Recent talk: The Good, The Bad, & The AI: Exploring Generative AI in Education

Estimated Reading Time: 25 minutes

It's been feeling like "another week, another talk" the last couple of weeks, even while navigating COVID.  This talk was for faculty in Boston University's School of Social Work.  This talk felt a bit more smoother and natural than some of the others that I've done and I think it's both a mixture of doing better with my slides as my visual guide and how I arched the discussion.  I also made parts of this more interactive which also helped.  Additionally, the group was an energetic crowd for a 90-minute session.  

I leaned into doing the annotated slide deck and folks seem to find that to be a key piece in helping them both in the moment and as a follow-up if usage stats tell me anything.  

The Good, The Bad, & The AI: Exploring Generative AI in Education 

Hi all,

I’m Lance Eaton and I’m excited to talk with you today about the role of AI in education and pull together some of the ideas that I’ve been seeing in the discussion over the past year.

Today’s talk is an extension of work I’ve been doing for 15 years in education and technology. I’ve had the privilege and desire to explore the intersection of education and technology in many ways, learning about both its possibility and of course, its problems and challenges. During the past decade, I’ve worked on different projects such as exploring hybrid flexible course design in the mid-2010s, well before the pandemic.  I’ve examined how digital technologies can change how we think about and perform service learning.  I’ve considered the power and importance of student agency and how open educational practices such as open pedagogy can improve student learning and meaning-making. I’ve also explored how surveillance and data practices can be another form of power-hoarding over student learning and agency.  

Often, I’m trying to figure out where is the balance of technology as a means of improving opportunity and agency for students while recognizing that the tools themselves are not neutral and come with trade-offs that can be problematic.

All of which is to say that when generative AI was propelled onto the scene last November, I had lots of different feels and thoughts about it and those continue to evolve through today.  

Let’s get started!

Our agenda for the next 85 or so minutes will focus around these questions.

General Questions About Generative AI

What is and isn’t generative AI?  This will be a little bit of a Q&A interactive session for us to get on the same page about the tools we’re talking about and our understanding of them.

What are some examples of using generative AI? I’ll show you a few general prompts that I’ve used and the output by ChatGPT to give folks some variation of what these tools can do.  

Why generative AI challenges us?  I’ll get a bit deep here about how the tool represents something different for us.  

Why generative AI might be useful for teaching and learning overall?  I’ll provide some guidance and strategies on what to do.  

How might you approach thinking about generative AI?  This final part will have you discussing with others where you find yourself and what you do next.  

At the bottom there and throughout the slides is that resource link.  I’ll put it in the chat too.  Keep that link–bookmark it.  It’s what I call my annotated slide deck which includes the text of this talk as well as lots of resources including prompts, links, and responses from that I’ve gotten from generative AI.  The resource is covered with a Creative Commons By ShareAlike which means you can share it with others and use materials from it. 

Here are some of the currently most popular AI tools.  I’m curious about folks familiarity with them.  I’m guessing everyone has heard of ChatGPT, let’s go through the rest.  For the next 30 seconds, put into the chat which of these you HAVEN’T heard.  

That is, if you haven’t heard of Claude, then put that into the chat.  Write in all the ones you don’t know.  You can enter each individually or submit them all at once.   

Now, that’s quite the variety of experiences here in the room. If you are familiar with one of the ones on this list, feel free to provide others with a 1-2 sentence description of what it is.  And, of course, no cheating by going to ChatGPT or another tool and having it generate an answer!  There’s also a chart in the annotated slide deck for more information about these.

These tools are becoming increasingly ubiquitous. And the AI Hype of the last 10 months feels different from prior cycles.  

First, generative AI is different from the AI we’ve been using for the last decade.  When we interacted with Alexa or Siri, or advanced recommendation systems, these didn’t feel as impressive or easy to use right out of the box.  But generative AI–AI that can generate text, audio, image, or video through the use of everyday language–that feels different because of its ease of use.  

