AI in Higher Education Newsletter

June 20, 2026 · Vol. 21

A weekly brief for the Management Department, Isenberg School of Management, UMass Amherst. By Matthew D. Langenkamp / 雷邁德, prepared in collaboration with Thea 🪻✨.

← Back to archive · ← Back to Dictionary

Executive Summary

This week’s central higher-ed AI story is assessment – not cheating in the narrow sense, and not another round of “students are using ChatGPT,” but whether our assessments still produce credible evidence of student learning.

EDUCAUSE’s new Impact of AI on Learning Assessment report, its June assessment-design symposium, Indiana Kelley’s AI Playbook, and fresh student-survey data all point the same way: detection is losing standing as the answer, and assessment redesign is the stronger institutional response. That means assignments where students disclose AI use, show process, defend reasoning, and demonstrate judgment rather than merely submit polished artifacts.

For Isenberg this is an opportunity as much as an integrity problem. Our memos, decks, analyses, and case work remain valuable – but they now need a second layer of evidence: can the student explain, defend, revise, and take responsibility for the recommendation?

Overview

The useful shift this week is from AI policing to learning evidence. For two years the faculty conversation centered on detection – which detector works, how to catch misconduct without falsely accusing honest students. Those questions are understandable but no longer sufficient. If a final product can be generated, polished, translated, summarized, and rhetorically improved by AI, the product alone carries less evidentiary weight. A polished memo may still be useful; it simply proves less than it used to.

The better question is what kind of evidence would convince us a student actually learned something: process notes, drafts, revision history, source-checking, AI-use documentation, short oral defenses, in-class explanation, targeted follow-up. None of this requires panic. It requires designing assignments with the evidence problem in mind.

1. Assessment Validity Is Becoming the Core AI Issue

EDUCAUSE’s Impact of AI on Learning Assessment (June 1; a 2026 survey of 438 faculty and staff) makes a practical point: AI is reshaping assessment fast, and faculty are responding unevenly. Many use AI to build assessments; many believe students use AI to complete them; and most want discretion over when AI is appropriate. That last point matters – neither universal permission nor universal prohibition will do. A finance model, a strategy memo, a written reflection, an oral defense, and an in-class exam are different kinds of evidence and should not share one AI rule.

EDUCAUSE’s June 9 and 11 symposium pushed the same line – skeptical of surveillance-first responses, focused on redesigning assignments so learning goals, feedback, and responsible-use policies fit together. A workable design sequence: What outcome am I assessing? Would AI use invalidate, strengthen, or merely change the evidence? What process evidence should accompany the product? What follow-up question would reveal whether the student owns the work? That beats starting with “How do I stop AI?”

New this week - student data: The Lumina Foundation-Gallup 2026 State of Higher Education study (3,800+ students) finds AI use now routine – roughly two-thirds use it weekly or daily for help with coursework – even though many students say their institution discourages or prohibits it. The gap between student practice and institutional policy is itself the thing to manage.

2. Kelley Gives Business Schools a Usable Assessment Formula

Indiana University’s Kelley School of Business has been unusually direct. Its faculty AI Playbook – a living guide, recently updated and amplified this month by Tom’s Guide, though the primary source is Kelley’s own document – states that AI detection tools are not approved for use at Kelley or IU, naming GPTZero, Turnitin AI Detection, and Originality.AI. The rationale: the tools are highly unreliable, produce both false positives and false negatives, raise privacy concerns, and are especially shaky on short-form writing and for multilingual students.

The useful part is what Kelley recommends instead – assignments that are traceable, defensible, and answerable:

This is not merely an integrity strategy; it is a professional skill. In consulting, finance, operations, marketing, entrepreneurship, and general management, a polished deliverable is rarely enough – a manager has to defend assumptions, answer objections, explain trade-offs, and stay accountable when the recommendation is challenged.

The implication for our courses is not that every assignment becomes an oral exam. It is that major AI-vulnerable assignments should carry at least one accountability mechanism – a five-minute follow-up conversation, a short in-class explanation, a process memo, a version history, a targeted reflection. That single move shifts the assessment from “produce a plausible artifact” to “show that you understand and own it.”

