ChatGPT and your edTPA/CalTPA

I'm not saying don't use it. I'm just saying you're probably doing yourself more harm than good.

Hey everyone! Nia here, your neighborhood TPA Coach & teacher trainer. (More at www.ggloweducation.com for 1:1 sessions and task reviews!)

Let’s get into today’s topic: ChatGPT.

I believe that ChatGPT started to get a hold of the teacher-prep world sometime in 2023. Now understand when I say all of this: I’m not the righteous ChatGPT police. Do what you need to get to the finish line.

Do I think that no matter what, you should be able to understand what you’re teaching and why, and that it’s terrifying if you’re steps away from receiving your license but are fully incapable of discussing teaching practices? Well, yeah, of course.

Do I understand that in general though, that’s not the case and a lot of you may be using ChatGPT to help you through really stressful moments? Yes.

So with that understanding, I’m not here to bash you. I’m here to help save you a heap of time and stress, as many of you are steps away from ChatGPT-ing yourself into a failed TPA.

So, what’s the deal?

The problem isn’t that your scorers will realize you’re using ChatGPT and fail you (not yet, at least). The problem is that ChatGPT is giving you guys responses that are crappy, often.

Whether you’re doing the edTPA or the CalTPA (they’re about 90% similar), you have to understand that both are extremely context specific. In order for ChatGPT to even slightly respond to a question correctly, you’d need to feed it a ton of supporting documents and prompt it EXTENSIVELY. There is no quick and easy way to do your TPA with ChatGPT, because you’d need to collaborate with it and keep it on track.

If you’re awesome at analysis, comprehension, and communication, then ChatGPT prompting is probably something you can do well with. If you struggle with those 3 things, *and* you don’t even know what’s required of you for your TPA, then realistically speaking, you will struggle to ask ChatGPT to create something high quality, because you can’t communicate what you don’t know and are unfamiliar with.

(As a random side tip, for those of you whose first language is not English but are required to do your TPA in English, and if you are extremely expressive in your native language, it could be helpful to write in the language you are most expressive in, and ask ChatGPT to help you translate it to English while keeping the key TPA terminology.)

What most of you are doing is giving the AI a sentence or two, maybe even your lesson plans as well, and telling it to respond to the TPA question at hand. That is so severely far from what you’d need to do for ChatGPT to produce a partial or full response that would get above a 1 on the rubric.

Realistically speaking, even properly prompting and editing with ChatGPT would take you 30+ minutes per question just to get something kind of barely feasible, and that’s on the light end.

The Pattern

I’ve noticed that what ChatGPT ends up doing (every. single. time, I might add - I’ve worked with dozens of you recently that have tried it, and I’ve yet to see a ChatGPT commentary that would pass on its own with minimal editing/prompting) is that it latches on to the same two or three key details that either you have given it, or that it has decided makes the most sense. It will then repeat these details ad nauseam, with an intro and conclusion to whatever paragraph it spits out.

Since you’re stressed and tired, you’re thinking “Wow! It made so many words! Awesome!”, deciding to stick with what it gave you. However, at the very least you need to always extensively add to and edit whatever you receive, and only use it as a base.

In addition (and I think this is most important), it literally isn’t answering the questions. It will answer a question, sure, just not the one it’s supposed to.

In the past year and a half, I’ll occasionally have clients come in that want me to review their work, or show me what they have so far so we can work through it together — with one catch: it was entirely written by ChatGPT. They’ll sit quietly, thinking I won’t notice. If the work is good, then it really won’t matter anyway. But if the work isn’t good……

There’s always the silly tells: The random lists, bullet points (please don’t have lists in your TPA commentaries), and paragraphs with bolded headers (quite funny, honestly). But the big tells are the robotic writing that’s painful to read and somehow manages to either evade the key points that needs to be discussed, or evades going into details.

“The lesson plan meets the learning standards and objectives because they leverage hands-on approaches with robust, multi sensory avenues that combine literacy with active learning. It focuses on (insert your skill/focus of choice here) for comprehensive support”.

Those sentences up there? It’s fancy garbage. A whole bunch of nothin’. Now imagine having your entire portfolio be full of that, and only that.

You need to be explaining yourself, making connections, and going into analysis. ChatGPT does a good job at naming a bunch of what’s, but not being able to accurately get into the why’s.

Imagine instead if that above response described the common core standards in detail, connected it to the objectives in a thorough manner, named key activities, why they were chosen and why they were relevant, and how that increases comprehension for students - with tangible examples and real evidence.

The Bottom Line

Think of it this way - there’s some things that are simply intense. Like lawyers preparing for the Bar. You’re not getting around something that is intensive and prepares you for your license.

So what now?

Write as much as you can. Be specific, excessively so. The realness of your words will fare far better against the rigid TPA scoring. Then, and only *if* you feel you need it, you can ask ChatGPT to help you refine your writing, without removing any key details and content, based on the parameters present (rubric scores, etc). I put *if* because you risk ChatGPT making your response worse, not better, about 50% of the time. Therefore, only do this if you are willing to take the risk, or if your analysis skills are good enough for you to catch changes that are unnecessary and manually reverse them.

You’ve got this!