Will Generative AI Enhance or Stunt College Learning?
1h 19m 10s
The debate centers on whether generative AI enhances or stunts college learning. Critics argue that widespread AI use leads to cognitive surrender, where students passively rely on prompts to complete assignments, avoiding critical thinking and deep engagement. They highlight that AI undermines the essential friction of learning—such as struggle, error, and effort—that fosters understanding and growth. This is especially problematic for students from disadvantaged backgrounds who face greater academic pressure and fewer resources. Supporters counter that AI can democratize access to information, support diverse learning styles, and improve educational equity when used responsibly. They emphasize the importance of AI literacy, teacher-led pedagogical design, and student agency in learning. The discussion reveals that AI doesn’t inherently harm or help learning, but outcomes depend on how it’s implemented—such as through structured classroom practices, oral exams, or human-centered teaching. Ultimately, both sides agree that education must evolve to maintain intellectual rigor, foster critical thinking, and preserve the human elements of learning, including mentorship and meaningful interaction. The conversation underscores the need for institutional reform, not technological solutions, to ensure that AI serves as a tool for empowerment rather than a shortcut to academic success.
Whoever says saving money is hard doesn't know that new customers who bundle and save with
progressive save hundreds because switching your home and auto insurance is better than cutting
down on the things you love like those ice lattes that help you get ready to face the day.
So switch your home and auto to progressive and see if you could save and keep sipping away.
Get your quote today at progressive.com progressive casualty insurance company affiliates and other
insurers not available in all states. Artificial intelligence in the halls of academia has been
called many things a personalized tutor a tool that students should master a creative partner
and a shortcut that enables cheating and degrade learning. Welcome to the debate over AI and higher
education. The number show it's already here about half of college students use AI regularly for
school work bringing with it fears of cheating and cognitive offloading. So is it time to power
down the laptops and return to in person blue book exams not so fast. Some more in that students
who graduate without AI skills will be unprepared for jobs where they will be expected to work with AI.
That's not all. We've been told AI will transform fields from medicine to law and education
is no exception. Supporters see an opportunity to make learning more accessible than ever.
So now open to debate in partnership with the Stavros Nyarchos Foundation Agora Institute at
Johns Hopkins University as part of the Hopkins Forum series asks this question. Will
generative AI enhance or stunt college learning? This is open to debate. I'm John Don Van. Today we're
debating will generative AI enhance or stunt college learning. We're here at the Johns Hopkins
University Homewood campus in Baltimore for the seventh installment of the Hopkins Forum,
which is a partnership between open to debate and the Stavros Nyarchos Foundation Agora Institute
at J.H.U. The Institute draws inspiration from the ancient Athenian Agora. It's an academic and
public forum with the goal to foster informed inclusive dialogue as the cornerstone of robust
democracy. Let's please welcome our debaters to the stage Gary Marcus.
Kathy O'Neill. Ravi Pennsay. Eddie Malone.
Thank you all for joining us. I'd like to just have a little chat with each of you for people
to get to know you a little bit better and Gary I'll start with you. You're a professor emeritus at
NYU. You are a scientist. You're a best-selling author. A well-known AI commentator,
particularly on your popular substat called Marcus on AI. Good name. Makes sense.
But you're not on the sidelines. You also founded an AI company that was later acquired by Uber.
This summer you were named a global ambassador for Pope Leo's and cyclical on AI. What does it
mean to be a global ambassador? What does that entail? It's still not entirely clear, but I'm an
ambassador not to the Pope, but to the document that the Pope wrote about AI, which I think is one
of the best things that has been written about AI. I think it is humane as its title suggests.
I'm a big fan of it. Did you have input on it at all? I did not have any input, although I spoke
to someone who spoke to him. So maybe very indirectly, we can hope. All right, thanks very much,
Gary. Your debate partner today is Kathy O'Neill. Kathy, it's great to have you here. You are a
mathematician. You are a data scientist. You are known for your books. Weapons of math destruction
and the shame machine. You are no stranger to academia in the past. You are a math professor.
