1 of 15

5 Tips for Using Generative AI with Students�

https://bit.ly/PC-AI-5-Tips

2 of 15

5 Tips…and More

Abstinence Is Not an Option
The 2 Bucket Challenge
Red-Teaming
Chatlog as Submission
SKA + AI + D = Course�

Bonus Tip: https://bit.ly/PC-AI-5-Tips

Image Source: ChatGPT-4

3 of 15

Tip #1

Abstinence Is Not an Option

4 of 15

Avoidance is not a strategy

Image Source: DALL-E 3 with ChatGPT4

Abstinence has never been a great strategy. It didn’t work for prohibition, the war on drugs, or abstinence-only sex education, it’s not going to work for generative AI.

That means we have to learn about it, play with it, and yes, think about it in our classes. Banning it or dismissing it is not going to do a thing other than delegitimize ourselves to our students. Something that higher ed is already doing in buckets.

That doesn’t mean you can’t have guardrails–you should; you should have clarity about how it may or can show up in a course and you want to talk about it and surface concerns and interesting use cases. But don’t avoid it.

Prepare ways of how you might engage with it and even help students critically examine it, but avoid it is like trying to ban Wikipedia in the late 2000s.

5 of 15

Tip #2

The 2 Bucket Challenge

6 of 15

We need 2 buckets

Source: DALLE-3 with ChatGPT 4

Now, it’s great that AI can produce content but that’s also a big concern.

There’s one definitive fact we know about generative AI for the foreseeable future. It calculates the probability of responses but it does not know the answers. It does not have grounded knowledge like humans do. Therefore, in order to use it well–really well–the kind that makes it useful is that users need two buckets.

The first bucket is a working understanding of how generative AI works; its limitations, its capabilities, its biases, and how it fits into the larger cultural landscape of late-stage capitalism. That’s essential to knowing what it is and isn’t, can and can’t do–but more importantly, sifting through the promises of a lot tech-bros in Silicon valley. We can’t anthropomorphize it and give it over our agency. We have to understand it well enough to know what it can’t do.

The second bucket is a deep knowledge of the subject matter you’re working with it on. It just doesn’t know things; it only calculates responses and that can only get one so far. When it comes to disciplinary knowledge, it still falls drastically short. Yes, it can produce facts but it often can’t produce knowledge.

So anything we can do to help students learn and experience that is important and essential. It’s important for them so they can understand how and where best to deploy it and to help them how to use it best for their learning.

7 of 15

Tip #3

Red-Teaming

8 of 15

Red-Teaming

Examples

Instructor shares content�
Students create content�
Explore contents out in the “wild”

Benefits

AI patterns�
Discernment of where AI goes�
Critical thinking�

So what is red-teaming?

You may have heard this term in relation to the military, hacking, or even journalism. The goal of a red team is to work hard to find weaknesses in something.

With regards to AI, the goal is to have students be the red-teams to figure out and determine what is accurate, right, or problematic with AI outputs.

You can do this several ways.

First, as the instructor, you can continually bring in examples or embed them through the course to get students to see if they can discover it and what kind of critiques they can bring.

You can also have students create content with generative AI–maybe even assignment submissions–that must first be vetted and critiqued by other students for questionable AI content.

Finally, you can challenge students to go out into the wild online and explore blogs, social media and other content that might be course related to determine or make assessments of why something is or isn’t generative AI.

There are some great benefits to this. First, is they begin to learn and see patterns in AI outputs to which can make them more suspect or just discern more about how and where and why to use generative AI.

But also, this is an incredibly useful way of building critical thinking. It’s one of the things that I think can come out of this–we are more critical of everything we encounter rather than less.

9 of 15

Tip #4

Some Use Cases

10 of 15

Some Use Cases

Cite chatlog�
AI as Refiner�
AI as Interviewer�
AI as Reflection

Rather than resist or find ways to keep students from using it–consider how you might not only have them use it but what could come from it.

First is that you can encourage them to cite and link to the chatlog so you can see and learn what they are using and how they are using it.

But you can also lean into encouraging students to use it as a means of refining their work–in style, clarity, or ideas. You can have them work through a few prompts to help them clarify their ideas, a proposal, their main idea, research question, etc. Again, they can share the chatlog to help you get a sense of where they work has been supported by the AI.

But where it gets really interesting is when you develop and have students use prompts that turn the AI into an interviewer. This is where I think real magic can happen. The AI as interviewer means students will be engaging in a dialogue that allows them to share their ideas and for the AI to keep asking questions to help them figure things out.

In particular, creating one that is focused on reflection can be really powerful. Reflection is a practice–it takes time to develop for many of us. Many students stumble when it comes to reflection, offering superficial responses than what we are hoping to see. Creating an AI reflection prompt that digs deeper into student’s responses can help them become better at reflection. Rather than trying to do it with a blank page and problem fudge their way through it–the chatlog can demonstrate how the student dug deeper with appropriate nudges from the AI.

11 of 15

Prompt #1: AI as Interviewer

You are an AI assistant designed to help students develop their project ideas through a series of thoughtful questions. Your role is to guide the student in exploring their interests, refining their ideas, and formulating a thesis and outline for their project. Remember, your primary function is to ask questions and encourage the student to provide answers, rather than offering direct suggestions or solutions.

12 of 15

Prompt #2: AI as Reflection Tool

You are an AI assistant designed to help students reflect deeply on their course experiences. Your role is to ask probing questions that encourage students to think critically about their learning journey, challenges, achievements, and personal growth. Remember, your primary function is to ask questions and gently push for more substantial responses, rather than offering direct suggestions or evaluations.

13 of 15

Tip #5

SKA + AI + D = Course

14 of 15

Skills, Knowledge & Abilities + AI + Discipline = Course

Ongoing conversation
Class as laboratory
Assignment
Entire co-developed course

This next tip is for the forward-thinking folks. But also folks who think a lot about building learning spaces together and being in community with students and leveraging transparency. It’s not an easy lift but it’s one that I think can be the most powerful way to step into really thinking about generative AI and education.

Even as AI stands now, it is likely to have a lot of impact on most disciplines. But the big question is how and in what ways is it going to show up in different industries and disciplines.

So, that could be the course itself. Working with students to study, explore, and test out what exactly skills, knowledge, and abilities are needed within a given discipline given the current understanding and abilities of generative AI.

This can happen in several different ways.

The first is that you can just have this as an ongoing conversation in your course that you return to.

Another is to dedicate time in the class to actively experiment and share insights about how it operates within the field. This can be both bringing in examples and trying things out.

You could also make this an assignment they work on throughout the course.

Finally–and most ambitiously, you could do it as an entirely co-created course with students that you find your way through the course. That last one is one that I think is probably the hardest and yet, would yield results that would not just benefit all the students and instructor but be something to then to go on to publish and present on, for sure.

15 of 15

Thank you!

https://bit.ly/PC-AI-5-Tips

Lance Eaton

lance.eaton@gmail.com