In Virginia, SB394 establishes a legal framework to guide artificial intelligence (AI) use in schools, setting standards for division-managed platforms, requiring the Virginia Board of Education to provide guidance to divisions for the safe, ethical, and equitable use of AI, professional development pathways for educators on AI literacy, and creating the AI Innovation in Education Pilot Program to evaluate and scale AI applications across the state to support instruction, tutoring, student engagement, and teacher support.
Additionally, the Virginia Secretary of Education provided guidelines for AI integration in education, which outlines guiding principles such as do no harm, promote integrity, and the use of AI to enhance student success while including strategies to support integration such as providing professional development and facilitating information sharing.
Overall, Virginia is taking a balanced approach allowing public school divisions to innovate and explore while also protecting students. Only a few states have formally adopted AI education laws rather than disseminating guidance, making Virginia a national leader regarding AI in the classroom.
However, like other states, the policy is largely focused on generative AI (GenAI) use in the classroom – meaning predominantly student and teacher use for instructional purposes – rather than taking a comprehensive examination of the many ways that AI is or could be used to advance educational priorities and improve educational opportunities for all.
The focus on instructional use of AI creates a significant gap around the use of AI for equitable decision-making.
There are many highly valuable ways that AI can support decision-making. For example, Alpha Route is an AI-enabled transportation management tool that uses advanced analytics to examine routing algorithms and optimization to increase efficiency for school bus routes and reduce costs for districts.
For one district, the tool tested thousands of possible scenarios given the current routes and resources of the district. It found a new option that reduced the fleet requirements by 21 routes without changing bell times or bus stop locations. This solution eliminated the bus driver shortage problem for the district and with the reduced fleet saved the district over one million dollars per year.
While not an AI application, researchers at George Mason University partnered with a local district to use advanced analytics to examine the effects of a new policy that allows students to opt-in to intensified English Language Arts in middle school regardless of prior achievement.
The advanced course has the benefit of adding two to three additional months of learning for students. The highest estimated effects from the intensified course were among students with average or slightly above average achievement scores, suggesting that this group of students might benefit the most from taking the course.
However, researchers found that students from lower-socioeconomic backgrounds and higher achievement scores still did not enroll in advanced coursework, even under the opt-in policy. These findings provide education leaders with critical information about this new policy that can inform decision-making moving forward.
While the opt-in policy demonstrates benefits, more effort is needed to encourage academically average students and high-achieving, low-income students to register for the intensive option to benefit from the policy.
How Should Policies for AI Decision-Making Be Addressed?
As the two examples demonstrate, there are many benefits to using AI and advanced data analytics for decision-making in education by reducing costs, improving efficiency, and identifying additional policy and practices needed to make learning opportunities more equitable for all students.
At the same time, using AI in this way also carries risks.
In the example of Alpha Route, there was still a need for significant human oversight and involvement in the bus route audit and the final decision regarding which option to pursue from those generated from the AI tool. With the second example from George Mason, there was no personally identifiable data used in the analysis and student privacy was protected.
Maintaining the human in the loop and protecting student data are central to the current AI policy and governance discussions. However, with the focus on generative AI in the classroom, we risk losing sight of the possibilities of AI for decision-making and the safeguards needed for doing so. In many ways, AI use for decision-making carries potentially greater risks as it can impact all students and families in a district.
As these conversations on AI policy in education continue to unfold, the next iteration of AI policies will likely include a more expansive review of AI use, including its use for decision-making.
We should not wait for that to happen. Divisions can plan for this now by being explicit on safeguards for AI for decision-making and preparing leaders for the change. While educators are being trained on AI literacy and ethics for the classroom, administrators and division leaders should be trained on new innovations and workflows that leverage the power of AI to support equitable decision-making. There should also be more opportunities for divisions to share their learnings, concerns, and novel ideas to spur greater innovation and improvement in education for the state.
In doing this, Virginia can continue to be a national leader in AI policy for education.
BETH DAVIS is a Postdoctoral Fellow at EdPolicyForward: The Center for Education Policy at George Mason University.
