AI in Academia: A Practical Conversation

Introduction

> SYSTEM: Generate an engaging introduction about AI in academia
> CLAUDE: I've prepared four versions of the introduction. Would you like formal, 
> casual, humorous, or technical?
> HUMAN: Just pick one and make it good. This is tomorrow and I'm tired.
> CLAUDE: Got it. Here's a balanced approach that should work well...

Well... I guess that answers the question of whether I used AI to make this presentation! Let's call that our first lesson in "Things Not To Do."

Insert "How it started/How it's going" meme here

Hey everyone - thanks for being here today. Quick confession: this is actually version 12 of this presentation. The previous 11 kept getting more complex and theoretical until I realized: we don't need another academic lecture on AI. We need a conversation about how we actually use these tools in our everyday work.

So I'm going to share what I've learned so far, but I'd also love to hear from you during our discussion time. What are you curious about? What's working? What's frustrating?

As someone still figuring this out myself:

1. I'm experimenting with these tools almost daily
2. I've had 6 successes, 12 failures, and about 83 "interesting learning experiences" (I made that number up, but it feels right)
3. I'm here to help you avoid my mistakes and learn from what actually works

Today we'll talk about three things:

• How AI is changing academic integrity (and what we can do about it)
• How to actually talk to these AI tools effectively
• How to use AI for administrative tasks without compromising confidentiality

Section I: Academic Integrity and AI

The Reality Check: AI Detection Limitations

[Lower voice conspiratorially]

Let's address the elephant in the room: AI detection tools are imperfect. Here's why:

The Truth About "Catching" AI-Generated Content:

• Running AI text through Grammarly often disguises AI patterns
• Making small manual edits breaks detection algorithms
• Asking AI to "write like a B student" creates more human-like errors
• Detection technologies lag behind generation capabilities
• Even GPTZero and Turnitin have significant false positives/negatives

Sonic crayon drawing AI image

Insert "AI detection vs reality" meme here

[Normal voice]

But here's the thing – trying to police AI use is like digital whack-a-mole. What if we pivoted entirely?

Embracing AI Collaboration: A More Practical Approach

The more productive question isn't "How do we catch AI users?" but rather:

1. What skills must remain human? (critical thinking, personal reflection, original insight)
2. What tasks can we comfortably delegate? (drafting, outlining, research summaries)
3. How do we assess learning in an AI-enhanced world?

Example: The Transparent Assignment Design

Instead of:

Write a 5-page analysis of Hamlet's character development.

Consider:

Use an AI tool to generate an initial character analysis of Hamlet.
Then, critically evaluate this analysis, identifying three specific
areas where you disagree with or would expand upon the AI's
interpretation, citing specific evidence from the text.
Document your AI interaction by submitting the prompt(s) you used.

Interactive Activity: Assignment Transformation

The Human-AI Division of Labor

Tasks Better Delegated to AI:

• Initial drafting and outlines
• Summarizing straightforward content
• Generating practice questions
• Providing background information
• Formatting and citation checking

Tasks That Should Remain Human:

• Making conceptual connections
• Applying theories to personal experience
• Ethical judgment and reasoning
• Creative innovation and original insights
• Critical evaluation of AI outputs

Section II: Talking to AI (Without Sounding Like a Robot)

The Three-Part Conversation Starter

When you're talking to AI, think of it like meeting someone new at a conference. You want to:

1. Introduce yourself: "I'm teaching a first-year biology lab..."
2. Set expectations: "I need help creating clear instructions for..."
3. Explain your goal: "...so my students can complete the procedure safely."

It's not rocket science - it's just good communication!

Let's See This in Action

Boring Prompt:

How do I make a better rubric?

Conversation Starter:

Hi there! I'm teaching public speaking to first-year students who get super 
nervous about presenting. Could you help me create a rubric that's encouraging 
but still has clear standards? I want them to know exactly what an A speech 
looks like compared to a C speech.

Let me show you what each version gets you...

[Show comparison of outputs]

Did you notice how much better the second response is? The AI has context now. It knows:

• Who your students are (first-years)
• What challenges they face (nervousness)
• What you're trying to accomplish (clarity + encouragement)

Your Turn: Let's Practice Together

Section III: The Privacy Thing (Because We All Have Secrets)

What Happens in Claude, Doesn't Always Stay in Claude

Let's talk about what might keep you up at night: privacy concerns.

Quick question for you all: By show of hands, who has used ChatGPT or Claude for something work-related in the past month?

[Wait for hands]

Now, who has read the terms of service to see what happens to the information you share?

[Usually fewer hands - laugh]

Here's the reality check: Many public AI tools can store and learn from your inputs. That recommendation letter with a student's personal details? That sensitive departmental planning document? They might not stay private.

The Recommendation Letter Dilemma

Let me show you two ways to approach this:

The Oversharing Way:

Write a recommendation letter for Jane Smith, my student with a 3.8 GPA who 
has been accepted to Stanford Medical School but struggled with organic chemistry
and has anxiety issues that affected her performance in my class.

The Smart Way:

Help me draft a recommendation letter template for a pre-med student. 
I'll explain that they overcame specific academic challenges while 
maintaining strong overall performance. Give me placeholders where I should 
add personal details myself.

See the difference? The second approach gives you a useful template without exposing anyone's private information.

Quick Poll: What's Most Sensitive in Your Role?

What's Next? A Conversation About Tomorrow

[Note to self: This is getting meta - version 12 of the presentation is discussing what happens next!]

So, what did we learn today?

1. Fighting AI use is probably a losing battle - but we can adapt our teaching
2. Talking to AI is just having a good conversation - be clear about what you need
3. Privacy matters - be thoughtful about what you share

But I'd like to end with some questions for us all to consider:

Questions Worth Asking

• What aspects of your work should always remain human?
• What tasks are you happy to delegate to AI?
• How do we help our students navigate this new world?
• What skills will matter most in five years?

I don't have perfect answers to these questions. Nobody does yet. But I think they're conversations worth having together.

One Final Thought

The writer Douglas Adams once said: "We are stuck with technology when what we really want is just stuff that works."

Maybe that's our job now - figuring out how to make AI work for education, not the other way around.

What do you think?

[Open for discussion]

Want to Keep Learning?

If you're interested in exploring more:

• Our teaching center offers monthly AI workshops
• The library has created a great AI research guide
• I'm happy to share these slides and resources
• And yes, you can always ask Claude yourself!

My email is on the screen - I'd love to continue this conversation. Thanks for being here today!

For Reference: AI Tools Cheat Sheet

The best tool? The one you actually use.

Version Notes

v1-v11: Various iterations with increasing complexity and pretentiousness quotient. 📈

v12: Complete structural reorganization. Dropped meal metaphors. Added fourth-wall-breaking meta-joke. 🍽️➡️💬

v13: Added placeholders for memes and drawings. 🖼️

v14: Reduced pretentiousness, increased conversational tone. 🗣️

v15: Added audience interaction elements. 👋

v16: Streamlined tool comparison section. 🔧

v17: Enhanced meta-commentary about version development. 🤔

v18: Added version notes thanks Claude. 🤖

v19: Added emojis and dark mode optimization. 🌙

"If the slides get weird, just roll with it. That's part of the AI experience."