Using AI to Analyze Student Data, Part 2

TL;DR (full transcript and prompts below)

This video demonstrates a multi-step process for using AI to analyze student data. This is the part 2 of data analysis with AI.

In this example, we analyze student assessment data from two assessment periods (initial vs. 9-week math scores) to evaluate student growth and instructional effectiveness, identify students’ growth patterns, and plan next instructional steps. Easily modify the data information based on your assessment data and type.

Step 1: Upload Assessment Data

  • Dataset: Upload the data file (but remove any PII – personally identifying information first).
  • Task: Create a table showing how many students moved up 2+ levels, moved up 1 level, etc.

Step 2: Changes in Proficiency Levels

  • Goal: Quantify student growth between assessments.
  • Insight: Celebrate improvement, but recognize that one-level growth may already represent maximum achievable progress for high-level students.

Step 3: Graph of Student Gains

  • Create a histogram of counts for each change category.
  • Label bars with student counts for clear visual communication of growth trends.

Step 4: Comparison Bar Graph + Trendline

  • Create a bar graph comparing initial vs. 9-week levels with student counts.
  • Add polynomial trendlines for each series to illustrate overall progress.
  • Interpretation: A rightward shift of the trendline indicates overall learning gains.
  • Caution: Growth potential is capped for top-level students (e.g., Level 4 → Level 5).

Step 5: Instructional Effectiveness

  • Evaluate: Is instruction working?
  • Findings: Ask AI to characterize your instructional effectiveness based on the data.
  • Typical recommendations:
    Low-level non-progressors need targeted interventions, small-group remediation, formative feedback.
    High-level students (Level 5): Enrichment, extension tasks, peer mentoring.

Step 6: Setting Goals & Target Students

  • Get the current proficiency status.
  • Use the data to set a goal by percentage and student count for the next period (or ask AI to set a goal based on student progress).
  • Example: Determine how many need to move up to reach 75%.
  • Identify which students are most likely candidates to reach the proficiency level.

Step 7: Planning Instruction

  • Focus core instruction on students near the proficiency level (e.g., Level 4) and who are showing improvement.
  • Continue differentiated support for both low-level and top-level groups.
  • Regularly monitor target students to ensure progress toward goals.

Conclusion

This structured data analysis process uses student-level data to:

  • Measure instructional impact
  • Inform differentiated instruction
  • Set measurable growth goals
  • Establish core, general curriculum
  • Support continuous improvement in teaching effectiveness

Transcript

How are your kids doing? What are their proficiency rates? What are their scores looking like? What are their assessment outcomes?

Well, generally when we talk about that, we say, “Well, I got 30 40% of my kids are proficient. Hurrah,” or whatever the case may be. But that’s not really the best question.

The best question is “How well are students responding to the instructional activities in the classrooms or all the things that we’re doing to them and for them and with them. Are they working or not? Are the kids making progress based on those experiences?”

So we are going to look at a little bit of data carefully and try to answer that question and make some decisions that we can use as the educators.

Step 1: Upload Assessment Data for Analysis

Let’s start with the file of data. What is it? It’s initial and nine-week scores.

Prompt

The uploaded file contains students’ initial and 9-week math scores. The possible score range is 0 to 60, broken into 5 equal levels: from levels 1 (lowest) to level 5 (highest). The file has a table with min and max scores for each level. Create a table that lists the students who moved up 2 or more levels, 1 level, 0 level, and went down a level.

And then I say a little bit about the assessment, which obviously you would change based on your assessments. A little bit about the file that I’m uploading.

And then importantly, here’s what I want to know. How many kids have moved up in their scores from one level to the next? How many moved up two levels? How many moved up one level? How many didn’t move up at all? And did anybody go backwards?

Step 2: Changes in Student Proficiency Levels

So, let’s answer that question first. And that’s going to be our starting point for eventually targeting students, identifying interventions, and setting goals. All right. So, this will take just a moment, but I see that it’s already read the file.

Obviously, when you do this, you’re going to have your own assessments. Levels are this, that, and the other. Or maybe your assessment results simply say level one, level two, and whatnot. And the AIs can use that just as just as fine, just as well.

So, kids who moved up one, kids who moved up two, three kids didn’t move any, and a little table here.

Now, this is a bit of a misnomer because we tend to celebrate those kids that made that big jump in growth and say, “Okay, but a lot of kids just made one level. Let’s keep pushing them.” As we’re going to find out when we dig a little more deeply, in some cases that is as much as we could ever expect from those kids, and we’ll figure out why.

Step 3: Graph of Student Gains

So that’s a bunch of numbers, but something I want to turn in, something I want to put in a report. I want to give it to my principal. I want to share it with the parents. I need a visual representation and in this case a histogram is going to work great. Basically how many here? How many here? How many here? So, let’s get our AI to knock that out for us.

