This week is my last professional development trip during the busy back-to-school season. June, July, and August are always the busiest time of year for me as I deliver keynote speeches, hands-on workshops, and presentations about teaching, tech, and AI.
This is my third time visiting this school, and my point-of-contact asked early on …
“Can we do something with AI and data?”
My answer? Yes, ABSOLUTELY. This seems to me to be a vastly underutilized area for school leaders AND teachers.
As I’ve been preparing for that presentation, I’ve been doing a lot of thinking about it. In today’s newsletter, I’ll share a few ideas — including one example right out of my own classroom.
📋 POLL: As always, I’m asking you to share in our weekly poll … but this week, we can really use it! (Especially if you’ve used AI to help analyze data.) Be sure to vote — and, if you have it, share an example!
In this week’s newsletter:
💡 Did you know that in Kira, you can...
📚 New AI resources this week
📢 Your voice: Seeking teacher buy-in about AI
🗳 Poll: AI to analyze data
📊 Data + AI: Get insights on just about anything
💡 Did you know that in Kira, you can...
This message is sponsored by Kira
Build an interactive lesson from scratch or adapt one from materials you already have, and assign it without leaving the platform
Give students their own personalized guide, not a general chatbot
Differentiate without rewriting anything — same lesson, three reading levels? Just a few clicks.
Create standalone assessments that mirror your state tests, in content and in format
Track mastery across skills and assign personalized practice
Curriculum. Assessment. Intervention. All in one place.
Learn more about:
📚 New AI resources this week
1️⃣ AI Tutors Not Yet a Replacement for Humans, Research Says (via The 74) — Recent research on AI tutors including Amira and Khanmigo found surprisingly low student engagement and little evidence that simply providing access produces better learning.
2️⃣ Can AI Adequately Assess Critical Thinking? (via K-12 Dive) — New research found AI could evaluate some elements of student critical thinking reasonably well but struggled with others; even after training and prompting, no measured subskill exceeded 78% accuracy.
3️⃣ In This District, AI Is a Teacher's Personal Assistant (via District Administration) — Washington's Peninsula School District is rolling out personal AI agents for educators while deliberately starting with principals and instructional leaders before focusing heavily on student AI use.
📢 Your voice: Seeking teacher buy-in about AI
Last week’s poll: How do YOU seek teacher buy-in?
🟨🟨🟨⬜️⬜️⬜️ The time you can save (4)
🟨🟨⬜️⬜️⬜️⬜️ What it will make possible (3)
🟩🟩🟩🟩🟩🟩 Lots of practical examples (7)
🟨🟨🟨🟨⬜️⬜️ One-on-one conversations (5)
🟨⬜️⬜️⬜️⬜️⬜️ Other ... (2)
Other: Honesty- being open and honest about my feelings about it, allows them to enter into the conversation not being worried I am trying to force them into believing or behaving a certain way. — K. Kohn
What would you like to read in AI for Admins?
What’s a topic you’d like to see covered here? Hit REPLY to this email and let me know.
Have you done anything you’d like to share with the AI for Admins community? Hit REPLY and let me know.
Would you like to write a guest post to support and equip AI for Admins readers? Hit REPLY and let me know.
🗳 Poll: AI to analyze data
Instructions:
Please vote on this week’s poll. It just takes a click!
Optional: Explain your vote / provide context / add details in a comment afterward.
Optional: Include your name in your comment so I can credit you if I use your response. (I’ll try to pull names from email addresses. If you don’t want me to do that, please say so.)
How do you use AI to analyze data?
📊 Data + AI: Get insights on just about anything
Use AI to analyze data and give you suggestions. (Image: ChatGPT)
For the longest time, data has been a pain in education.
I mean, it can be great. There are trends and anomalies and insights that you can glean — if you’re able to pick them out.
But for some of us, data has been associated with a two-inch ring binder full of papers … landing with a resounding “thud” on our desks after someone says: “Hey, can you take a look at these numbers and tell me what you think?”
(As if “take a look at these numbers” is going to be easy and take about 90 seconds …) 😒
Thankfully, it doesn’t have to be as painful anymore.
Lots of the artificial intelligence models that we have ready access to are built to handle large amounts of data … to tell us what it means … to tell us what might happen as a result … to suggest strategies based on the data.
After we wrap our brains around a few things about that data and AI, it can be very enlightening.
First: What kind of data?
That very much depends on what your role is and what you’re trying to accomplish.
There are lots of datasets we can put our hands on that can give us some insights into what’s happening and what we can do …
Global data: PISA and TIMSS data, UNESCO education statistics, global literacy and numeracy trends.
National data: NAEP (“the Nation’s Report Card”), stats from NCES (National Center for Education Statistics), national graduation rate trends, College Board/ACT national percentiles.
State data: State summative tests, state report cards, state longitudinal data systems.
Community data: Census data, housing data, languages spoken at home (via ACS), other federal/state datasets that tell you about your community.
