Instructional Governance by Design: A Framework for AI in Computing Education

AI tools need built-in governance, not just policies: Rather than relying on separate rules or guidelines, the paper argues that governance should be "encoded" directly into how AI teaching tools work—through their design choices about what they can do, who approves their outputs, and how they interact with students.

Different educational tasks require different levels of AI authority: The framework identifies six key dimensions for configuring AI tools, including how much authority the AI has in grading or feedback. For example, a code-style coach might have more autonomy than a system generating forum responses, which the paper shows should require staff review before students see them.

Human oversight must match the stakes of the decision: The paper emphasizes that accountability structures should scale with importance—high-stakes decisions like grades need stronger human control and approval gates, while lower-stakes uses like reflection prompts can allow more AI autonomy, demonstrating that "one-size-fits-all" AI governance in education is inappropriate.

Student learning engagement depends on tool design choices: How an AI tool is configured affects whether students actually learn—the framework highlights that tools should preserve "learner agency and cognitive engagement" rather than simply automating away student thinking, suggesting that more AI automation isn't always better pedagogically.

Institutions can use this framework to make intentional choices: By analyzing five different AI tools used in actual courses, the paper provides educators with concrete design questions to ask when adopting or building AI teaching tools, moving beyond reactive responses to proactive alignment of technology with educational goals.