You taught math at Barnard. You also work as an algorithmic auditor. Very nice illiteration,
but I have no idea what that means. Can you fill me in? Oh, yeah. In fact, I was going to
mention it during one of my opening statements. I worry about harm of algorithmic systems and I
often work on lawsuits related to people being harmed by algorithms. So looking down the road all the
time. I worry about things. I keep me up at night. Actually, yeah. I'm sure we're happening now.
We're going to hear some of that in the course of the other debate. Yep. Let's go to the other side.
We want to welcome Ravi Pensei. Ravi, welcome. You are a professor of electrical engineering and
computer science at the University of Michigan. You are also in charge of the University-wide
strategy on AI and other technologies. So you're right in the midst of the thing that we're debating.
It seems like a big job. You mentioned also the first university in the world.
I'm reading to provide a custom suite of generative AI tools to staff and students. Plus,
I heard the other day that you actually channeled me and moderated a fireside conversation. How
did that go to being a moderator? Piece of cake, isn't it? I mean, watching you perform not very well.
But we had the pleasure of having Nikesh Arora, Chairman and CEO of Palo Alto Networks.
And we had a terrific fireside chat as part of Cybersecurity Month.
Excellent. Thank you very much. And I want to welcome your partner, Edward Maloney.
Eddie, I want to share that you are actually a late minute addition to this debate.
That seat was supposed to have been occupied by Roman Chaudhary, but she fell ill.
And we're all hoping that she feels better. But we're absolutely thrilled and delighted to have
you doing this. And so appreciate that you were able to step in and nearly the last minute,
because you have a lot of knowledge and experience on this topic. You're at Georgetown.
You're also the executive director of a center that works on innovation and teaching methods.
You're also founding director of the Learning Design and Technology Program.
The thing I like about you, we only met today. But as an English major who still does not regret it
myself, you are an English professor. And you also teach classes on post-modern literature
and narrative structure. But you're also in this field of working with AI. How did those two
worlds talk to each other? Yeah, that's really good question. How much time do we have?
I think that was a hard one to answer. You know, apart, I think everything we think about in
the humanities is about what makes us human, but also how do we communicate as humans? How do we engage
with each other? How do we think about our impact on the world and how do we think about the world,
the impact of the world on us? And there's nothing that I think brings that out more than technology.
Technology is a fundamental tool that we use to engage as humans, to engage with each other,
to think through our world, to, you know, in this moment in time, the technology is incredibly
fascinating for what it means to that endeavor. Yeah, I love that. Thanks so much. And again,
thanks for stepping in as you did for us. Our live audience here at Johns Hopkins is playing a
critical role tonight. They've been asked for their opinion on this topic before the votes,
before the debating actually began. We asked them to do that by vote. We're going to get those
results in real time. And we're going to ask them to vote again after they've heard all of the
arguments. Again, that will come in real time. And we're going to see which side prove most persuasive
by how many people change their mind. So that makes this a competition as a friendly one. And that
means we do have some rules, no personal attacks, sound topic, actually listen to what the other
side is saying. The debaters are here to show that we can disagree with each other and do so in
good faith. So let's get started. We're going to go to our opening round and our opening round,
each of the debaters takes the floor for a period of four minutes each to tell us why they are
taking the side and making the argument they are. Again, our question is, will generative AI
enhance or stunt college learning? Gary, you are up first. You are arguing that generative AI will
stunt college learning the floor is yours to make your case. Hello, Baltimore.
I grew up here, as many of you know, my late mother lived here for 60 years, just a few blocks away.
And I wish she could be here tonight. Three years ago, I had a debate with open debate and with
John, the parallels tonight about whether or not generative AI would on net benefit society.
The jury is still out, but between the massive wave of cyber hacks, deep fakes, and wholesale AI
generated disinformation, I have my doubts, which is not to say that no AI could ever help society,
maybe future improved versions will bring the great medical and technological progress we were
hoping for, but that's about hope and not was actually been delivered. We can say the same about
generative AI in education. It's getting widely used, but it's fraught with problems.