I mean we’re several years since the launch of the Metaverse and can anyone tell me what that really is and if they have visited it?  

The lift to figure out what the Metaverse is, how to access it, how to create in or with it, and why it would be better than other things–that’s a lot of figure out.  

But the generative AI comes in the form of a chatbox…something that’s been around for decades.  We all know: see a textbox on the computer, enter text.

Generative AI took something terribly complex and nuanced and made it usable in a textbox.  That’s a devilishly easy invitation.  

What exactly is generative AI?  Feel free to put your answers in the chat.  I’m almost tempted to play a game of “wrong answers only” but I won't.  

So here is the answer I’m providing and how I go it.  First, I asked several generative AI looks including ChatGPT, Claude, Bing, & Bard.  They all gave decent answers of varying and similar degree.

But then I poured all of them into Claude and told it to:  “Take these 4 definitions of generative AI and come up with the cleanest, clearest, and simplest explanation of generative AI that is under 50 words, can be understood by a 8th grader and maintains the full integrity of what generative AI is:”

It gave me that answer on the right:  “Generative AI is a type of AI that can create new content like text, images, and videos after learning from large amounts of data.”

I wasn’t that excited with that answer. I felt like it had limitations for really understanding it.  I asked it to elaborate.  It came up with:

“Generative AI refers to artificial intelligence systems that can generate new, original content such as text, images, audio, and video. These AIs are trained on large datasets which allows them to learn patterns and relationships in data. They can then use what they've learned to produce new examples that are similar to but not exactly the same as the training data.” 

That’s definitely a little more clear idea of what it is. There are definitely things that I or others would nitpick and challenge with this definition, but that’s the case with many loaded terms.  

Now, let’s do a little myth busting shall we?  Throw into the chat–true or false:

Generative AI may become becoming sentient.  That is, alive.  

False.  And also, it is not a person or sentient–no matter what all those sci-fi novels, movies, comics, TV shows, and Silicon Valley Dude-Bros tells us.  We have a long history of telling ourselves that our objects are real–going back millenia.  What we hear and see in the media is often sound bites and misrepresentations of what generative AI is.  

Ultimately, generative AI is a complex technology with lots of possibilities, but sentience is not one of them in the immediate future.  Much of the hoopla out sentience is projection and that’s been a habit with our technology for a long time.  

Ok, what do you think about this one?

“Generative AI thinks to arrive at its answers, like humans do.” 

False.  Generative AI doesn’t think like you or I do.  First of all, we all think differently–that’s why we talk about neurodiversity.  But our mind is composed of whole things be it words, images or sounds, tastes, etc.  By contrast, generative AI works through mathing the hell out of things.

For instance, when working with text, a generative AI will first have access to what is called a “large language model”--that is a massive data set of text.  What the generative AI does is to go into that data set and mathematically analyzes the relationships across that text.  Now, it doesn’t analyze the words but rather strings of characters.  Those strings of characters are referred to as “tokens” and are short, something around 4-5 characters long.  

The AI calculates the relationships in probabilities of those tokens across its data set and also with any prompt you give it.  So when ChatGPT gets a prompt by a user, it’s mathematically examining the relationship among tokens and then searching its large language model for similar relationships to help generate the most probable answer.  

It’s not thinking, it’s mathing.

Ok, what about this one?  “Generative AI can lie.”

This one is tricky but it is true that Generative AI doesn’t lie.

Lie comes with intention and intention is human.  

Can it give false information.  You betcha!  But when you hear language about it lying or hallucinating–it’s important to understand the speaker of those words are assigning human characteristics to AI that confuse our use of it.  It will present false information because it’s not thinking and rationalizing or proving things in the ways that we humans think about it.  It’s mathing relationships and that may lead it to come up with poorly calculated wordings.

So while it doesn’t lie, you still shouldn’t trust it.  

Finally, how about this:  Generative AI will take all our jobs.

