Researchers propose human-centered AI model for online student support

Two University of Phoenix College of Doctoral Studies scholars have proposed a 16-stage framework for using artificial intelligence to support online students without removing educators from key decisions. The model matters as colleges weigh how to use generative AI and predictive analytics in ways that are timely, personalized and subject to human oversight.

Pamayla E. Darbyshire and Carl Beitsayadeh published the conceptual paper, “Enhancing Student Success through GAI and Predictive Analytics,” in the International Journal for Educational Media and Technology. According to the University of Phoenix, the journal issue included refereed papers first presented at the 2025 Teaching, Colleges, and Community Worldwide Online Conference.

The paper argues that institutions often assess predictive analytics and generative AI as separate tools. Darbyshire and Beitsayadeh propose linking them in a closed-loop support system for online higher education, with predictive insights, generative feedback, faculty judgment and institutional governance operating together.

Darbyshire said AI in education should start with the learner’s experience. She said the framework is intended to identify student needs earlier while keeping faculty judgment and ethical review central to the process.

“For online learners, timely support matters,” Darbyshire said. “The goal is not to replace the human relationship in learning, but to help educators respond with greater context, clarity and care.”

How the proposed system works

According to the paper, the framework draws on systems theory and the learning analytics cycle. It describes a continuing process that includes data intake, predictive modeling, AI-generated feedback, instructor review, monitoring and institutional refinement.

The authors say institutional systems such as student information systems, learning management systems and analytics platforms can reveal patterns that suggest a learner may need support. Predictive models could flag signals including lower engagement, late work or falling performance, then classify those signals into risk tiers.

Generative AI could then help create targeted support, the authors write. Examples in the paper include personalized messages, formative quizzes, study plans and recommendations for learning resources.

The framework does not place those AI outputs on autopilot. Darbyshire and Beitsayadeh say faculty should review and contextualize recommendations so student support reflects both data and the learner’s circumstances.

Beitsayadeh said current discussion often treats predictive analytics and generative AI as separate technologies. He said the proposed model combines them in one adaptive system, where data-based signals, AI-enabled support, faculty judgment and institutional oversight form a continuous improvement cycle.

Governance and safeguards

The paper says responsible use of AI in online higher education requires more than buying or deploying new tools. Darbyshire and Beitsayadeh identify secure and interoperable data systems, clear rules for data access, audit trails, faculty development and governance as necessary parts of implementation.

The authors also emphasize ethical safeguards. According to the paper, institutions should address transparency, fairness, bias monitoring, student trust and the continued use of human discretion.

The paper lists several practical steps for institutions considering AI-supported student services:

• Set policies for AI use, data access and model evaluation.
• Train faculty to interpret predictive analytics and AI-generated recommendations.
• Monitor systems for bias, fairness and unintended effects.
• Design tools that support instructor judgment rather than replace it.
• Create feedback loops to refine interventions over time.

Darbyshire and Beitsayadeh present the framework as a planning model rather than a report on a deployed system. Their central recommendation is that AI tools for online learners should be tied to educator review, institutional policy and ongoing evaluation.

This story draws on original reporting from Phys.org.