Data Report · RedactSure Research
Chris Sowa · Founder, RedactSure
Published
Last updated
Does the Need-to-Know Rule Survive AI? A Former County Superintendent on Student Records
From HAL: Who Signed Off on This?, with Dr. Judy White, retired Riverside County Superintendent of Schools
Episode: Need to Know: AI and Student Records, with Dr. Judy White. Listen on Libsyn. The show is also on Spotify and YouTube.
In brief
-
Student records now combine academic, health, social and family information in one place, which is why staff need more privacy training, not less.
-
Districts that suffered breaches often responded with so many layers of security that collaboration around the student suffered.
-
AI should act as a teaching assistant, not a teaching replacement, and training teachers to use it well costs time and budget.
-
AI can surface a pattern of referrals. It cannot tell you what a child is living through, so the decision stays with a person.
-
Her rule for records: if something is not needed to know, take it out.
-
Policy has to be in place before the incident, as schools learned with social media.
About the episode
Dr. Judy White spent 42 years in public education, from counselor and principal to district superintendent, and finished as Riverside County Superintendent of Schools, with responsibility for about 430,000 students. Today she leads MOSAIC, a nonprofit focused on racial healing and student recognition, and works with Think Transcend to develop current and aspiring superintendents.
On HAL: Who Signed Off on This? she explains what changes when a student's academic, health, discipline and family information sits in one record, why that makes AI both useful and risky for overloaded teachers, which decisions must stay with people, and why the old rule of keeping only what is needed to know matters more now.
Highlights from the conversation
What changes when everything about a student is in one place?
Chris Sowa: There's a lot of process and documentation that the average parent and certainly students probably don't appreciate that staff are managing: IEPs, disciplinary files, medical records, academic records, maybe even family information. All that information is put together. Can you talk a little bit about some of the administrative challenges?
Dr. Judy White: Today, because so much information is stored, there is information from the health to the social sciences to academics. All of these things, but now they're integrated around that student. And so parents sometimes don't mind you knowing their academic deficiencies, but they don't want you to know something about the house. And then what happens, you do have all the information in one place. And it makes it difficult because you cannot really responsibly respond unless you know the total child, unless you know everything that's going on.
But that means you have to actually teach more privacy laws for your teachers. You have to talk to your staff about confidentiality at a greater level.
The consolidated record is what lets a school respond to the whole child, and it is also what makes a careless AI prompt so costly. A teacher who pastes part of that record into a public tool to draft a parent letter exposes far more than an academic note.
Should leaders be concerned about teachers using AI with student information?
Chris Sowa: Because they're so overburdened with all the things on their plate, teachers are reasonably trying to leverage tools like AI to bring information together, maybe to make a response for a parent or to summarize it. As somebody who's seen all this data, are you concerned about this? Do you see an opportunity here?
Dr. Judy White: Yeah, I am concerned. So we're trying to find this balance, having technology support, but not take over. However, there are people that do not mean well, and they will access different data points and use it against what we're trying to do. And so this data security is a high priority.
And some districts, their data has been hurt. And then they've had to put levels and levels of security in order to make sure it doesn't happen again. And then they end up restricting their opportunities for collaboration and for knowledge about every topic within a student's profile.
Many districts live with that trade-off. After a breach, the natural response is more locks, and the cost is that the people who need to work together around a student can no longer see what they need. The goal is to control what each person and each tool sees for each task, not to shut everyone out.
How much AI training do teachers need?
Chris Sowa: How much development do you think we need to provide our faculty and staff within the school system, and do you think they're getting enough support considering how fast AI is moving?
Dr. Judy White: I do not think that they're getting enough support. In order to properly train, you have to capture time. And then you work in your curriculum and content. So when do you capture that? And so there's a cost. You bring your teachers in during the summer for a full week of AI, that has to be paid out of the budget. And in some cases teachers are volunteering to come to professional development because they want to be knowledgeable. I believe that's going to have to be a budgeted [item].
But when you're talking about professional development, you're also talking about a human element. AI should act as a teaching assistant, not a teaching replacement. Human educators are essential for emotional support. We don't want AI to replace them. Teachers also have to be taught how to guide the students in the use of AI so that they're not allowing them to do the work. We want to make sure that they're able to still think critically.
Which decisions must stay with a person?