Some of those stem from technical limitations. Current systems are pattern matchers,
that are outstanding at mimicking how people talk, but with understanding that still remains
superficial. They speak authoritatively, but an unpredictable fraction of what they say is either
bafflingly stupid or just wrong, as they say in the military, frequently wrong never end out.
Nobody wants that in a teacher. Worse AI exacerbates certain human
foibles. Most of us, most of the time, are lazy by default. And what has happened with AI is that
the vast majority of people use chatfots as a substitute for thought and learning. When students
simply type in a prompt, they tend to learn little of lasting value. They don't develop critical
thinking skills, and they don't develop problem solving skills. We call that cognitive surrender.
AI users who give in to the machine give up on the chance to become critical thinkers.
A few days ago, I talked to a smart law student in Canada with a PhD who tutors law students,
many at the top universities, they are both undergrad and grad.
She sees chat GP as an unmitigated disaster. Students often completely outsource their
learnings chat GPT, and they see their job as just typing in a prompt, like turn this legal
decision into an L.A. law episode, rather than learning to write or think on their own.
They're baffled when they have to explain in their AI written work. It's kind of terrifying
to imagine what the next generation of lawyers will be like. Worse teachers often give in,
accepting AI written essays, which leaves the few bright students who resist
feeling like they've been cheated by the teachers.
I know there's some good, at least one study
says that students who interrogate AI systems
and take them skeptically as a fallible tutor
outperform those who use it passively.
But most kids, most of the time, are passive consumers,
and they will learn little that sticks.
If I made the rules, I would not outlaw generative AI,
but I would introduce it only one students could use it
responsibly and with care.
And only after they'd already learned their profession
and not before.
In some, a few students may benefit,
but most come away with little.
I certainly don't think this generation of undergrads
is better off for having been immersed in generative AI.
And most of them, they don't think so either.
And I thank you very much.
[APPLAUSE]
Thank you, Gary.
Next year to argue that generative AI will enhance college
learning, I want to give the floor to Robbie Pensay.
[APPLAUSE]
Well, good evening.
I bring you greetings from University of Michigan.
How is everybody this evening?
Great.
All right.
I like that enthusiasm.
Acharya Devabhava.
Teacher is divine or puts simply honor the teacher.
My mother taught me to see a teacher
in every one I interact with in all my interactions.
So I see teachers in this half-elow audience
and my fellow debaters, and I'm ready to go to class
and learn from all of us and all of you together.
Take a moment and think about your most impactful teacher.
Was it somebody who challenged you?
Was it somebody who took special interest in you?
Was it somebody who dazzled you with their inspiring lecture?
Or was it somebody who saw something in you,
your potential promise of doing great things
that you didn't even see yourself in the first place?
It doesn't matter what it was.
The bottom line is that you established a relationship.
There was a connection, a human connection.
And that's the heart of our argument tonight.
Generative AI is the most impactful technology
of this century and can be a powerful force for learning
if used responsibly, if used startfully,
if used legally and always with humanity at the center,
always with humanity at the center.
So taking a step forward, we have seen this before.
Think about that student middle of the night
who is worried about that calculus problem,
or in my case, it was thermodynamics.
Think about that student who needs help
in an alternative form to just better understand the subject.
Think about that student who has physical disability
and may need information in an alternative form.
Generative AI has the ability to provide all of this information
at scale to all of us, including us, debaters here.
No institution, no university has that ability
to do that at scale, currently for all
of the individuals who are involved.
So that's the power of generative AI, if used responsibly.
And I realize that sometimes generative AI tools
will provide produce files information.
The answer to files information is not let's just not use it.
The answer is AI literacy.
We need to interrogate, we need to question,
we need to double check resources,
and we need to see through the fluent nonsense
that LLM scan produce, but that's part of our literacy.
That's part of going to college and learning.
Answer to bias is not let's shut it off.
It is thinking about transparency and accountability.
