From Homework To Practice Assignments - GenAI in PHY 1000

Practice Assignments with Structured AI Support

I recently shared my own AI-adapted assignment structure with colleagues at the for inclusion in their . I really enjoyed the structured way they are capturing various ideas, so here I have recreated what they did in .

You’ll see my thoughts on the structure and purpose of these assignments, my student-facing “Generative AI and This Course” language and then a detailed breakdown of how and why I created this structure.

A big thanks to Helen MacDermott and JT Torres!

Faculty Reflection

In designing these Practice Assignments, I wanted to solve a fundamental challenge: how can students use generative AI to genuinely enhance their learning without it becoming a crutch that prevents actual understanding?

The three-step structure emerged from recognizing that the key to learning is repetition with feedback. By forcing students to first attempt problems independently, then explicitly seek help (whether from AI, peers, or instructors), and finally reflect on what changed in their understanding, students develop both physics knowledge and metacognitive awareness.

The [FLAG] system ensures I can provide targeted feedback on the most meaningful learning moments while keeping grading sustainable.

Students consistently report that this structure helps them distinguish between problems they truly understand versus ones where they’re just following steps.

What Students Create

Weekly practice assignments demonstrating conceptual physics understanding through a three-step process: independent attempt, getting help (including strategic AI use), and reflective revision with flagged responses for instructor feedback.

Why This Assignment Matters

Students learn to use AI as a learning tool rather than an answer generator, developing both physics reasoning skills and metacognitive awareness of their own learning process. The structured approach ensures repetition with feedback鈥攖he foundation of genuine learning.

The Three Steps at a Glance

First Pass 鈥� Independent attempt, no help and no consequences for failure.
Get Help 鈥� From the instructor, peers, resources, or AI via a Learning Prompt.
Final Version 鈥� Revised answers, with reflections on three responses marked [FLAG] for instructor feedback.

Generative AI and This Course

Course-level AI statement shared with students.

What is the purpose of a class like this?

There is a General Studies requirement that every student must take a course in the Natural and Physical Sciences category, and that is the most common reason for students to take this class. General studies programs are extremely common across all colleges and universities in America, and I agree with the idea and goals of those programs wholeheartedly. From that perspective, you are taking this class in order to build a firm foundation in skills and content that will prepare you to learn, experience, and grow throughout life.

What you gain from this course isn’t intended to be specific skills that you use in narrow circumstances. Instead I hope you will not only learn interesting physics concepts, but also improve your ability to reason, problem solve, shift perspectives, and understand the world around you.

How does generative AI fit into a class like this?

At this point in human history, it is possible to use generative AI tools, primarily Large Language Models (LLMs). This is incredibly new territory, and we are all working to understand if and how we should use these tools, including the ethical implications. For this course, here is my broad perspective on using generative AI tools:

Generative AI can help you learn. If you use good prompts, I think the gAI chatbots can definitely help you learn the physics concepts we discuss. When done well, I even think it can help you learn the thinking patterns and problem solving skills that are just as important as the concepts and facts in this class. Using good prompts is really important here and I will do my best to give effective guidance.

Generative AI can ruin your learning. The presence of free generative AI tools while you are trying to learn can feel like having a free robotic junk food vending machine right outside the front door of your home. It isn’t a perfect analogy, but the temptation to go with the fast, free, easy “solution” is very powerful. Throughout this course, I’m going to ask you questions that I think will either help you make progress toward learning or that are intended to assess what you have learned. If you were to take those questions and simply ask a generative AI tool for the answers, you could come away thinking you had learned, but I think that would be an illusion.

Part of my role is to help you make good choices about when and how to use generative AI. Obviously, my education and training for this job didn’t include “how to use generative AI to improve learning without ruining it.” But as we all grapple with the way these tools are going to change education, I do think it is part of my job to help you use these tools well and to put some barriers and friction in place to make it harder for these tools to ruin your learning.

