Understanding How AI Works Makes It More Effective in Lesson Planning

If you're like me, you've been inundated with news, opinions and cluttered Canva designs full of bulleted lists about how AI can “transform education.” I knew that I couldn’t ignore this new generative AI phenomenon, so of course I began experimenting.

One symptom of being in the edtech sphere for so long is a deep sense of skepticism, and along with that skepticism is a deep sense of remorse that so many of us educators, myself included, fell for the shiny objects without thinking about the long-term repercussions for student data, privacy, and security. It’s also not clear that many of the tools we ran to adopt had that large of an impact on student learning. This time around, I vowed to take things more slowly and to make sure that I had fully explored some of the tools inside and out, including their privacy policies, before using them with my students.

Once I understood that any large language model (LLM) I used was basically going to regurgitate the internet back to me, I began to think about all of the web searching I had done over the course of my career to find projects and lessons that would help my students engage with and understand the concepts and skills I wanted them to acquire and incorporate into their thinking. What if an LLM could help me with that work?

Thinking Beyond ChatGPT to Plan a Unit

While ChatGPT is often the LLM that most people head to, I came across an LLM known for its better writing style and ability to read uploaded PDFs called Claude.ai. I also learned that it used “constitutional AI” guardrails to guide its behavior and didn’t use users’ chats to train its model like ChatGPT does by default. The last thing I wanted was for my queries and conversations to end up in someone else’s chat with the tool.

I decided to give a wild idea a try: I use a unit planning template from the Understanding by Design (UbD) framework created by Grant Wiggins and Jay McTighe. This backward-design framework calls for identifying the underlying, deeper questions that guide a unit of study and pinpointing the precise skills and understandings students should have by the end of the unit before creating a “transfer task” (project) where students can show what they have learned by engaging with and incorporating their new skills and understandings. I wondered if Claude could help me with those parts of the unit as well.

This past school year, I wanted to hone my art students’ drawing skills. This can be a very stressful unit for students and daunting to teach, so I asked Claude for the top five drawing skills that high school students should have. I reviewed the list and chose a few that felt right to focus on with my students. I was able to plug those into my UbD template. I then asked it to provide five essential drawing concepts for high school students to see what kinds of big-picture thinking would help my students build stronger drawing skills.

I chose a few of those concepts to include in my UbD template. The next step was to develop a transfer task–something students could do that would allow them to show what they had learned throughout the unit. I asked Claude to provide a list of five potential transfer tasks for the unit. Luckily, by now, it had been learning about my unit and could use our previous conversation thread to inform its replies.

I landed on my transfer task and plugged it into my UbD planning template. I uploaded a PDF of the unit so far and told Claude how many class periods I had for the unit and how long each class period was. It gave me a breakdown by day that helped me get started. This was barebones, so I added more instructions–telling it to provide an objective and list of materials for each lesson. By the end of the process, I had my entire UbD completed and my pathway through the unit.

How to Get The Results You Want From the LLM

This was my very first attempt at using Claude. Here is what I learned and did differently the next time around:

• Make sure to keep what you use simple and short. I was so excited by what Claude was pumping out that I used way too much of it.
• Have the LLM cross-check any lesson planning jargon you may use. For example, true essential questions are hard to draft so by adding “Please use the Understanding by Design framework to write the questions” can make sure that the LLM is cross-referencing specific uses of the jargon that align with your goals.
• Using Claude.ai, upload your partially completed unit and ask it for three potential transfer task ideas based on the unit's essential questions, skills, and understandings. Choose and rework one as necessary and enter it into your unit plan. I wish I had done this for my first attempt.
• Upload this new version of the unit plan and ask Claude to create daily lesson plans by telling it how many class periods you have and how long each class period is.
• Revise and adjust this timeline and lessons as needed.

Use AI as a Thought Partner

It’s important to remember that your expertise matters. The internet is full of problematic pedagogy, and replacing your years of experience or content knowledge with the text that an LLM spits out is similarly problematic. Cross-check the information it gives you with sources that you trust. Treat any LLM as a thought partner–a tool to help prod your thinking and get you thinking outside the box.

There are many tools out there that will write lesson plans for you or create worksheets. Be mindful of how these tools work and how these lessons are created. Be critical about whether they’re pedagogically sound and match your own pedagogical approach.

I find that LLMs allow me to save time brainstorming so that I can focus my attention on creating deeper ways for my students to engage with the unit. Rather than generating fill-in-the-blanks or comprehension questions, try asking an LLM to design an activity where students move around the room, or how your content could be connected to the world that students live in.

While unit planning is always iterative, I find that the iterative process feels a little more exciting and faster with the help of an LLM thought-partner. It’s rare that the tool will generate something I can use right away, but it pushes my thinking and helps me focus and streamline my units on what really matters so I can spend more time on designing truly engaging and relevant lessons.