MAKING IT EASIER TO BE A BETTER TEACHER

Noteworthy News

Here’s our weekly roundup of interesting education stories from around the world. Click each link to learn more:

• Universities are being told to accept less private school kids - why the sudden push?

• Are your students always absent? This NFL team is officially trying to solve the issue.

• 3 bears wreck havoc on this school - and no, Goldilocks had nothing to do with it.

TECH TALK

Not Even AI Can Help Our Students with Math

It is clear that our students will one day be working with AI; but up until now, it’s struggled greatly in helping learners with one particular subject - Math.

Current versions of the world’s most famous AI model - ChatGPT - famously struggle to correctly provide help with math equations.

Well, that officially changed last week thanks to their updated model, codenamed Strawberry. 

The Solution: OpenAI o1 (Strawberry)

On September 12th, OpenAI (ChatGPT’s parent company) launched a series of new models that think before they respond. Well, they don’t actually think. But they do take their time to reason through more complex questions.

We covered the last update only back in May (click here to read), so considering we’re just in September and already at the next major version, it’s clear that these models are evolving quickly.

It is a little creepy - but the reason for the change is because previous models were answering certain math and science problems incorrectly. This meant that students who were using it for extra help were also being led astray.

Why the extra time?

Besides trying to make it seem like real conversation, the extra pauses allow the models to try different strategies (in the background) and recognize its own mistakes before replying.

GPT-4o, the previous model, isn’t going away. OpenAI is saying to use that model for the usual quick tasks and browsing. Strawberry is reserved just for complex tasks and questions.

How About YOUR Classroom?

To give you an idea of what’s been improved - the previous model (GPT-4o) only solved 13% of math questions correctly in the International Mathematics Olympiad exam; this new model, although it takes longer, is able to score 83%.

If you’ve got the Plus version of ChatGPT, you’re only allowed 30 messages per week (for now); simply select it in the model picker. Note: ChatGPT Plus will cost you $20 USD/month. If your school foots the bill though (which many are starting to), it’s an interesting evolution to try.

We’re not saying in any way shape or form to use this to replace your lessons - but it can be a useful math practice tool for students.

Here’s how:

Strategies:

No matter a teacher’s personal opinions on it, it’s hard to argue against the fact that our students will be utilizing AI tools in their not-too-distant futures. We owe it to them, as their teachers, to guide them on how to properly and ethically do so - no matter how crazy each upgrade gets.

BRAINY BITS

The Ethics of AI - Statistically Speaking

As AI models rapidly develop - along with the many fears that surround it - teachers, parents, and students alike are concerned about the ethics of those responsible for coding the very models we use.

The Study: The Ethical Wisdom of AI Developers

No one will ask a developer how moral issues were considered when creating any product. To find out why, this week’s paper explores what ethical wisdom exists in the AI developer community.

Very little research has targeted this community before. This experiment included semi-structured interviews with 40 developers from around the world who were actively involved in AI development throughout 2022. 

The interviews asked these developers questions on awareness of broader ethical issues, how they navigate these challenges at work, and about any moral barriers they’ve encountered in developing their models.

The Results:

Researchers found that AI developers are aware of the ethics they must navigate, but are faced with inconsistent resources and a heavy lack of any moral guidance and training.

The industry is fixed on rapid innovation at the moment, which leaves no time for reflection, dialogue, and ways to incentivize developers to take appropriate training. 

This  concern is brought up in their interview results. Here are their most important data takeaways for educators:

• 7.5% have left previous employers because of ethical concerns.

• 67.5% admitted that they have worked with models that had unclear ethical motives.

• Most developers look to other team members to discuss ethics of their project.

• 2.5% said that they would look to the impacted community (like teachers) for guidance.

• 10% said they would look to help from moral/ethical experts.

• 10% reported to experience mental distress due to ethical concerns at work.

In Your Classroom:

As AI integrates itself more and more into our schools - and the lives of our students - it is our duty to understand how and where these models come from.

The same way you wouldn’t choose a novel for your class without first knowing the story, we should be approaching AI with a similar precaution. 

Here are some ways to tackle these classroom concerns:

Strategies

First, guide your students on how to actually use AI ethically and efficiently. This might mean teaching ourselves first. AI tools are not meant to replace the way we think, but accelerate and improve the work produced. 

The output of any AI model will only be as good as what we input into it. This is an important idea to teach to help curb using these tools just to cheat and find answers.

Second, especially for middle school and above, bring these ethical discussions to your classrooms. Don’t simply state that ‘using ChatGPT is bad/good’; instead, showcase the critical thinking skills they will need to come to that conclusion themselves.

For example, evaluating the ethical use of ChatGPT in various careers. This can be turned into a great assessment that can help students better understand how to use these tools, but also enforce critical thinking development.

No matter the method, studies like this show us the ethics surrounding AI don’t start and end in the classroom; they start written in the very code of these models. Just like we would examine the root of any other classroom tool, it is important for teachers to stay up to date with emerging AI models as well.

REFERENCES

This week’s issue adapts information from the following sources:

Tech Talk:

OpenAI. (2024). Introducing OpenAI o1-preview. Retrieved from https://openai.com/index/introducing-openai-o1-preview/

Brainy Bits:

Griffin, T.A., Green, B.P. & Welie, J.V. The ethical wisdom of AI developers. AI Ethics (2024). https://doi.org/10.1007/s43681-024-00458-x