A quick guide for NCEA Maths/Science students on how to use ChatGPT to supercharge their study.

The most effective studying method for NCEA students is, and always has been: doing practice exams (bonus points if it’s under time pressure), checking your answers, and adjusting your approach depending on how you performed.

The purpose of this guide is to show you how to incorporate ChatGPT (and other Large Language Models – ‘LLMs’ for short) into that study method.

Why bother? Well, ChatGPT (and other LLMs) are very good at giving you the reasoning behind an answer, and setting it out in a very understandable manner (as best they can anyways). By doing this, they help you avoid the frustration of having no way of figuring out where you went wrong on a question. They're nearly like a tutor, of sorts.

They don’t always nail it, but they’re a lot better than nothing. And can save you a lot of time and frustration.

In this quick guide, I’ll show you show to use LLMs to supercharge your NCEA Maths and Science study.

Is my Math/Science problem appropriate to give to an LLM to solve?

LLMs can deal with a pretty wide range of problems, with decent results.

Because LLMs can process photos/screenshots of problems, you can give it problems which have primarily visual (non-word) elements. An example of that is below:

Something to note is that you should only use an LLM if you have access to the correct answer already. This is so that you can audit the LLM’s response. LLM’s aren’t always right, and Math is one of the LLMs’ weaker subjects. Where LLMs add so much value is with helping you understand where you went wrong – not with finding the answer itself.

Also, make sure you attempt the problem itself before giving it to the LLM. Even if you have no clue and make a complete dog’s breakfast of it, you’ll still learn more if you have a go yourself, first.

In my experience, LLMs perform poorly on Level 3 Calculus papers (Differentiation, Integration and Complex Numbers). I’d suggest that LLMs are best used for NCEA Level 1 and 2 papers.

Outside of that - give the LLM a go. It costs nothing except a minute or so. The flowchart below summarises when it's appropriate to use an LLM.

Firstly, which LLM should I use?

Any of the most popular free LLMs should do the trick. For free and easily accessible LLMs, you could use:

• Google Gemini – This would be my first pick, only because there are no limits on the number of messages you can send or images you can upload.

• ChatGPT – Limited to 2 image uploads per 24 hours.

• Claude.ai – Limited to 6 messages every 4 hours.

The graph below shows how different LLMs compare in terms of their ability to solve complex Math problems (June 2024). Given how quickly LLMs are learning, I don’t know how reliable that data still is.

Finally, if your problem is just a raw mathematical equation (no visual/word elements), best results will be obtained from using a tool like https://math.microsoft.com/en or Wolfram Alpha.

Quick note: There are many LLMs out there which claim to be tailored specifically to solving Math questions. I’ve played around with heaps of these and can say with certainty that they don’t perform any better than the general LLMs listed above.

The process

Step 1: Give the problem to the LLM

Either take a screenshot of the Maths/Science problem and ask the LLM to solve.

OR

Take a photo of your attempt at the problem (including your working), give the LLM the correct answer, and ask the LLM to tell you where you went wrong. An example of this is below:

Step 2: Audit the response

This needs to be done every time you use an LLM. We can’t (yet) rely on LLMs to be accurate all of the time. The success rate is closer to 76.6% (see the LLM comparison graph above).

To do this audit, check that the LLM’s:

1.       inputs are correct (see ‘A’ on the image below); and

2.       outputs reflect your expected answer (see ‘B’ on the image below).

Once you’re satisfied that the LLM has passed the audit, you can analyse the LLM’s method to see how it reached its answer. Ask yourself how you can adjust your own method/approach accordingly, for the next time you try a similar question.

Step 3: Keep pressing (optional)

If you’re still stuck, there are a bunch of different additional ‘prompts’ you can give to the LLM to help you.

For example:

• “Show me a different method for solving this problem."

• “Explain this particular part of your answer in more detail: [ ].”

• “Break your method down into steps so it is easier for me to understand.”

If you're still getting nowhere, and willing to look into getting one-on-one tuition (from a human being), get in touch with us. We provide one-on-one NCEA tuition for Christchurch students across all school subjects.

Summary

In this blog I’ve shown you a quick way to incorporate LLMs into your study process for Maths and Science.

Study aside - AI/LLM literacy is becoming an increasingly marketable skill as workplaces become more hybrid (powered by both AI and human personnel). Utilising LLMs to become more efficient learners is a skill that all students should be developing.