For the foreseeable future that seems highly unlikely.  It will disrupt and augment a lot of different types of work and part of the challenge for all of us will be to stay abreast of it to understand its impact on our industries.  Any knowledge worker is going to have to consider and contend with how these tools can amplify their work and not replace it.  

At its core, it is likely to help us do a lot of things we didn’t do before and that can create new opportunities. But yes, there are jobs that will be lost.  However, there will be new jobs that emerge.  The history of the world is filled with technologies that would supposedly ruin all employment and yes, there’s still plenty of work to be had.  

What are some examples of using generative AI?

Now, let’s take a look at some of the things that generative AI can be helpful with.  The next are just a couple examples of how I’ve used ChatGPT.  I show them as ways that might be helpful and insightful for how you might start to think about using generative AI if you haven’t tried it yet or if you’re looking for some basic ways to get started.  The annotated slide deck will have a lot of additional prompts and guidance on how to use it.  

So here’s one that I use now at the beginning of each semester and I’m sure many of you will to.  We all do this and here is an example of how I can save time with generative AI.  I asked ChatGPT to give me all the Tuesdays between the start and end of the semester.  This saves me toggling back and forth between my syllabus and a calendar to get these days.  I also asked it to include holidays and such so I can keep that in mind as well.

You can see, here are the results that I got in less than 30 seconds.  Now I can copy & paste that list into my syllabus or anywhere else I need to.  Generative AI can help with simple but tedious tasks like this.  

This one is a little more advanced.  I’m asking it to produce a list of ways to use generative AI for the general profession or education professional.  I’m also adding some guidance about what to include–not just telling me what it can do but also telling me how much time it will save me, how hard it is, and how to do the thing.  This is one of those things that you can do and customize it to your own personal situation, editing out certain details and adding new ones that fit you.

This is a limited view of the results, the full are in the resources.  But as you can see, it provides a couple examples that folks can start to use right from the start.  It is focusing on both ChatGPT and Google services because I asked it too and I would encourage that for such prompts, you focus in on the tool to get best results.

Now this one, I focused in very simply.  I asked it to basically create a guide for how to create a literature review in a intro to social work graduate course.  I asked it to explain each step and also to include the rough-amount of time each step will take and when in the course they should complete it. And to me, this is the kind of stuff that I find incredibly useful with generative AI is to help me get out a rough draft or to see what it generates so I have something to work with or against.  It’s likely to give me ideas and insights about the process.  Such as including how much time it may take does make me think differently about a course. 

Here’s the response. What I find interesting is that in the first 9 weeks, it’s expecting students to spend a minimum of 55 hours on this project.  That may be true–that may be the assumed amount of time but what does it mean to write it out. How does that impact what we think about everything else in the course?  To me, this is the thing. It gives me ideas and it also challenges me to think about its responses and our own assumptions about what happens in our courses.  It’s in those assumptions that we embed the hidden curriculum and the assumptions we put upon students to figure out. Therefore, I think it can make me challenge my thinking in many ways when I use it.

Why generative AI challenges us?

Those were a few examples to consider to get a sense of why folks are using it and what value they find.  I want to switch now to some of the bigger questions that we as educators are navigating with this new technology.  These are the hard bits and I want to acknowledge them and possibly shed some light for many of us for our mixed up and tangled feelings about this technology.

The tool comes with a lot of baggage.  I mean a lot.  And that baggage is different than other technologies in that we’re learning about the problems of the tool just as we’re learning about it. We know that there is all sorts of bias based upon the large language models that were used and also the bias that the companies themselves knowingly imbue into these tools. That alone raises concerns.  

Then, in the last year, we’ve learned about exploited workers, environmental degradation and resource depletion, and significant energy use that’s likely to contribute to climate change.  We’ve also learned that it may have misused or be abusing copyright of creators, authors, and artists to fuel its creations.  And the level of privacy and what is used with our input into the generative AI doesn’t leave us particularly safe.  

Right out of the gate, it raises our hackles because it’s another product from Silicon Valley tech bro offering utopia in the form of something that promises to help so long as we don’t care about human rights, environmental preservation, or intellectual property.  