Prompt

Create a histogram to show counts for each category of change in level. Label the bars with the student count.

Lickety split. How many jumped two levels? Jumped one. Didn’t change at all. And happily, there won’t be any who went back. So, let’s get our histogram.

Then we’re going to find where this could, if we’re not careful, lead us into a trap as we are looking at our data.

(Come on, histogram! This should be an easy one right here. Bob needs coffee apparently. Oh, there we go.)

In the nick of time. So, there’s my moved up two, moved up one, moved up none. This is a nice visual representation. and tells the story of my data. Very nice.

Step 4: Comparison Bar Graph and Trend line

But like I mentioned, there can be a trap. And here is where we’re going to find it.

So I want a bar graph pre-est levels, post- test levels. Okay, fine. Give me the student counts in there. But I want a trend line, a polynomial trend line because it’s that trend line that’s really going to help represent the shift in student levels over time.

Prompt

Now create a bar graph that shows students per level comparing initial and 9-week scores. Label the bars with student counts. Add polynomial trendlines for each series.

Oh, that was fast.

So orange bars pre, blue bars post. Great. And we can just sort of visually see that there’s been movement going up. But now look at our trend lines.

The dark blue trend line is for the pretest. And you see it bottoms out at zero at the highest level because there wasn’t anybody there. But then the orange one in this case is our post placement. And we see that that has shifted over to the right. That’s what’s going to show us that, overall, there has been progress generally among students in the classroom.

But we know we’ve got some kids that didn’t make any progress.

Now, I did mention in some cases one level jump is the max. We cannot expect any more. You’ll see we’ve got four kids there at the pretest who are at level four. They can only go up to level five. They cannot make more than a one level increase because there’s no higher level.

So when we say, “Gosh, a lot of kids just made one level!” in some cases, that’s the max. That’s as far as they can go. So we have to be very careful about making assumptions about the amount of change over time because it can lead to misunderstanding. We have to look at this to really know what’s happening.

Step 5: Determine Instructional Effectiveness

Let’s bang into four. And here’s the question we want to know.

Describe the efficacy of instruction. Is it working?

And then for those who started low and stayed low, what can we do? And we know we have a few of those kids, but we also have kids at the top. What do we do for them? We can’t say, “Hey, you maxed it out. Go play.” We need to do something, a next step for them. So, what might that be?

And this is kind of a multi-part prompt here.

Prompt

Based on these results, describe the efficacy of the math instruction. Recommend strategies for students who started on a low level and have not progressed in level. Recommend strategies for students who have attained level 5.

What’s the efficacy? What about the low-low kids, and what about the high kids? Because the ones in the middle are doing pretty well, has been broadly effective. That’s great. I’m happy to hear that.

Students with no or low growth, here are some things that we can do for them. Students who reach level five, here is some stuff that we can do for them. Overall, demonstrating positive impact, etc., etc.

So, some good understanding of what all those data mean and what we can do with them. And that then brings us to the last one.

Step 6: Set Goals and Identify Target Students

What percentage of students are at level four or greater? And I’m going to just do this. I’ll tell you the answer. We’ll cheat. The answer is 53%.

Prompt

What percentage of students are at level 4 or greater at 9 weeks?

I want at least 75% of students at level 4 or greater in the next 9 weeks. How many students need to move up to level 4 to achieve this goal? Identify the students who are the most likely candidates based on 9-week score and change in level. Explain why you selected them.

So, while that’s coming up at 53, what do we want to know? Well, 53% is pretty good. Let’s bump it up for the next nine weeks. Let’s say 75% in level four or greater. How many kids are we talking about? And which kids are the most likely candidates?

The reason why we ask this is simple. The kids who are low and stayed low, we need to do something very specific for them. The kids at the top, they’re doing great. They’re already at the highest level.

But when we think about what is our core instruction, it’s that group in the middle.

Step 7: Determine the Core Instruction

Whom do we need push? Whom do we need to adapt our core instruction to? It’s those kids, that group that’s going to push us up to level four and help us achieve our goal. So that becomes our base instruction.

And then some special things on both ends.

The core instruction needs to fit these particular kids that we’re going to identify, and these are the kids we need to watch and monitor most carefully to make sure they are, in fact, going to get over that hump and into level four.

So, who are they and what do we know about them?

We have 23. We need seven more. All right, who are the seven? One, two, three, four, five, six, seven. There they are. right there, plus some things about them.

Why did we choose them? They’re close to the level four cut-off, and they’ve been showing growth over time. So, the instruction is working, and they’re almost there. Those are the kids we’re going to really focus on.

Conclusion

So, there you go. New strategies for data analysis with AI. I hope you found these techniques useful. Take care.