The answer to overuse, which can negatively impact
our critical thinking skills, is to be very aware of when
to your generative AI and when not to your generative AI.
That's part of our intellectual skill
that we all need to learn.
So my argument tonight is, if used responsibly,
if used responsibly, generative AI can enhance learning.
So I request you to vote in favor of enhance.
Thank you.
Thank you, Robbie.
[MUSIC PLAYING]
Great meeting.
Who took notes?
Anyone?
With Monday notetaker, the meeting ends,
and the plan already exists.
Action items, owners, deadlines.
Go to Monday.com, the AI workspace platform.
Next up, Cathy O'Neill is arguing
that generative AI will stunt college learning.
The floor is yours.
What?
I love public debates.
I'm so glad to be here.
This is awesome.
Listen, it's uncomfortable to learn.
It is an uncomfortable friction.
Learning is friction.
I'm a mathematician.
I should know most of my time as a student was spent feeling
like an idiot, right?
The payoff is that once you do figure something out,
you get this amazing feeling.
Aha, I get it now.
That aha wouldn't have happened if we didn't have the friction.
And the problem is that AI removes friction.
As long as we have access as a society
to the frontier models that are either cheap or free,
we're going to have that possibility.
And when I say, we, I'm talking about college students
who already are doing this because they're busy people.
They have a lot of homework.
It's difficult.
They don't think they're cheating.
They think they're just getting a little bit
of a conversation going with the AI,
but they are losing that friction.
And we know that.
That's that we have studies for that.
But I want to frame out an even bigger issue,
which is not a new issue.
And usually that's true.
These AI issues, these AI failures usually
point to something that has been going on for a long time.
Which is that learning itself is a feedback loop
that requires friction.
Because, as I said, learning is hard.
It's uncomfortable.
It makes you feel like not doing it.
Why would you do that unless you had the social structure set
up to do it?
Like your parents wanting you to get good grades.
Your teachers wanting you to do your homework,
get good grades.
Your peers who are also doing homework with you.
And they are doing well.
And you want to be one of those kids that does well.
Well, one of the problems with AI,
the access that we have to AI,
is that it has circuited this feedback loop.
This feedback loop, which has worked not perfectly.
I'm not going to argue perfectly.
After all, let me just pause there.
We know that there are better resource kids
that have tutors, right?
That have better teachers that have smaller classrooms.
That have peers that are more competitive.
That's why parents move to those towns
if they care about education in order
to get their kids into those schools where those things happen.
And so yes, when we have a tool that students can use
that short circuits that lets them get good grades
without actually learning, then the feedback loop breaks down
because their good grades make the teachers think
they're doing OK.
Their good grades make the parents think they're doing OK.
They might even think they're doing OK themselves
because they're getting good grades.
So where is that feedback loop?
How are they going to learn?
How are they going to be forced to feel uncomfortable?
How are they going to force themselves to feel uncomfortable?
And I was finished by mentioning that this is related
to my work as an algorithmic auditor.
I asked the one question over and over again.
It's a really dumb job really.
I just asked the same question over and over again,
for whom does this fail?
And I will just say that when I think about college education
or high school education, and I think about Americans
or all over the world, what I keep coming back to
is this will fail for kids that aren't
going to be given the resources of personal attention
to make sure you're being uncomfortable sometimes.
You're forcing yourself to do the uncomfortable task
of not understanding something, figuring it out
for yourself, getting that aha moment.
And that's going to be most kids.
It's going to be most kids.
Most parents go to jobs.
They're busy lives.
They use school, well, they want their kid to do well in school.
They want their kid to get a good job.
But they do think of it as kind of as a babysitting service.
Go to school, get good grades.
If they're getting good grades, they don't think twice about it.
Those are the kids I'm worried about.
For whom does this fail?
I think about the kids who don't have those tutors,
who don't have intellectual debates every night at dinner
like my kids had with me, who don't see intellectual product
productive struggle on a daily basis at their house.
They're going to, where are they going to get that?
Cathy, I'm sorry you've hit time.
Oh, we're done.
