Analyzing Brazil's PISA Report with AI 📊

I'm from Brazil So I asked both models and in the prompts I'm going to share everything with you.

ask you to analyze and provide some insights and to an action plan to improve Brazil's results.

let's start.

With Claude's Fable 5.1.

this is it that you are seeing here.

So the first two pages it's showcasing for us or highlighting the results in mathematics, reading and science showing Brazil versus OECD average.

Brazil is way below on everything, mostly in math, and also highlights that 72% of Brazilian students below the basic level in math, 51% in reading and 55% skipped a full day of school in the two weeks before the test.

All much worse than when compared to OECD average.

The second slide will talk about some evolution throughout the years showing pretty much that Brazil is not evolving quite so well, especially in the past 10 years.

And it shows here comparison versus other countries for math and reading.

if we are to show the same two first slides for the GPT6 Astra, this is what it shows.

So it tells us the learning gap starts with the basics students below pisa's basic proof, level in math and reading the same that we saw before telling us that results broadly unchanged since 2015.

And the second slide showed an evolution chart from 2015 not from the 2000s and a comparison between main countries here like benchmarks Singapore, Japan, Estonia and the average pretty much telling us the same thing, right?

way below the average and we haven't improved throughout the the years.

To be honest, so far I think GPT6 Astra did a better job at telling us with less information what is going on.

Now let's compare the action plan both of them have come up with.

So on the left side we have and on the right side here we have GPT.

And you can see the first difference is.

Claude's action plan contains 8 points while only 5 on GPT.

you can see for yourself the points.

Arguably is more comprehensive in the plan.

it talks about attendance.

GPT also talks about attendance.

It talks about having a target one floor targets that is like minimal accepted for the basic level in a way.

GPT is also talking about the basics but not necessarily as a target.

This is kind of unique for Claude which is a daily deep reading of long tests on paper.

And it says that reading trends move with math trends.

There's a strong correlation for lack of better word here of 80%.

So if we are constantly improving the reading capabilities that will also reflect on better math results for the kids.

The number four here is very similar to the number three which is diagnosing catch up in math grade six to nine.

and this is like a tutor or coach of small groups.

The difference is the teacher feedback.

And in all of them you're going to see like a why this step which is the evidence both models use to come up with that idea.

Money and best teachers to the poorest quarter barter.

This is something that ZPT did not suggested.

Claudes, this is not really an action point for me.

Number six is like count the 15 year olds that Pisa never says which is 25% or not in the score.

Let me explain this.

So only 76% of Brazil's 15 years old teenagers are in grade 7 or higher.

So Pisa will test then the rest are not in the score.

That's way below the average here.

So this is pretty much telling us how important it is between grades 1 to 5, catch up and fix the issues of those kids before they start to to fail soon as they are failing soon.

Interesting.

This is kind of not so obvious and something that GPT also didn't do.

So the family and growth mindset.

this is somehow in the number four here as well.

So let's go to the number seven.

So they're both talking about the importance of attendance and communication to and from the parents, from the school to the parents as well.

that's very interesting.

I like the Claude's better.

The measured and improvements in publishing both of them have mentioned as well.

So overall I would say that I liked the comprehensive guide from Claude's better but I like GPT approach to try to summarize it and give it to me in a way that is really actionable and everybody can understand.

And there's no room for errors, no excuses.

But do you think I'm gonna make everything available that I did like the prompts which were quite simple.

This was a, let's say, entry level, back and forth and those PDFs as well.

you are interested, just comment Pisa, down below and thank you for watching.