Issue 03
Members

AI as Intervention: What Every Student Success Leader Needs to Know Now

AI in higher ed is being deployed as an efficiency tool. It should be deployed as an intervention tool. These are not the same thing.
The Big Idea

AI cannot replace the human relationship that keeps a student enrolled. It can free up the humans who build it.

Social Psychology Foundation

There is a finding in social psychology that consistently surprises people who have not read the research: the quality of a single relationship with one institutional representative is among the strongest predictors of student persistence. One faculty member who knows a student's name. One advisor who remembers what was discussed last semester. One residence hall director who noticed the student seemed off.

This is not a feel-good finding. It is a structural one. Humans are social animals, and our institutions — despite their best intentions — are often structured in ways that make genuine relationship formation between students and institutional agents unnecessarily difficult. Advisors carry caseloads of 400 or more. Faculty teach three courses of 30 students each. The ratio is wrong, and AI is not going to fix that by itself.

What AI can do is reduce the administrative burden on the people who have the capacity to build those relationships, freeing advisor time for the conversations that matter instead of the paperwork that does not.

"The question is not whether to use AI. The question is whether you are using AI to replace human connection or to protect the time available for it. The answer determines whether your implementation helps students or quietly hurts them."

✦ Free Preview Ends Here
The full analysis, frameworks, practitioner tools, and recommended resource are available to members.

The institutions that retain the most students are not the ones with the most sophisticated data systems. They are the ones that have built cultures where advisors have time to notice, capacity to act, and frameworks to guide their response...

This Content is for Members

Subscribe to The Praxis Brief for full access to every issue, the complete archive, practitioner toolkits, and monthly deep-dive PDFs.

Every issue
Full archive
Monthly PDF
Practitioner tools
Premium content — you are reading the full issue as a member.
Higher Ed Application

The AI Implementation Decision Framework: Before You Buy Anything

Every AI vendor in higher ed will tell you their product improves outcomes. Before any purchase decision, run every proposed tool through these four questions:

  • 1
    Does this tool reduce advisor administrative time or does it replace advisor contact?Administrative time reduction is a net positive. Contact replacement is a net negative and should require extraordinary evidence before adoption.
  • 2
    Does the risk score generate a human conversation or an automated message?Automated outreach triggered by algorithmic risk scores has consistently underperformed human-initiated outreach in controlled studies. If the AI produces an alert that results in a form letter rather than a phone call, the intervention value is near zero.
  • 3
    Does the tool explain its recommendations or only produce them?A system that says "Student X has a 68% departure risk" is less useful than one that says "Student X has had no advisor contact in 34 days and missed 3 assignments in the past two weeks."
  • 4
    Who are the false positives and what happens to them?Algorithms trained on historical data encode historical inequities. Students from under resourced high schools, first generation students, and students of color are frequently over flagged — and over-flagging leads to intrusive outreach that can itself signal to students that the institution views them as deficient.
3 AI Tools with Actual Evidence
  • 1
    Intelligent Tutoring Systems for foundational coursesPlatforms like Carnegie Learning and Realizeit have robust evidence bases showing 15 to 25% improvement in gateway course pass rates when implemented with fidelity. The mechanism is mastery-based progression.
  • 2
    AI assisted financial aid communicationChatbot systems that help students understand their financial aid packages in plain language have shown measurable impact on FAFSA renewal rates and award acceptance. The intervention is not sophisticated — it removes friction from a high stakes, high-anxiety process.
  • 3
    Predictive scheduling and enrollment guidanceSystems that flag students enrolled in course sequences with historically low completion rates and proactively suggest alternatives reduce the invisible wall problem where students fail predictable failure-point courses without warning.
Practitioner Tool

AI Policy Review Checklist

  • We have reviewed the vendor's equity audit and understand which demographic groups are over or under flagged
  • The tool's outputs generate human contact, not automated messages, as the primary intervention pathway
  • We have defined what success looks like for this implementation and will measure it independently of vendor provided metrics
  • Front-line advisors were involved in the procurement process — not just informed after the decision was made
  • We have a sunset clause or annual review requirement built into the contract
  • Students know they are being monitored and have a pathway to contest or opt out
  • The implementation does not reduce total advisor to student contact hours
EDUCAUSE (2024). AI in Higher Education: Perspectives and Practices
educause.edu · Free download for members
The annual EDUCAUSE AI report is the most comprehensive snapshot of how institutions are actually deploying AI — not how vendors say they are deploying it. The gap between these two things is significant and instructive.

Next issue: the one thing social identity theory tells us about student experience that your data dashboard will never show.

— Dr. Corey Sims