Yep, but thank you very much.
Yes, thank you.
[APPLAUSE]
Next up, we go back to this side arguing
that generative AI will enhance college learning.
Please welcome to the floor, Eddie Maloney.
[APPLAUSE]
Thank you very much.
It's great to be here with you.
I drove up from Washington at rush hour,
so I've been thinking a lot about friction.
I very much appreciate you introducing the topic, Cathy.
One of my colleagues in an initiative
that I belong to, a Purple Oshandra,
argues that's higher ed, institutions of higher ed,
universities, colleges, were actually
created to reduce friction for learning.
We bring all the books together in one place in the library.
So people have access to them.
They can come to one place.
We bring all the smart people together.
We bring all the laboratories together.
We try to create a space where people can access learning
and reduce friction, even though we
know that friction, in many ways, is fun.
mental to learning higher education has been established
really structured in many ways to reduce that friction.
It also comes with a lot of constraints, right?
We have faculty who have limited time
and we have more and more students
over the years our student body has grown.
It certainly grew in the 1950s in the United States.
And so we had to reduce friction
to deal with some of those constraints.
So we created spaces, well probably not like this.
I don't know how many lectures happened in here,
but a lecture hall is very similar to this
where people are talking at you because it scales well.
We have essays and exams as evidence of learning
because they scale well.
There are things that we are doing
to recognize the constraints of higher education
that ultimately are intended to reduce that friction,
that friction for gaining access to knowledge
and learning, which is interesting because in many respects,
it's at odds with how we think about the process
of learning and what learning ultimately means.
And we know with AI that it can exponentially reduce friction,
depending on how you use it, transactionally,
it can take that friction and eliminate it completely.
We can turn out products really quite well.
But the problem there is that those structures
of higher education that were developed to reduce friction
also needed to create products
that didn't really fundamentally represent learning, right?
They represented a particular kind of output that scaled,
as I said, but not necessarily with something
that fundamentally represented the process of learning
that iterative process.
I don't know if I can ask questions, but I won't.
I was gonna say, if anyone knows how many hours
go into a college education,
it's about 1,800 hours of see time at a university
when to get a bachelor's degree, which is a lot.
But if anyone also knows how many hours
an electrician has to go through an apprenticeship
to become a master electrician,
it's about 8,000 to 9,000 hours.
And those hours are iterative.
Those hours are spent engaged with a master
who is helping that person teach and learn, right?
It doesn't scale, it's a one-to-one relationship.
And so the friction that we've reduced in higher education
is now multiplied by AI.
Because of how we're using AI, we're using AI
to solve the problems that we created in the structures
to reduce friction in the first place.
There are a lot of different ways of thinking about how
AI can, in fact, enhance friction.
We talk about something like productive friction.
What it means to introduce dialogue and problem sets
and conversations that can be meaningfully pushing a student
or a faculty member as they're thinking
about producing new knowledge in new directions in new ways.
And one of the things that AI is doing
is not just disrupting that product,
but it's disrupting the structures of higher education
so that we need to rethink what we're doing
in order to actually invest and think about learning
and put learning first.
And in that respect, I think AI is not just enhancing learning,
it's asking us to really pay attention to some questions
we should have been thinking about for a long, long time.
There are many, and I'm sure we'll get into this conversation
many different ways in which we can do that,
but I'm out of time.
So I'm gonna stop here, but thank you.
(audience applauds)
Thank you. I know we move on to round two and round two
is where the debaters have conversations with one another.
They can question one another, challenge one another.
But I just wanna summarize what I think I've heard
in the opening arguments.
I've heard Gary Markis and Kathy O'Neill argue
that AI is problematic in the classroom
because in the college classroom,
because it fundamentally undermines
fundamental processes of education that learning is hard.
AI makes it seem like learning is gonna be easy.
There is what they call a cognitive surrender,
the outsourcing of the work,
being focused just on the prompt,
not understanding necessarily the output
and not needing actually to grapple with it.