Generative AI challenges our notions of work.  By doing something that we have largely assumed was only capable by humans–that is create text, visuals, audio, and video with little expertise–it creates an opportunity for folks to wonder why they have to do certain types of work.  Like the calculator, many folks are asking, why can’t I just use this. And we can come up with answers about the importance and ethic of writing as thinking or creating visuals as part of the process of learning or becoming.  What does it mean then to “do the work” or “show your work”?  Some folks will reasonably challenge the real material value of “doing the work.”.  We may buck at such challenges but folks will have different reasons for doing so that we should be mindful of.   

 I know, in academia, we cherish the process–we love to do the work because we know there is deep learning in there.  But you’re teaching social workers–folks who are going out into the world to be overworked and underpaid.  Folks who are going to often have more cases than is appropriate or allows them to be the person they want to be in this role. In this world, the process doesn’t make sense–there’s no real time for it. The process is the path to burnout.  So yeah, they may choose to lean on shortcuts if only to maintain what is often an unmaintainable disposition.  Convincing them not to do so feels like a really hard thing to do–on top of all the other things we have to do as educators.  

Then, of course, there’s the fact that it can be really hard to distinguish.  This creates two deep challenges.  The first means that there are no good automatic ways of catching it.  There continues to be a lot of companies trying to get rich off our insecurity but AI generative detectors are not going to be useful in any way that makes sense.  The only folks that are going to get caught with those are folks who are actually in need of help or false-positives of students. Those false-positives will also be directed more towards students who are multi-language learners.  

And that is the most harm we can do–accuse an innocent student.  Not just because it alienates them from the institution but also because there is no way to prove their innocence.  These machines work from probability, not facts. Therefore, a student is going to have to defend itself against a machine on probability. How?  Exactly–it’s a stacked deck.  

Yet the deeper challenge is that because AI is hard to distinguish, we are left wondering about our effectiveness to evaluate work.  Many folks will claim a Spidey-sense or just “knowing” when generative AI has been used.  I know I certainly at times think I know when a student’s writing is off.  But we don’t really know and we won’t be able to really prove it. This leaves us to a vulnerable space where we know but don’t want to say that there are possibilities of students fooling us and passing our class without actually learning anything.  And that idea challenges many of us as educators.  It can make us feel inept or wondering what we are doing in this work.    

And in this way, generative AI challenges power and the power of the learning space.  The power of us as educators to know and hold knowledge in a particular way.  What does it mean that students can choose to use this tool to challenge us or bypass us and our role as knowledge gatekeepers.

Now–I’m not saying that individually, we feel like we hold that power or we operate through that lens, but as representatives of a larger institution within a larger system–higher education–that is, in fact who we are:  Knowledge gatekeepers–deciding who goes forward with passing grades and who does not.  

Generative AI leaves us wondering about our ability to hold this role which means it represents some level of power change that we’re not entirely comfortable with.

It feels very much like the vast majority of mental work that gets turned into tangible deliverables for evaluation in higher education are very quickly becoming possible to being generated by AI. 

It’s not just the fact that it can challenge us in these other ways, but that it come at the end of a long train of technology, pedagogies, and world events that have asked–no, demanded–that faculty change much if not their entire practice.

Many of us have taught for years and developed a deep and rich practice and philosophy of teaching where our courses are interconnected webs.  Everything comes together in an alignment that we have been working on years to perfect.  That alignment deeply interconnects with learning outputs by students that are directly thrown into question as a result of generative AI.  

To pull on that thread, means to unravel all the other interconnected threads.  I don’t know that everyone fully appreciates that depth of that fatigue, frustration, pain, and yeah, even sadness.  It feels like we’re back at page 1 but that page 1 might have been written by an AI chatbot.

But teaching and how we show up to a class is so personal, so individual, and so deeply a part of our soul--that this isn’t just a pivot…it’s a paradigm shift.  

