A Framework for Institutional Change in the Age of AI

Synthesis: Perl-Nussbaum & Finkelstein (2026) propose a six-dimension framework for adapting institutional change models in STEM higher education to the realities of generative AI. Unlike prior reforms based on adoption technologies (stable, evidence-based practices like Peer Instruction or PhET simulations), generative AI is an arrival technology — it entered classrooms before pedagogical evidence could form. The framework identifies where existing change models break down and derives actionable design implications for leading change under genuine uncertainty. Central insight: "Institutions cannot wait for best practices, but cannot responsibly scale unjustified innovations. Neither banning nor uncritical embrace is tenable."

Six Dimensions of Reconsideration

Tools Dimension

1. Evidence Base: Prior models assumed reform begins with evidence-based tools; AI arrived without pedagogical evidence. Design implication: privilege humble, local inquiries — document what faculty are trying, share cautiously, avoid overclaiming generalizability.

2. Rate of Change: Prior models assumed stable tools; AI evolves on timescales of months. Design implication: organize reform around pedagogical approaches rather than specific AI tools, which may be obsolete before evidence accumulates.

3. Scope: Prior models targeted bounded interventions; AI's impact is broad and systemic. Design implication: adopt a systemic view — coordinate across courses and departments rather than responding piecemeal.

People Dimensions

4. Faculty: Prior models positioned faculty as adopters of proven practices; in the AI era, faculty must become inquiry-driven experimenters navigating genuine uncertainty. This connects to the Educational Development finding that faculty are "pragmatic realists" navigating competing priorities with scarce resources.

5. Change Agents: Prior models cast change agents as disseminators of best practices; when best practices don't exist, they must become facilitators of collective inquiry. This reframes Centers for Teaching and Learning from trainers to community organizers — a transformation anticipated by The University AI Didn''t Replace: Rethinking Universities in the AI Era.

6. Students: Prior models treated students as recipients; in the AI era, students are often ahead of faculty in AI use and must be engaged as partners in reform. This aligns with broader calls in Student Experience and AI Literacy for student agency in AI-era education.

Application: Physics Department Workshop Series

The framework was piloted through a faculty workshop series in a university physics department. Rather than prescribing AI tools, workshops facilitated collective inquiry around pedagogical goals, engaged participants as co-investigators, and organized around teaching approaches rather than specific technologies. This connects to the departmental-level work described in STEM Education and the workshop-based models in Educational Development.

Connections to Knowledge Base

• Extends The University AI Didn''t Replace: Rethinking Universities in the AI Era by providing a concrete framework for institutional adaptation, not just rethinking
• Complements Educational Development by adding the change-agent and student-partner dimensions missing from playbook-centered approaches
• Connects to A principled way to think about AI in education: guidance for educators and policy makers based on goals, models — the framework operationalizes Finkelstein's goals-models-technologies lens at the institutional level
• Relates to Higher Education and Teaching — redefining instructor and institutional roles in AI-era education
• Aligns with AI Literacy calls for faculty AI literacy as a prerequisite for meaningful institutional change
• Contrasts with Scaffolding and The Pedagogy of AI Mistakes: Fostering Higher-Order Thinking — while those address classroom-level AI integration, this framework addresses the institutional conditions needed for such integration to succeed
• Shares the systemic perspective of — institutional capacity is a bottleneck for AI in education globally

Open Questions

• How does this framework apply outside STEM — in humanities, social sciences, professional programs?
• What metrics track institutional adaptation progress under genuine uncertainty?
• How do different institutional types (research universities, community colleges, liberal arts) shape framework application?
• Can the framework be operationalized into assessment tools for institutional readiness?

What this means for practice

• Faculty developers. Anchor workshop design in pedagogical approaches and the department's own teaching problems rather than in specific AI tools: the framework's tools dimension argues that AI changes on timescales of months, so tool-anchored reform risks obsolescence before evidence accumulates.
• Faculty developers. Reposition the center for teaching and learning as a facilitator of collective inquiry rather than a broker of best practices — the physics workshop series surfaced what was happening in participants' courses instead of prescribing approved tools.
• Administrators. Invest in sustained spaces for faculty inquiry and in dedicated expert support rather than in tool procurement, following the paper's Science Teaching Fellows precedent of funding discipline-based partners in course transformation, and build capacity for continuous adaptation rather than one-time adoption.
• Administrators. Treat students as partners in the change process rather than its recipients: the framework holds that students are often ahead of faculty in AI use, so their existing practices are evidence about what is already happening in courses.
• Instructors. Document and share local experiments with their limits stated and refuse to generalize them; the framework's answer to an absent evidence base is humble, local inquiry.

Limitations

• No empirical test of the framework. The authors state plainly that they have not tested whether change initiatives designed in accordance with it produce positive outcomes relative to alternative approaches; it is a theoretical contribution.
• A single, brief case study. Application is illustrated through one faculty workshop series in one university physics department, so the six dimensions are shown in a single discipline and institution type.
• Context-bound evidence. The framework is drawn primarily from U.S.-based change initiatives, and the authors note that the relative weight of the dimensions and specific design implications may shift in different institutional contexts.
• Deliberately provisional. The authors concede that the dimensions most salient today may be reordered as the technology and practices mature, and present the framework as a starting point rather than an exhaustive or final model; the workshop materials are available only from the corresponding author.

Citation

Perl-Nussbaum, D., & Finkelstein, N. D. (2026). A Framework for Institutional Change in the Age of AI.