They do acknowledge that there is some good,
but they say that most of the kids,
most of the time, are gonna succumb to being lazy,
and lazy is hard to get around.
They all also made the argument that friction,
and I'm bringing this up because the other side responded
to it so directly that friction is part of the process
of learning for all of us,
and that AI shortcuts that feedback loop
that depends on friction.
It fails if they say, especially for kids
who don't have resources,
and that that's, I think, another piece of this
that we wanna discuss.
The opponents on the other side, Eddie Maloney
and Ravi Pense argue that, yeah, there are problems,
but it can be done right.
They talk about making it a powerful force for learning.
If you place humanity at the center,
they say that the AI can provide
just so much information at scale.
The argument about friction came back the other way
that the colleges have traditionally been there
for the sake of higher education to reduce friction
that that has been the tradition,
but they also say that AI is,
that the way the colleges have responded
to the need for learning, they've made mistakes,
and that AI as its functioning now
is actually exposing some of those mistakes,
and I think we wanna talk about that as well.
But I first wanna take to the side that's arguing
that a-generative AI stunts college learning,
I'll take this to you, Gary.
Your opponents are saying, yeah, but if we do it right,
it's gonna be okay, and that's a great big hole
to drive a big truck through on their side of the argument,
and I just wanna see what your response is to that.
- Well, I mean, first of all, better to do it right
than not do it right, and there's certainly some room
from renewing and doing it more right than less right,
but it's hard, and it is relative
to the current technology that we have.
So no matter how flawed or good the technology
is whether we're talking about this technology
or some other technology,
there's always gonna be ways of using it better and worse,
but this current technology remains flawed,
and it remains this way as Kathy notes
of avoiding the friction, the hard work of actually learning,
and maybe the responsible way to use it
is after you've actually learned your critical thinking skills,
but that's maybe after college for a lot of people,
and it seems to me like a promise or a note.
Maybe eventually we will find better ways to use this tech,
maybe eventually we will have better technology,
and then we'll be able to handle it.
The reality is that most students aren't going to use
these things responsibly.
Most students are gonna be like,
this is a little tool, I can use it to save me
from doing my homework, which is not actually
the responsible thing to do, but it's the tempting thing,
it's the easy thing, and so that's what happens
most of the time.
- So your framing is right now,
and for the foreseeable future,
the way that AI is available in the classroom
remains problematic, that we're way down the road
from learning how to do it the right way.
Do I have that correct?
- Yes.
- I just wanna take that to Robbie.
- So I think Kathy makes interesting points,
but respectfully, I disagree with some of them,
especially the notion that students are lazy,
that's a massive generalization.
I'm still in classroom every day.
My experience has been that students at Michigan,
and I'm sure that no exceptional Johns Hopkins
are committed, dedicated, and are willing to work hard
if you design experiences for them that are responsible.
So AI should be used as a learning mechanism,
not something you derive shortcuts with.
So the responsibility is also on faculty members
to design those learning experiences that are challenging
that are engaging,
and that allows students to leverage AI
when it is appropriate, and also turn AI off
when it is not appropriate.
And that's part of college learning.
I, yes, we all make mistakes,
but through those mistakes we learn.
Making mistakes and learning from those mistakes
is also part of college education.
So I sincerely believe that experiences can be designed
to do it right.
I mean, I don't know, that was a study,
Gary was referring to in his opening comments,
but Harvard had done an experiment in randomized controlled trials
with about 194 students taking an undergraduate physics course,
and they actually found when AI was integrated
and designed properly, the gendered AI tools,
that these students learning significantly improved,
and they actually learned faster
compared to their peers in an active learning classroom
who did not have access to the AI tools.
So there is evidence, and at Michigan,
we've been added since 2023, have you gotten everything right?
No, but through the experiences we are learning together,
we are co-creating together, and the goal is,
we have to engage with these tools together,
because if we engage, we can actually shape it.
Or somebody else is going to shape it for us,
which would lead to challenging results for everyone.
- Yeah, I'd love to reframe this a little bit,
because I feel like speaking as somebody who works
in lawsuits against big tech companies,
it's really hard for me to hear that students are lazy.
