Variation in School District AI Policy Within Contrasting State Policy Environments

• In Event: The Emerging Role of AI in Education and Workforce Preparation

Abstract

As generative artificial intelligence (AI) becomes more embedded in K-12 education through teacher and student use, school districts are increasingly expected to develop policy and guidance for its use in schools. That need is driven by AI’s potential academic benefits and its risks to students’ critical thinking and well-being. Yet, districts are developing such policy within state environments that vary in mandates, guidance, and implementation supports, and within economic and geographic contexts that may further influence their capacity to develop comprehensive AI policy. In practice, substantial variation in AI policy may influence whether, and in which settings, educators receive clear and actionable guidance for engaging with AI safely, responsibly, and effectively. Studying school district AI policy across contrasting state policy landscapes and district contexts is therefore necessary to better understand how state policy and local conditions influence district AI governance, and to inform more equitable state policy design and capacity-building efforts.

Accordingly, this study asks two primary research questions: first, what does the current landscape of school district AI policy look like with respect to how well district policy supports teacher AI competency; and second, how does district AI policy vary across contrasting state policy environments and local contexts? To address these questions, this study examines variation in school district AI policy across Kentucky, North Carolina, and Tennessee, which represent three contrasting state-level approaches to AI governance in education. Kentucky has taken a locally driven approach to AI policy, North Carolina has moved toward a more coordinated statewide approach, and Tennessee has passed state law requiring school districts to adopt some form of AI policy. Using a nested mixed-methods design, this study analyzes 946 documents collected from 424 public school districts across the three states. It applies a systematic, evaluative coding protocol aligned to UNESCO’s AI Competency Framework for Teachers to assess how well a district’s policy supports teacher competency for safe and effective AI use. Logistic regression is then used to estimate whether state context, district childhood poverty, and urbanicity are predictive of whether a district’s AI policy is supportive of teacher AI competency.

Findings indicate substantial variation in district AI policy across state policy environments and local contexts. Across the full sample, 87% of districts’ AI policy were rated unsupportive of teacher AI competency. A domain-level analysis revealed that districts were especially unlikely to address more complex ethical and pedagogical issues of AI use in schools. In terms of differences between states, districts in North Carolina had 12.7 times the odds of meeting the threshold of supportive AI policy compared with districts in Kentucky, while districts in Tennessee did not differ significantly from those in Kentucky. Further, childhood poverty was negatively associated with district AI policy supportiveness, whereas urbanicity was not a significant predictor. These findings suggest that district AI governance may be influenced by normative isomorphic expectations set by centralized state guidance but not by coercive isomorphic pressures. The results also suggest that districts serving higher-poverty communities may face greater barriers to developing comprehensive AI policy.

Author

• Mitchel Ryan Mandel, University of Louisville
  Presenting Author