The Algorithm in Classrooms

Imagine this: a middle-schooler in Florida logs into her math class online. Her virtual lesson is being customized in real-time, but there is no monitoring teacher. Rather, an AI is actively tracking her response speeds and which questions may be tripping her up. It collects data on where she is hesitating and how long she is looking at her screen, all in order to recommend her the perfect next problem set. Of course, her teacher is able to see the recommendations on her dashboard, and her parents get regular updates on her progress, however, none of them know exactly why these decisions were made, and how the algorithm may affect the students.

Across the country, this is the new normal for students; AI tutors have the ability to shape how and what students learn. However, this all has evolved so fast that policymakers do not even know what questions to ask about these new innovations. If we do not create regulations now, we are potentially allowing private companies to control our nation's public education.

These systems are called adaptive learning platforms. Many companies have become increasingly popular in schools, like DreamBox and Khan Academy. In fact, nearly 64% of middle-schoolers use digital learning on a daily basis. They work by tracking the behavior of students, as well as their performance in real time. They then use this data to make predictions about what a student is ready to learn at a perfect “Goldilocks” medium. These companies often emphasize this personalization, as it avoids the instances of too easy material leading to slowed progress and too hard material frustrating the student, which often occurs in the group settings of traditional teaching styles. DreamBox, for example, collects over 48,000 data points per student per hour, in order to achieve algorithmic near perfection. 

A Legal Framework Stuck in the Past

This technology sounds incredibly promising, and in many cases, it truly is. However, the problem lies in the lack of regulation regarding the products. There are currently no clear federal rules about how this kind of AI should be used in classrooms. FERPA, which is the Family Educational Rights and Privacy Act, written in 1974, only protects records controlled by schools and has no jurisdiction directly regarding third-party vendors.  COPPA, which is the Children's Online Privacy Protection Act, is just as outdated as FERPA. This act only applies to kids under 13 and only covers commercial websites and apps. It was not at all designed with classroom tools, especially those utilizing AI, in mind. As these companies continue to have, essentially, free reign over children's educational data, parents have become increasingly concerned. 69% have active anxieties about student tech privacy, yet no intiatives have been taken by governments to rectify this. To exacerbate this, the edtech companies, who are collecting huge amounts of behavioral data and storing it indefinitely in some cases, are rarely ever being transparent about how they are being used or whom it is being shared with.

Even more concerning is how little parents are told that AI is being used at all. For instance, in a study done by Common Sense Media, six in ten parents have heard essentially nothing about private companies storing their child’s data. However, this is not simply about discomfort. Data collection has become a real threat. The US Government Accountability Office has documented rising numbers of student data breaches across thousands of public schools. Personally identifiable information is often being exposed during these breaches, leaving students vulnerable to emotional and financial harm. 

AI Educational Potential 

AI and AI powered classroom tools do have their place, however. Students in personalized learning environments, where things like adaptive learning platforms are being used, often perform 1.5 times the national average in mathematics and reading testing. Furthermore, these tools keep students more engaged and likely to complete homework tasks, outside of school time. 78% of educators also believed that using these platforms felt more effective than teaching without their assistance. They are also incredibly powerful for students with learning disabilities. In traditional learning environments, children with learning disabilities, on average, reach 62% of the learning outcomes that their standard peers do. With adaptive technologies, they can reach upwards of 85% of the outcomes, a drastic jump which can make a real difference in their lives.

It is clear that it is time for Congress to update legislation, like FERPA and COPPA. They need to reflect the new learning environments entering US classrooms, like driven learning platforms. While these tools have great potential in increasing learning outcomes across all different types of students, that should not allow them to have unregulated access to our children's data. Students, and their parents, should have the right to opt out and have the ability to delete their data. Furthermore, the school districts using and the companies running these tools must have rigorous privacy and fairness review. With these tools, the United States has the possibility for a serious enhancement of their public school education.