I want to say it this way, big tech companies
are hijacking our brains.
And they are hijacking young people's brains,
just as much as us, by the way.
We probably all were on our phones a lot today, right?
But they're hijacking young people's brains
before they have developed the critical thinking skills
that we developed in school.
And so it is a critical, urgent problem for that generation.
My son, who's a freshman in college now,
was a freshman in high school four years ago.
His teacher, his English teacher, realized
that no one of his kids knew how to read.
Like, literally couldn't sit down to read a book
'cause they were scrolling all the time.
So he'd spend his time in class having them read.
That's what they did in class every day.
And they got better at reading.
But that's kind of a technology, right?
reading, it's kind of a technology,
but it doesn't remove the effects of everything else
in their lives.
And that's really what I'm saying.
We could have narrow AI in a classroom that's good.
It's not going to negate the effects of the AI effect
that they've been living through since childhood.
- I wonder, are we to some degree going to be talking
about screens in general in addition to AI
because it sounds like you're also talking
about screens in general.
I think that AI is a natural extension
of what we've been living through with social media
and phones, yes.
- Any response to what Cathy had to say?
- It does come down to how we think we use the technologies
and tools that we have available to us.
And it's certainly the case that we can use them poorly
and we can think about our engagement
with social media as an example of that.
I think culturally we've probably used social media poorly
and it's led to all sorts of detrimental effects
to our society, arguably, but there are also productive ways
of engaging with people socially and we know that.
And we know we can do that as a society
and we can build ways to do that.
And the fact that we may not be imagining those
and we may not be enacting those
is not necessarily an indictment on AI
or any technology that's an indictment on us.
It's an indictment on our engagement with
and our thought process to decide where we should be teaching
our kids about how to think about not only the tools
that they have in front of them,
but they're innate and learned capacities.
That takes time, that takes energy
and to my opening statement point,
we haven't built the structures of education
to enable that and so partly what we need to step back.
- Does that mean you agree with Gary that right now
it's pretty bad, there's a lot of harm being caused?
- I think right now, so one, I don't agree,
I actually don't agree that the tools
are horrible, terrible and bad.
I think they, you didn't say they were horrible,
terrible and bad.
- It's a matter also on net, right?
I mean, there are gonna be some circumstances
where they help and some, where they hurt,
the debate that we had three years ago
was about generative AI writ large
and whether it helped or hurt.
And obviously, there's some places it helps
and some it hurts and that's gonna be true in education.
Sometimes it's gonna help a student find information
they couldn't find, but for me on balance
the extreme temptation to outsource so much to it
is what carries the day and makes it net negative.
- So that's the question I'm taking to you.
He's describing this situation now.
A lot of damage is inherent.
And I'm just asking if you're conceding
that that's actually, I don't think that that's true
across the board.
I don't think it's necessarily true.
I think both those are different.
And I think we have plenty of examples
where we, our students, our colleagues
are building AI literacy and the ability to discern
where they can and how they can use these tools
effectively and meaningfully.
I have two students in college, two students, two kids
in college who happen to be students.
Both of them think about their engagement with AI
in ways that I'm actually pretty impressed by it.
It's not transactional.
They tend not to take a prompt, stick it into AI,
take the results and turn those results in.
They see these tools as tutors.
They see them as helping them think through structure.
They think of them as debate partners,
which we don't always have the luxury of tables
across the stage to be able to talk to people.
They think about these tools as actually helping them
practice things that are harder for them to perform
in spaces like this or in a classroom.
It gives them access to information knowledge
in new in different ways, which used to be linear,
but now actually is much more complex.
So the tools themselves, I don't think are inherently wrong.
It really comes down to how we use them.
I think even that concept, when someone says,
I used AI, what does that mean?
There are so many different possible ways
of thinking about that.
And so when we think about it in education,
we think about it in learning.
I think it really does come back to that sense of,
what does it mean to use these tools effectively,
meaningfully, thoughtfully, and we need to get to a place
where we're doing that.