3. The Question Is No Longer Whether AI Is Coming to Campus

EDUCAUSE Review’s June 16 “current state of play” piece treats AI as a campus-wide transition – teaching, research, advising, administration, policy, the student experience – and warns that many institutions are still acting tactically: a pilot here, a policy statement there, a training session somewhere else. The result is fragmented practice and mixed signals for faculty and students alike.

For a business school this is an opening. We already teach the vocabulary needed here: governance, process design, incentives, risk, accountability, adoption, organizational capability. The classroom AI problem is a live case in management education. Isenberg does not need a grand theory before improving practice – just a shared operating grammar: assignment-level AI labels; disclosure treated as documentation, not confession; process evidence for AI-vulnerable work; a defense moment for high-value recommendations; privacy and vendor caution before requiring tools; faculty discretion inside a common departmental language.

4. The Economics Are Starting to Matter

The Chronicle’s June 16 piece on whether colleges can afford AI is a useful corrective to the easy “AI-first” marketing. At scale, AI brings recurring inference costs, vendor dependence, training needs, data-governance and accessibility obligations, cybersecurity exposure, and support burden. That shapes pedagogy: if one course requires a paid tool, another recommends an institutional chatbot, and a third assumes students have premium frontier models, the department can accidentally build inequity into its assignments.

Tool access has to be designed, not assumed. Before assigning AI work, a faculty member should be able to answer: Is the tool available to every student on reasonable terms? What student data enter the system? What happens if it changes, fails, or becomes paid? Does the assignment measure judgment or tool access? Is there a no-cost path that still meets the objective? The cost question is pedagogical, not just budgetary – a business school that teaches strategy should be wary of confusing a vendor stack with a capability.

5. Faculty Need Infrastructure, Not Just Encouragement

Two Chronicle pieces converge on faculty development. One has professors “walking a tightrope” as they decide how to use AI without weakening learning; the other floats an “AI librarian” – someone fluent in tools, sources, information literacy, citation norms, and privacy who can help faculty and students navigate the terrain. The title may not stick, but the function is real, and librarians and instructional designers are often closer to that work than the central IT office.

AI has blurred old categories. Brainstorming with AI is like tutoring; summarizing sources is information work; drafting a memo is communication work; critiquing a recommendation is managerial judgment – if the student stays in charge. Those uses need different norms, and the most useful support is close to the course: assignment review, sample language, tool comparisons, privacy guidance, a small repertoire of AI-aware designs. The most important support may be social rather than technical – credible local examples from colleagues who tried things, found the failure points, and adjusted.

6. Higher Education’s Human Advantage Has to Become Explicit

Inside Higher Ed’s June 18 essay argues that as AI grows more capable, the human dimensions of education – trust, mentorship, judgment, feedback, belonging, responsibility – matter more, not less. They are part of the educational product, not decoration around it. That connects directly to assessment: if a student can produce a polished answer with AI, the faculty role shifts toward helping the student understand, defend, revise, and own it.

Business schools should be especially alert. Our graduates will enter workplaces where AI generates memos, decks, code, market summaries, and strategy language. The scarce capability is less “can you produce text?” and more: Can you frame the right problem? Can you tell when an answer is plausible but wrong? Can you verify claims under time pressure? Can you explain the trade-offs to other people? Can you take responsibility when the recommendation matters? That is the human advantage worth naming – not anti-AI, but AI under disciplined human supervision.

7. Practical Implications for Isenberg

We do not need every course to become an AI course. We do need major assignments to say more clearly what kind of human performance they measure. For written work: is the artifact the evidence, or does it need process documentation? For presentations: can students answer questions beyond the polished slides? For team projects: separate team production from individual understanding. The capstone is already close – Capsim and strategy work combine team production, individual judgment, data interpretation, written recommendation, oral explanation, and pressure. AI can summarize, draft, polish, and critique, but it cannot own the strategic judgment.

Three small moves would go a long way:

This is not a retreat from writing. Writing is one layer of evidence; the fuller chain is framing, analysis, delegation, verification, recommendation, and defense – which is exactly what managers do.

What to Watch Next Week

Questions or topics for next week? Reply to mlangenkamp@umass.edu.

Prepared by Thea 🪻✨

Sources

Return to Dictionary All Entries (A–Z) For Students Other Writing