And the lift to reinvent our approaches is hard, scary, and exhausting.  I wanted to name and acknowledge that.  I wanted to thank you all for showing up here, showing up for your students, and looking to find a way forward.  

What are we to do as educators?

Ok, so let’s take a breath.  Also, this is the part of the presentation where I include pics of my pets as part of my slides….because I’m a professional–and figure a little brevity can be helpful here.

What are we do to as educators?  I’m not looking at the chat but I bet someone is going to respond to that rhetorical question with a reference to retiring.  

What follows are some of my best tips to get situated and more comfortable with this new age we’re in.  

 No–there isn’t a perfect plan.  

And let’s rejoice with that!  

No, seriously. Nothing will solve this and so with that comes the relief that you–the individual sitting here listening to me are not solely responsible for solving it.  Because here is no solution.  That doesn’t mean we give up responsibility but that also means we’re not solely responsible for solving it. None of us are going to do this perfectly and that’s ok.  

Seriously, I give you permission to fumble through this so when you do–if someone actually asks; which they won’t–you can say that Lance said I could stumble through this.

This is something that will be a little bit harder. But you’re going to have to get comfortable with being a uncomfortable.  Often, we’re not going to know and we’ll have to sit with that.  I think this also means we’re going to think about how we engage with students and possible usage that is allowable or even allowing some level of generative AI in work.  Not because I think that is necessarily ideal, but I think that’s the only way to normalize practices of identifying usage.  That is, we can only encourage students to indicate they have used these tools if we make it ok for them to do so.  Otherwise, we’re playing the suspicion game and in that path lies madness.

Look around at this room.  There’s a lot of wisdom in this room.  In fact, there’s lots of conversations going on right now in social work and folks are sharing lots of tips and practices.  Find your listserv, your Facebook or LinkedIn groups, your professional organization newsletters–dare I say to go onto Twitter–dumpster fire that it is?!?!  

Seriously, find your professional communities–they’re out there and sharing.  Learn and share with them! So much of what I have learned in this, it is because of my network.  In fact, I literally co-authored an article about how open education–sharing my work with others and learning from others’ shared work–has helped me figure out my own strategies with generative AI. It’s the very thing we tell our students to do and we need to also do it. 

Yes–I know.  You have to try it out.  At least for 30 minutes, but I’d encourage you to make a plan to use it for 10-15 minutes a day for a few weeks.  Play around with the prompts in annotated slide deck, make up your own, learn from others.  Test it out, see how you can use it, see where it doesn’t work.  Find others who are using it substantially and learn from them.  Create an informal group where you can share your chat threads and results.  You will want to use it if only to learn and understand it better for yourself but you will also want to use it for a lens for your class to think about where it might make sense to allow for its use or where it feels like it would interfere with your intentions.

As you learn, it can be really helpful to clarify and come to understand what your relationship with generative AI is.  Write it out if that helps.  Explain to yourself where you are and aren’t comfortable with it and why.  Verbalize it to help you understand where your struggles lie.  

Given that, consider where it will and won’t make sense to use both in your personal and professional life.  That can help you figure out how to orient to your teaching and your students.

I know some will want to come from a top-down approach to this. They will want to decide how the tool will be used.  I encourage instead for you to talk with your students and learn what they know and think about these tools.  What are their concerns and challenges with it?  Where do they find it helpful?  

I learned from one of my students that she used it to organize her notes.  She knew she wasn’t going to organize them and so put them into ChatGPT and asked it to organize them.  That’s a really smart approach.  Imagine a student who has such challenges.  Sure, they can struggle through organizing their notes or most likely do what I did–struggle through the messy notes the night before while trying to submit the paper.  What if by organizing the notes into a coherent form, particularly in relation to a topic they want to write about, it helps them more quickly focus on the paper itself? 

We don’t know all the ways that students will creatively deploy these tools and how that may help them.  Given that, talk with them and collectively figure it out.  