Chris Sowa: Could you give us some examples of decisions where AI might inform you or bring data together, but you'd say, I need my principal, my teacher, or myself to actually make those hard decisions? Those critical, human, consequential decisions?
Dr. Judy White: In the superintendent's role, whether it was at the local district or the county, we remain at a seat of appeal. Parents can come and say something you're not considering, or you didn't follow a policy or a process.
With AI, you are able to capture incidents that the students had since the very beginning. If the teachers are putting referrals in the system, when you get to a place of expulsion, you're able to see a pattern. However, AI won't tell you that […] It won't tell you certain things. And so you as a decision maker have to make sure you take the time to find out all the information. Did you as a school district intervene when you saw that student just putting his head on the desk every day, trying to manage that on their own? Or did you just wait until they're striking out? So we do have to pay attention to all types of human behavior and provide opportunities for discussion.
The pattern is the AI's contribution. The context, and the decision, belong to a person who can be held to account at the seat of appeal. That division of labor is the same one any AI decision process needs: the tool assembles, a named person decides.
What should be in a student's file at all?
Chris Sowa: When I was a kid, I was handed a paper file when I went from middle school to high school and told to take it to the nurse. Nowadays I would imagine that's electronic. How have your teams thought about protecting that health information?
Dr. Judy White: If something's not needed to know, take it out. The most effective system I've seen is when people treat the students as individuals and the team sits together and meets around that student. The nurse, the counselor, a teacher, and you sit there with the vice principal or administrator and talk: are we setting this student up for success? And you have to have a spirit of confidence. You can't go out and share everything that you know. These professionals have taken an oath to be able to discuss amongst themselves for purpose, but not necessarily to get outside that room.
Chris Sowa's follow-up on the episode put the same rule in AI terms: an AI organizing information for a teacher or a student needs certain things to do the task, and it does not need the rest, so the rest should be taken out. Need to know works as a design rule for AI just as it does for people in the room.
What did schools learn from social media?
Chris Sowa: What do you think needs to be in place at the school level and at the superintendent or district level, in terms of protections and policies?
Dr. Judy White: I do believe that social media has impacted greatly, and too often students actually feel that the social media information is more true and more believable than even what they hear in school. So you have to actually intentionally instruct them to find out what the difference is. We've had to have restrictions on personal social media for teachers. Teachers felt that their social media was private, and they would put personal feelings about educators as well as students in that setting. And then they say, well, this is my private Facebook, this is my private social media. And students actually have access to that.
So you have to have policies in place before it happens. You sit down, you develop the policies, and then you have to have opportunities to teach them the policies.
How do students learn to question AI?
Dr. Judy White: One of the best examples I saw was at a conference. They asked the students to put a question in and then they compared their answers. And then they showed them where some places were errors, some places were hallucinations. And then the students themselves started saying, what can we do to check information? And they created from that experience rules for their classroom. So you've got to give people experiences so that they will be able to build those critical thinking skills on their own.
What school and district leaders can take from this conversation
-
Treat the consolidated student record as the most sensitive thing a district holds, and train staff on confidentiality accordingly.
-
After an incident, control what each person and tool sees per task instead of locking everyone out.
-
Budget for AI professional development rather than relying on volunteers.
-
Let AI surface patterns, and keep consequential decisions with a named person who can hear an appeal.
-
Apply need to know to AI: give the tool what the task requires and take the rest out.
-
Write AI policy before the incident, then teach it.
Where RedactSure fits
Dr. White's rule is the design principle behind RedactSure: if something is not needed to know, take it out. For a district, the AI works on tokens in place of student identifiers, so it can help with the paperwork without the real values entering the model. Permissions stay exactly as they are for the principal, counselor or administrator, and a named person approves anything consequential.
Frequently asked questions
Should teachers use AI with student records?
Dr. White sees AI as a teaching assistant, not a teaching replacement. Teachers need training and clear policy first, and student information should go only where it is needed for the task.
What decisions about students should stay with people?
Discipline, expulsion, appeals and early intervention. AI can show a pattern of referrals, but it cannot tell a decision maker what a child is experiencing at home, so a person must gather the context and decide.
What does need to know mean for AI in schools?
Keep in a record, and give to any tool, only the information required for the task. A team around a student shares what it needs inside the room. An AI helping with a task should receive what that task requires and nothing else.