Tracking student progress and flagging kids at risk of failure. More customization of lessons to meet individual students鈥� needs. Professional development tailored to individual teachers. Automatic essay grading.
Those are just some of the tasks in K-12 education that experts say could鈥攐r already are鈥�being performed with the help of artificial intelligence. AI is already transforming retail, agriculture, medicine, and other industries. Its impact on K-12 education is only expected to grow, along with nearly everything else in the economy.
With that in mind, the White House released a earlier this month. Here are some critical facts educators should know about it.
The AI bill of rights is centered around five principles
You should be protected from unsafe or ineffective systems. That means, among other things, that AI systems need to be tested before they are rolled out, and then carefully monitored to make sure they鈥檙e working as intended.
No one should face discrimination by algorithms. Systems should be used and designed in an equitable way. AI systems reflect the biases of the people who program them. That鈥檚 why, for instance, an algorithm that鈥檚 designed to decide who gets a financial loan may inadvertently disadvantage Black borrowers. Having people from different backgrounds design AI-powered systems is one possible solution.
You should have protections against abusive data practices and agency over how data about you is used. AI systems rely on data, and student data privacy is obviously a big issue in any tech that鈥檚 powered by AI.
You should know that an automated system is being used and understand how and why it influences outcomes that impact you. K-12 schools could play a big role here too, in helping students understand the technology and how it impacts the world around them.
You should be able to opt out, where appropriate, and have access to a person who can quickly consider and fix problems you encounter. That would seem to imply that companies that create learning software with AI would have to respond quickly to any problems educators or parents raise.
The AI guidance has no real legal authority
The bill of rights is simply guidance for areas of the economy that rely on AI, though that鈥檚 increasingly in nearly every area of the economy. If anything, its principles may apply to AI use by the federal government, . But it鈥檚 not going to force Facebook or Netflix or even a state criminal justice system to make changes to the way they use AI, unless they voluntarily decide to embrace the principles.
The U.S. Department of Education is one of a number of agencies that is supposed to follow up on the bill of rights. It is expected to release specific recommendations for using AI in teaching and learning by early in 2023. The recommendations should include guidelines for protecting student data privacy when using AI.
What data privacy experts see as problematic
Amelia Vance, the founder of Public Interest Privacy Consulting and an expert on schools and data privacy, thought the general tenor of the document was the right one, but she wondered just how much outreach the White House had done to K-12 education groups, given some of the examples used in the guidelines.
For instance, in elaborating about data privacy, the document said that ideally, data should be most accessible to those who work directly with the people that the data pertains to. And it gives, as one example, a teacher getting more access to their students鈥� data than a superintendent.
鈥淭here are many school districts who have decided that they want the superintendent or principal to have access and be able to see across the schools [how] the teachers are serving their students,鈥� Vance said. 鈥淚t just raises some really serious questions again about who they talked to鈥� in making recommendations for K-12.
What鈥檚 more, it might not be practical for schools to always get parental permission before allowing students to use learning technology that relies in part on AI, she said. But that鈥檚 how some might interpret the guidelines.
鈥淚 think it鈥檚 largely the same reason that many superintendents and teachers are struggling with parents wanting to be able to individually approve the reading their kid has to do,鈥� she said, referring to a push by parents in some communities to review curricular materials before they are used with students. 鈥淚t鈥檚 often impractical. It is difficult for teachers to build their curriculum. It鈥檚 difficult for the school to move forward to make sure that everyone is learning the same things and that learning is provided in an equitable way.鈥�
What do people from companies that create tools for student learning think?
Having guidelines can be helpful to companies, particularly those that want to reassure schools they will safeguard data and root out bias.
鈥淚f somebody wanted to build an AI system, there鈥檚 some nice guardrails there to help you build a better system,鈥� said Patricia Scanlon, the founder and executive chair of SoapBox Labs. It designed a natural language processing technology specifically for children鈥檚 voices that is used in educational products developed by McGraw Hill and other companies.
Like other international companies, SoapBox Labs, which is based in Ireland, will have to comply with pending European guidelines for AI, which may be stricter. What鈥檚 more, unlike the White House AI bill of rights, those guidelines may come with an enforcement mechanism.
Earlier this month, SoapBox Labs became the first company to receive the Prioritizing Racial Equity in AI Design Product Certification, developed by two education nonprofits, Digital Promise and the EdTech Equity Project.
School districts may feel more comfortable using certain products if an outside evaluator confirms that they meet certain privacy and bias mitigation standards, Scanlon added. 鈥淚t can give some confidence so not everybody has to be an expert in AI,鈥� she said. 鈥淚 think the stakes are just higher in education than they are for your Netflix recommendation,鈥� which can also be driven by AI algorithms.
