ChatGPT for teachers - Grammar and Language evaluation

  All educators have to correct students texts all the time, and that takes a lot of time, frankly. Today, we will be diving into a use case for educators, and that is using ChatGPT for grammar evaluation. I'm

Christian, Principal Project Manager at appliedAI, and welcome to the video series, How to Master AI Tools at Work. Today, I'll show you how I can use ChatGPT to evaluate the grammar of students texts. That's a very interesting use case for teachers who very often take a lot of time assessing and evaluating texts from students, in particular, their language proficiency.

But before we do that, let's jump into the registration process of ChatGPT.

Okay, head over to the OpenAI website. And what you find on this website is this sign up button. Click it and follow the process.  The first thing that the system prompts is your email address. So I just enter my email address in here  and click the continue button.  I then have to provide the system a password.

Let's use a very simple one for now, but always make sure that you have a very secure password. Let's just go for this one.

Okay,  that's the password. I hit continue.

And the ChatGPT asks you to verify your email address to make sure it's valid. So jump over to the email address itself. I look into my inbox. ChatGPT asks you  In my email provider, I can find the email here. My junk email is always double check if it's in the spam folder or not.  Then I have to verify email.

I click it  and I'm being redirected to the JTPT web page.  Of course, I do have a name. That would be Christian  and a birthday.

And I hit agree.

And boom, that's it. We've signed up for ChatGPT and now you can use the system and become creative and use it for all kinds of use cases.

The ChatGPT UI looks very simple. First, it prompts us and tells us that we should not share sensitive info. We will come to back to that in a second, but for now, let's just click. Okay, let's go.  And let me zoom in a little bit so you can see it clearly.  What you can see here is a little chat box further down below this page, and we'll input all information into that chat box.

But before we do so, let's talk about technique, because you can prompt the system very differently. And one technique that's often used is the so called RTF framework, which stands for role, task, and format. This is just one of many options, and if you're interested in more prompting techniques, you can click the link here above.

Here's the prompt.  And let's talk about the prompt in detail. The first part is the R part of this technique, where we tell the system what kind of role it has. And here we say, you are an expert in education and a language teacher.  Fantastic. Then we need the task itself, what the system is supposed to do.

In this prompt, we tell it, you will evaluate the grammar and language uses of students homework that I will provide to you. This would suffice. This would be enough. But.  Very often you want to be more specific. For example, we tell the system, your feedback should be directed or tailored to a 10 year old student.

Of course, your context might be different. Your students might be older or younger or some other context. Just be very specific on the kind of thing you are looking for. Also, I'm telling the system here that it should be factual and not use words of encouragement. Why? Because the feedback is not. given to students.

That's the task part of the RTF framework. There are different ones we can use. We can prompt the system to output bullet points. It could be a poem. It could be a text file just as JSON and many different things. But for now, let's stick with the idea of tables. It should output a table.  That has a few columns.

The first column, uh, the corrections or the actual texts and sentences from the students. Then we have a column about the language improvements, where the system identifies the problems the students had. And the last column is for specific feedback, how we could improve and help students write better texts.

That's the prompt. So we've established RTF, the role, the task and the format. And now that we're ready, we can hit the enter button and see what happens. Thanks.  What you get is a very short response and the system tells us that  it is happy to provide feedback to us and it's waiting for the student's texts.

Let's have a closer look at the table. In particular, sentence three. Here it says, Then I go to the park with my friends. And it should say, Then I went to the park with my friends. And that's also the feedback we get from the system. Or further down this table, we find two other mistakes. Here the student said, we ate together in the table.

Well, that doesn't sound like it's the right preposition. And that's also the correction. So it should say, we ate together at the table, which is the feedback we get here. And finally, we've got this sentence. Finally, before bed, I read a book. And  this should be, of course, read a book with an A. And that's also the feedback we get.

As you can see. We get very specific feedback on the language use of our students. And if you have a larger class of many students, you would input their texts sequentially into the system. This is just one of them. And if you have another text, you just go up there, you copy the other text, put it into the system, and then you will get a different table with different correction results.

Trust is a very important topic for us at appliedAI. It's the mainstay of a company, and there are reasons for that.  If we have another look at ChatGPT and go into the settings, we find that ChatGPT in OpenAI is actually using our data to train their models. That's their default option. And we can opt out of that option if we wanted to.

So if you don't want your data to be trained for their systems and for their models, you click this toggle button here and then your data won't be stored and won't be used for them to train the models. Thanks. So you have more control over your data. Keep in mind though, that these settings are only stored for your particular browser or your system.

So whenever you log in with a different computer or with another browser, you have to deselect  that feature again, and opt out. There are more things to discuss when it comes to data usage and OpenAI. And if you want to have more information about this, have a look at the link below this video. There are other issues we need to consider when it comes to trustworthy AI, and one of them is data privacy.

We do have GDPR regulations, and we need to make sure that we treat personal data carefully. And it can happen that your text contains names. Emails or other personal information. And you want to double check if you actually insert that information into the system. So be very aware of the kind of data that you input into the system and remove it if necessary.

Then we have this other topic of transparency and school policy.  You are working in a educational institution and your school might have particular policies as to how we can use Gen AI tools or ChatGPT for your work. So it's good practice to align with your school of how we can use these systems for your craft and for your work.

As a responsible educator, try to ask yourself, should I disclose the use of these systems to my students and others in your institution? And there are also other interesting problems that we have to tackle. If we move back. And then we have this topic of overcorrection  here. It says, yesterday was super fun, exclamation mark, and you can also see that in the corrected version, we don't have that exclamation mark anymore.

It replaced it with a dot. That's clearly a case of overcorrection and sometimes that's what ChatGPT does. It over corrects what we're looking for. We didn't prompt the system to evaluate the style of the text, only the grammar, but still it corrected the style to a certain extent. Be aware that this problem exists and make sure that you are the human in the loop.

Meaning that you're in charge of what happens to the feedback. And that's the last part when it comes to trustworthy AI. Always make sure that you double check the grammar corrections. You have a certain expertise and you should use your expertise if you're using the system. This is maybe a side partner.

That you have that gives you some hints and ideas that expedites your process, but don't leave your expertise at the door. Use it and never fully, fully trust the system. The most important message, really, when it comes to trustworthiness and risk is that Keep your expertise. You're in charge. We always want to have a human in the loop using these systems because ChatGPT hallucinates sometimes and comes up with things which are just factually not correct.

So make sure that you decide what feedback is being given to the students and use your expertise wisely. at your best.

For this use case, there are also alternative tools. There's, for example, BART by Google. There's Bing by Microsoft or Lama by Meta. Choose the tool that you're most comfortable with.  That's a wrap. Thanks so much for tuning in and we hope you found this video helpful. If you want to have more of that kind of content, hit the like button, hit subscribe, and also share your insights and feedback in the comment section below.

Thanks a bunch. See you next time.