Finding our way through this is going to be challenging–in part for reasons I’ve already highlighted.  But also, because it continues to change.  One way that will significantly help you and your students is to build trust and community with your students.  

Realize that one of the best ways you’re going to get students to use it less is to make sure you are building connection and trust in the classroom. When students feel connected and that they can trust you, they will ask.  It’s happened in my classes where they asked if they could use AI for an assignment.  I explained why I don’t think it’s ideal on a given assignment but still gave them the final say.  They opted to do the assignment without AI. It won’t happen every time, but it will more often happen if we’re centering the relationship rather than the output.

Ok–but what about what to do with it?  Well, here are 3 solid ideas but of course, there are more in the annotated slide deck.  

Infinite example creator:  If you are trying to come up with case studies or examples to break down certain concepts or ideas.  It can quickly create examples, case studies or other things where you want students to dig it.  It can also create bad examples.  We can often ask students to use their work in the future as exemplar work–but we don’t often have bad examples to show as well–in part, because it feels wrong to ask a student to use it in that context.  But generative AI can create these now.  

Course Promptbook:  Collectively, you can build a prompt-book that you and your students can use to further their knowledge and ideas. Maybe it’s a set of prompts to further explore topics, generate ideas, or bring to class to discuss.  

Red-teaming:  Have students use it to create content and then have other students challenge and find faults or limitations with it.  That is, draw out the issues of the tool by using it and letting students go to town with finding its flaws and limitations around course content.  

So these are a few to think about but there’s many more to consider.

How might you approach thinking about generative AI?

Up next is an opportunity for you all to give some thought and consideration about where you are with all of this.  I’ve talked long enough. It’s time for you to think about these things.  

We’re going to do an activity right now.  First, take a moment and think about which group you find yourself in.  

If you can, write it down. Here are our 4 groups.  Trying to determine which one you find yourself closest to.   

1. Lean toward engaging and using generative AI in relation to teaching and learning
2. Learn more and use generative AI more and maybe dabble with it in teaching and learning
3. Learn more and use generative AI more but not ready to bring it into teaching and learning
4. Lean away from generative AI and not wanting to use it in teaching and learning

Which group do you find yourself in?

Take a moment and write down your thoughts with these questions.  

1. Are you settled in the group that you find yourself in?  Why or why not?
2. Would you want to be in another group? Which group & why?
3. What would it take for you to move to that group?

I’ll give you a minute.  

So I’m going to put you into breakout rooms of 3 for 10 minutes.  Be prepared to report back.  

• For 1 minute each, share which group you’re in and why.  
• For 2 minutes each, share what it would take for you to move to another group and why?

 Any questions about what we’re doing?

Don’t forget you can always go to slides 38-40 to see the questions and I’ve also put them in the chat.  

Who wants to share?  We’ll hear from reps from 3 different groups and the rest can put their experiences in the chat.  

Thank you for participating in that activity.  I think it unsurfaced some key considerations.  Part of my goal in that activity is to help you all imagine what you might need and where you might need to go–to help you speak to that and speak to others about it since often we aren’t always aware of what we need or want to do.  You now have a clearer sense of that.  I hope through this session you’ll have some ideas about what to do next.  

We’re in a weird space in higher ed. There’s been lots of changes over the last 15 years and the cycle of change seems to be hyper-accelerated with something like generative AI.  It can feel scary and overwhelming and challenging in ways we didn’t anticipate.  

BUT we can find our way through it. It won’t be easy but it will be meaningful.  It will require us to keep calm, build community, and work with students to figure out what this new paradigm looks like.  And students are hungry for that–we know that they seek connection in an alienated world that demands much of them and delivers very little.  So it is ok to feel all the feels and know that you won’t figure everything out, but figuring it out with others is what’s gotten us through so much in the past and will guide us into this strange weird place.   

---

Did you enjoy this read? Let me know your thoughts down below or feel free to browse around and check out some of my other posts!. You might also want to keep up to date with my blog by signing up for them via email.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.