School data: School-level test and growth data, attendance/absenteeism data, discipline referrals, course enrollment, school climate surveys, etc.
Classroom data: Gradebook patterns, unit/chapter test results, quiz scores, participation patterns.
Formative assessment data: exit tickets, warm-ups, results from a Kahoot! game … quick in-the-moment data.
Platform/app data: Usage and mastery reports from learning/practice apps, analytics from your learning management system, etc.
Of course, it’s important to know that the data that you have actually relates to the problem you’re trying to solve. There’s an old computer programming saying: garbage in, garbage out. If the data coming in isn’t what you need, the insights and results will be useless.
Second: Keep everyone safe and data private
Having lots of data at your fingertips can be really helpful. But don’t get carried away and start using it in places and in ways that you’re not supposed to.
Protect PII (personally identifiable information): This is any data that can be used to distinguish or trace someone’s identity. Here in the U.S., we’re bound to protect it. (And even if you live where it isn’t, you should really consider it.) Strip out names, student IDs, phone numbers, social security numbers, etc.
Respect intellectual property: Data belongs to someone. In many of the examples above, the data is publicly available and you can use it. But if it’s more proprietary, make sure to get permission to use it in the way you intend.
Understand how the AI model will handle it: Not all AI tools handle data the same way. Drop some data into the free version of ChatGPT (or another similar AI app)? That data will be retained and used to train future AI models. (Not exactly like the AI app being able to look it up … more like a human who has read something and remembers the general nature of it.)
An example: When you use a school Google or Microsoft account, that account has certain data privacy protections. When you use Google Gemini or Microsoft Copilot with your school account, your data won’t be used to train AI models. (See below image.) Now … to me, this doesn’t mean you can just throw any data you want into it. We’re still legally bound not to share student PII — even if the AI model isn’t training on it. But if you strip out PII and use a model like this, you’re mitigating lots of the risks.
Notice the little shield icon in front of “Ask Gemini.”
So … what can you do with data?
You can just say: “analyze this data.” But that means the AI model has to make judgment calls about what you really mean.
Instead, you can ask questions like these …
Descriptive questions: “What actually happened here?”
Diagnostic questions: Why might this pattern exist?”
Comparative questions: “How are these similar/different? Where’s the gap?”
Predictive questions: “What might happen if this trend continues?”
Prescriptive questions: “What should I do differently?”
A few examples …
Diagnostic analysis of student work — A teacher collects student work … then photographs or scans the entire stack. (My favorite: use the Google Drive app on my phone OR use the teacher workroom photocopier’s scanner feature.) The teacher asks AI to identify patterns or concepts that might be worth reteaching or clarifying.
Comparative analysis of data sources that disagree — Imagine that your state test results say that students struggle in one area … but the classroom unit test says that they’re fine. Upload both tests (if available) to an AI model along with the academic standards they’re supposed to measure. Then see how AI analyzes their alignment to the goals and how they measure them.
Predictive analysis of students on track to graduate — This one can be telling, but you have to be really, really careful. A middle school collects attendance data, course grades, and an academic benchmark. They connect each data point to students through anonymized student numbers (so they aren’t passing PII off to the AI model). It can help counselors and teachers identify who might need additional support to succeed.
I recently tried #1 in my own class. After having my students fill in a graphic organizer — where they look at an image and write sentences/pick out words represented in the picture — I scanned the whole stack of student work. (I blacked out their names beforehand.) Then I asked Google Gemini …
I used Gemini to analyze student responses.
Here’s how it responded:
AI suggested a few things I could re-teach my Spanish class.
These don’t surprise me! But it is nice to have insights based on data. And really, that stack of student work had a ton of data points to give me those insights!
Remember what the data means (and doesn’t mean)
I mentioned that with #3 above, you have to be really, really careful.
First of all, remember what predictive analysis really means. It isn’t actually predicting anything. It isn’t telling the future. It’s just making best guesses according to the data and the trends that it identifies.
Here’s the other thing. AI predictions can sometimes reinforce inequities and magnify the problems that the system is already creating. Example: if it’s using disciplinary referrals, if certain students are being referred unfairly or disproportionately, it will highlight them and further pull them out, which could make things even more unfair for them. By the way: just saying “well, the AI algorithm said so” isn’t enough. You need concrete human judgment. Using AI doesn’t excuse you from doing human thinking.
It’s all still built on the expertise of the human
AI can analyze data. It can give you suggestions and tell you what patterns it notices.
It’s still up to the human to know what to do with the results … to know if the suggestions are actually helpful to the human students they’re working with … and to protect the human students from unfair or unsafe practices.
But with a plethora of data in our hands, we can get better and quicker insights — and actually use them right away — when AI helps us process and analyze them.
How are you using AI to analyze data? Answer today’s poll and tell us!
I hope you enjoy these resources — and I hope they support you in your work!
Please always feel free to share what’s working for you — or how we can improve this community.
Matt Miller
Host, AI for Admins
Educator, Author, Speaker, Podcaster
[email protected]