Guidelines by course component

WarmUps 鈥� No gAI (馃馃毇). These are learning activities where you reflect on the learning resources you have read and try to apply that knowledge to a few questions the night before we talk about them in class. You don’t have to get the questions right to get credit and what I really want is to see your genuine attempt at answering the question using your own understanding. Insight into what you and your peers already understand or are struggling to figure out allows me to plan and adjust accordingly. For these reasons, these activities should be done without help from generative AI (you don’t even need to use a tool that will correct your grammar!). But if you want to talk this through with generative AI after you submit your WarmUp, try using one of the learning prompts I’ve shared.
Discussion Questions 鈥� No gAI (馃馃毇). These are questions you answer and discuss during class time. You don’t have to get the questions right, just participate in the majority of the questions each day. Grappling with new questions along with other people that are learning the same things is a fantastic way to learn, and there is no need to bring generative AI into this process. These should be answered without using generative AI.
Practice Assignments 鈥� gAI allowed in Step 2 (馃–>2锔忊儯馃憖). When you do the Practice Assignments in this class there is a middle step where you should get help on any questions that gave you trouble. You are welcome to use generative AI tools as the source of that help, but you need to use them in the right way. Practice Assignments have three steps, and you will submit the assignment itself at least twice. (This is the assignment on this page 鈥� the three steps, the submission checkboxes, and the grading are all detailed below.)
Connection Activities (馃–>2锔忊儯馃憖). The first draft of a Connection Activity should be completed without the use of gAI. After that, I would encourage you to consult with other people or with gAI to make sure you have responded to the questions that were asked and that your responses are easy to read and follow. There will be a place for you to disclose what kind of help you got.
Exams 鈥� No gAI (馃馃毇). Summative assessments are the ways that learners demonstrate their knowledge, skills and abilities. For a General Studies class like this one, it is your personal perspectives, knowledge and reasoning that needs to be assessed, rather than a more practical application in which perhaps you should use all the tools available to you. For that reason you will demonstrate your learning using in-class Exams, and no tools other than a simple calculator will be used.

Step 1: First Pass

Assignment Instructions

The first thing you will do in each Practice Assignment is check a few boxes at the top to indicate how the assignment has gone so far, and (when relevant) how you got help in Part 2.

The seven checkbox options:

This is my first submission on this Practice Assignment.
This is a re-submission. I had the wrong idea on 3 or more items in my first submission and then I got help on those, as indicated below. For at least 3 of the initially incorrect questions I have put “[FLAG]” at the start of the explanations and then described what I learned and/or why I changed my answer.
This is a re-submission. I had the wrong idea on 2 or fewer items in my first submission and then I got help on those, as indicated below. I have put “[FLAG]” at the start of 3 item explanations. For questions I initially got wrong, I have described what I learned. For others I have added details about why I think the item was difficult or interesting.
After my first submission, I got help from Prof. Loats.
After my first submission, I got help from other people, such as classmates, friends, family, etc.
After my first submission, I got help by using a Learning Prompt and a generative AI tool.
After my first submission, I learned more using passive resources like websites, videos or textbooks.

On this first pass, that is just the box saying this is your first submission.

Step 1: First Pass (student instructions)

Give yourself at least a few minutes on each question, but not too much time (don’t spend more than 30 minutes total). Answer each question and describe your reasoning, if you can. If you find that you are just guessing, just say so. You will be able to view your responses later.

Be sure to get this first pass done several days before the Practice Assignment is due.

AI Permissions

None. This step is about discovering what you actually know without any external help. No AI, no resources beyond your own thinking. If you’re guessing, say so鈥攖hat’s valuable information.

Faculty Notes

Why no help on the first pass?

Learners need a grounded view of what they do and don’t understand at the start. Getting help at the start (from people or GenAI) will hide the real gaps in understanding, leading to the illusion of learning.

How I make it worth doing

The first pass is a formative assessment, purely there as a diagnostic for the learner and a baseline for the instructor. Getting learners to engage with these ahead of time is tough, I’ve found that repeated reminders by me in class, calendar reminders in Canvas, and an Announcement for the first few times is mostly enough. If students do only the first pass by the due date, I am generous with offering/insisting that they do step 2 and step 3 (for full credit).

What to keep if you adapt this

In my model, nearly all questions are Canvas fill-in-the-blank questions with dropdown choices followed by a free response field for which any text counts as correct. Converting questions to that structure can be weird, but it means that Canvas gives students an accurate read on what they got right or wrong, which I think is key. If instructor grading is needed after the first pass, I think this probably won’t work.

Step 2: Get Help

Step 2: Get Help (student instructions)

Assignment Instructions

Now that you have a sense of your own understanding of the questions, it is time to either get help or solidify understanding. You can get help from me (which I encourage), by talking them through with others (especially classmates), by finding additional online resources, or by using a Learning Prompt and asking a generative AI tool about the question.

NOTE: If you got all the questions right and you are ready to indicate which are the most interesting or difficult, you can skip this part.

AI Permissions

Permitted, but only through the course Learning Prompts. Two are provided:

Semi-Socratic Tutor 鈥� guides your thinking one question at a time and never hands you the final answer.
Misconception Hunting Tutor 鈥� diagnoses the misconception behind your confusion, then fixes it directly.

Faculty Notes

Why these prompts work better than just asking AI

Generative AI tools are generally sycophantic (they try to please and flatter the user) and are not good at helping people think their way to an answer. These prompts will get you what you need and make sure you are more prepared for future questions on the same topic.

Writing a learning prompt for your course

I adapted these from sources that I found online, but they are NOT carefully studied or stress-tested. Those students that used these never offered commentary one way or another. Some personal testing or literature searches are probably needed.

