses after students encounter challenges. Faculty review student feedback, analyze performance and revise learning experiences after barriers become apparent.
But what if institutions could identify some of those barriers before the first student ever opens a course?
As higher education explores artificial intelligence, much of the conversation has focused on generating content, improving efficiency and supporting students after they enroll. Another opportunity deserves attention: using AI to help institutions evaluate the quality of learning experiences before students encounter them.
AI doesn’t replace human judgement
Every course contains barriers its authors never intended and may not know exist. A lesson may use language that is clear to a subject-matter expert but confusing to a student encountering the topic for the first time.
Instructions may rely on cultural assumptions. Concepts may appear in an order that makes sense to the author but not to someone learning the material.
Traditionally, institutions find many of these problems only after students encounter them.
At Western Governors University, we are exploring whether synthetic learning agents can move some of that feedback earlier.
These are not digital replicas of individual students. They are specialized AI agents designed to evaluate specific dimensions of a learning experience.
One might examine language for potential comprehension barriers. Another might examine cultural assumptions. Others could evaluate accessibility, emotional friction or whether concepts appear in a logical sequence.
An orchestration system can determine which agents are relevant to particular content and bring their findings together for human review.
Consider a seemingly supportive instruction: “If you’re struggling with this material, reach out to your instructor or a trusted family member for help.”
A human reviewer may see nothing wrong with it. Different agents might identify different issues. “Reach out” is an idiom that may create unnecessary language friction. “Trusted family member” assumes a student has that relationship available. “Struggling” may unintentionally frame seeking support as evidence of failure.
None necessarily means the sentence is wrong. They are signals for a human course designer to consider.
The same approach can extend beyond individual sentences. A course can be clearly written and still have structural problems.
Students might be asked to apply a concept in one module that is not explained until the next. An AI agent designed to examine the relationships among concepts could flag that sequence before students encounter it.
This matters because AI should not become a substitute for human judgment in course design. It can give educators another set of tools for applying that judgment.
Another layer of quality assurance
At WGU, human experts remain responsible for deciding whether an identified issue requires a change. Our work also includes ethical review intended to identify bias and other unintended consequences as these systems are developed.
Privacy is equally important. The goal is not to create digital copies of individual students or expose personal student information every time a course is evaluated. Synthetic agents can instead represent specific dimensions of how learners may encounter content without functioning as replicas of actual people.
The technology is still developing. We have individual agents are reviewing course content and producing structured findings, while we continue working on questions such as accuracy, useful signal versus noise, scalability and how findings ultimately correlate with real student outcomes.
Those limitations are exactly why higher education should explore this use of AI carefully rather than wait for the technology to mature somewhere else.
Institutions interested in this approach can begin small. Identify one recurring course-design challenge, such as confusing language or concept sequencing.
Test whether AI can reliably surface potential barriers. Have faculty and instructional experts evaluate every finding. Then compare those findings with what real student outcomes tell you.
The larger opportunity is to change when quality assurance happens.
Higher education will always learn from students. Their experiences and outcomes remain the ultimate measure of whether our courses work.
But students should not have to encounter every avoidable barrier before institutions know it exists.
AI gives higher education an opportunity to identify some of those problems earlier. Used responsibly, synthetic learning agents could become another layer of quality assurance, helping educators test, question and improve learning experiences before students ever log in.
