Some student data is dangerous if it stays online. Not just risky. Wrong.
Charles Fadel writes about this from the Center for Curriculum Redesign. He calls it Cognitive Security. It’s about where we draw the line.
Schools track grades. Attendance. Test scores. That’s the old model. Easy to digest. Safe enough if locked down properly.
New tech is different.
Intelligent Tutoring Systems adapt to kids. Fast. They watch how long a student hesitates on a problem. They note abandoned tasks. Repeated errors. The slump in engagement over time. It’s subtle.
From that, systems guess things. Big things.
Cognitive difficulty. Emotional state. Personality traits. Anxiety risk. Motivation levels. Confidence. These aren’t grades. They’re psychological inferences. And that’s where the trouble starts.
The danger isn’t just hackers. Breaches happen. Bad news, yes. But the deeper problem is worse.
What if the data is safe?
What if a hostile actor or just a well-meaning but flawed system uses a lawful profile to steer a child’s development? A secure database is still a record. And a persistent psychological record of a nine-year-old is a trap.
Kids aren’t static objects.
They’re fluid. Changing. A label applied at age nine sticks. “Low engagement.” “Poor executive function.” It starts as a system output. Then it becomes a teacher expectation. A parent worry. A kid’s self-identity.
Labels stick harder than glue.
So the question changes. It’s no longer “How do we lock this database?”
It is “Should this info be digital at all?”
We need a split. A hard border between analog and digital.
Learning data fits in the cloud. Progress. Pacing. Short-term engagement needed for tomorrow’s lesson. Keep it tight. Keep it brief.
Session-only data should vanish after it serves its purpose. Scaffold now. Forget later.
But high-risk categories? Emotional state. Personality. Risk profiles. These never get digitized. They stay with the teacher. In their notebook. In their mind. Analog form only.
Is this anti-tech? No. It’s risk management.
A paper note is limited by design.
You can’t query it easily. Can’t sell it. Can’t merge it with external data. Can’t decrypt it in twenty years. It ages. It decays. It stays tied to human judgment.
A digital record is permanent. Searchable. Portable. It waits for uses we haven’t even imagined yet.
The table from Fadel’s work offers a boundary. It separates acceptable learning metrics from sensitive psychological profiling.
Think of fractions. A system knows a kid hasn’t mastered them. Useful. Actable.
That same system should not infer the kid is anxious. Impulsive. Low on persistence.
One supports instruction.
The other builds a dossier.
Turn a temporary moment into a permanent identity? That’s harmful.
Systems need to follow three rules.
– Necessity : Collect only what’s needed.
– Retention : Keep it briefly.
– Boundaries : Block personal inferences.
For the most sensitive stuff, encryption isn’t enough. The solution is non-digitization.
This isn’t about hating AI tutors. Personalization is good. If it stays educational.
If it turns psychological, it fails.
Kids need support today. Not a profile that haunts them forever.
Some data expires.
Some never becomes digital.