When the Socratic prompt frustrates students

One perk of going to GenAI for help is that the LLM will never get frustrated or exhausted. The people-pleasing nature of the tools means that they are not likely to keep students frustrated for long.

Step 3: Final Version

Step 3: Final Version (student instructions)

Assignment Instructions

As you do the Practice Assignment again, update your answers based on what you learned. You can do this as many times as you want and I will review your final submission. In your explanations, tell me your previous answer, your new answer, and the physics reasons for why you changed your answer.

In addition put “[FLAG]” at the start of three of your explanations, so I know which responses you would like me to read. You should [FLAG] responses where you learned something new during Part 2 and now have a different answer or explanation than you did before. If you got all the questions right in Part 1, you will instead [FLAG] questions where you tell me about why the question was either difficult or interesting.

Before you re-submit, update the boxes at the top of the assignment to indicate that this is a re-submission and how you got help in Part 2.

AI Permissions

You may use AI to help clarify your explanations or check grammar, but the ideas and physics reasoning must be yours.

Important: AI cannot write your reflections on what you learned—those must come from your genuine experience.

Deliverables

Updated answers for all questions
Explanations showing previous answer, new answer, and physics reasoning
Three [FLAG] markers on most significant learning moments
Completed checkboxes about process and help sources

Faculty Notes

Why only three [FLAG]s

“Grading” 3 questions is an unexamined habit from the way these assignments used to work on paper. Just 2 might work. A few more would probably be a bit better, if the instructor has bandwidth for it. With 3 [FLAG]s to check, I find that grading and feedback take me just a few minutes per student.

What the flagged explanations reveal

As with many metacognitive exercises, the results can be both really important for the learner, and somewhat obvious to the expert. Because I’m experienced in teaching this topic, I’m rarely surprised by which questions students get stuck on. But, as always, it is far more important for the student to realize and state it themselves than for me to point it out.

Why students report how they got help

I believe that near-universal disclosure is key to navigating the wicked problem of GenAI in higher ed. I want disclosure everywhere. From myself, from my students, from my colleagues, etc. I am not sure this disclosure changes student behavior, but as we wade through a world of AI slop, I want to emphasize that these tools can be used in transparent ways.

Grading (student instructions)

The points you earn on Practice Assignments come from having done the three-step process. For the questions that you specifically [FLAG], I will read about your process and offer some feedback.

My honest expectation is that you will get nearly full credit on Practice Assignments, unless you skip some of the steps, or leave some questions with wrong answers. If you submit the assignment without engaging in all three steps, you will lose at least 5 out of 10.

Practice Assignments count for 10% of your total grade.

Learning Prompts (student instructions)

Use prompts like these with generative AI (LLMs) to enhance your learning, rather than losing out on the chance to improve. Copy a prompt into your AI tool of choice, then tell it which question you’re working on.

Semi-Socratic Tutor

This will really push you to be the one driving the learning. It won’t give you answers and will always be asking you to think and explain. The downside is that this can be frustrating if you are really stuck.
Adapted from openNCCC, I worked with Google Gemini to draft this and then edited it to add context.

The prompt

I’m taking a college-level conceptual physics course called Physics of Nature. The course focuses heavily on concepts and ideas, with minimal calculation and heavy use of proportional reasoning. You are my conceptual physics tutor. Ask me one question at a time and help me learn by guiding my thinking (Socratic style). Don’t give me the final answer.

Start by asking what topic or question I’m working on and what I already understand. Then explain using simple examples/analogies and keep checking my understanding with questions.

If I’m stuck, give a small hint and ask the next question.

At the end, ask me to explain the idea in my own words.

Misconception Hunting Tutor

This comes from years of research on the idea that confronting misconceptions is key to learning new ideas, especially in physics.

I worked with Google Gemini to draft this prompt and then edited it.

The prompt

I’m taking a college-level conceptual physics course called Physics of Nature. The course focuses heavily on concepts and ideas, with minimal calculation and heavy use of proportional reasoning. You are my conceptual physics tutor and your job is to help me find any misconceptions behind my confusion.

Ask 2-4 short diagnostic questions (one at a time). After each answer, tell me what it suggests I’m misunderstanding.

Then give a short explanation that directly fixes that misconception, plus one quick “check” question.

First, ask me what question I am working on.

Notes and Disclosures

Generative AI disclosure:

After working with the folks at Washington and Lee University, I used Claude to help me create a similar post in 爱液A片’s WordPress platform. None of the text or ideas is AI-generated, but it was helpful for formatting.

After writing this piece I used generative AI to write a first draft of the short 鈥渢easer blurb鈥� that went out by email. Want to know more? Send me an email and we can chat!

爱液A片