And I think we are, I think many institutions,
are looking to reimagine their work right now
to engage with these tools.
That doesn't answer the future question,
what's going to happen when we're thinking differently,
when the kind of cognitive offloading that's been happening
changes how we actually structure and engage with things.
All of those things, I think will happen too,
but I'm happy to talk about why that can be positive as well.
I feel like one of the tricks, one of the traps
we might fall in on the stage is to think
that we need to solve the problems caused by AI with AI.
The truth of the matter is, I feel like read a book
when you come into English class,
was solving a problem of social media with a book.
The best ideas I've heard, personally,
about solving the problem of AI is blue book tests
and oral exams, and my problem is that.
I was just for people who weren't around
when blue books were a thing.
I remember them, but can you tell people what a blue book is?
Yeah, you actually write and you use your penmanship too,
in a notepad, I love notebooks.
I mean, I'm like an old school person,
but my point is that like, these are solutions
to the problem of AI that have nothing to do with AI.
I'm not saying there aren't good AI tools,
but I don't want us to think we need to solve AI problems
with AI.
Mark Zuckerberg went in front of the Senate
and when he was confronted many years ago
about the problem of fake news,
he's like, we're going to solve that with AI.
And I remember thinking, no, you're not, you liar.
And you know you're a liar.
But like, I feel like--
I actually wrote an op-ed at the time in the New York Times
saying that he was lying maybe eight or so years ago
when you did it.
And I said, there's no way he's going to actually--
He knew that.
--deliver, and he did not live.
He knew that.
But let's not box ourselves in.
That's what I want to say.
Let's think, how do we get kids--
how do we give kids the friction, that they need?
Is AI going to do it?
Not the AI that's being written now.
Not the Gen AI.
That's free now.
We'd have to really narrowly define a product.
Ravi, how are we going to do it?
How are we going to give kids the friction?
So I mean, do you like Columbia University,
I believe, now require--
no screens in the classroom.
Blue book exams are happening.
Real text on paper, read in class.
Those kinds of solutions are meant to reinstate this kind
of friction that's being talked about here.
What do you think of those solutions?
I think the way I look at it is, when
it comes to classroom, a appropriate faculty member
is in charge.
That's their domain.
What kind of policies they want to use for their classroom--
AI or no AI, blue book or no blue book,
or whatever they want to do.
It's up to them how to design those experiences,
because they know they're students better.
And the other thing I want to point out
is that when students are coming to college,
they are adults.
And we have to treat them as adults.
And we have to work with them.
We have to take them into confidence.
Some of you may think I'm naive.
But really, if you take them into confidence
and talk about integrity and how we want to approach this,
talk about this tool as a very powerful tool,
but also has potential negative harmful effects.
It can negatively impact our critical thinking skills.
So we have to be mindful and aware of that.
If you approach it that way, as a faculty member,
we can design experiences that will enhance learning.
And at the same time, control for some of the challenges
that you're talking about.
A faculty member is completely free to have them
read in the classroom, like you're
pointing out for what your children, that's up to them.
A faculty member is completely free to give blue book exams
if they choose.
Or they can continue giving take on exams.
Or they can evolve their pedagogy.
We cannot be teaching the same way we did 20 years ago
and claim that, oh, it's AI slop and people are just cheating.
I'm getting the same old papers that I was getting
over and over again.
So that's where we have integrity conversations
at higher level.
So how a class is managed is a faculty member's domain.
And I want to keep it that way.
I'm not going to go and advise a faculty member
what one should do in their classroom.
It's up to them to design the experiences they want
in consultation, working with students who are adults,
and who are actually quite committed.
So Gary, I hear Robbie talking about things
to do to make it work better.
And I'm curious--
and you've conceded that that can happen.
And down the road, that can help it will.
But right now, would your prescription for doing this better
include, I don't want to use the word ban
because it's hard to control, but fundamentally,
having a role that students can't use AI
for some period of time, for a period of time
during which they are developing the critical thinking
that Kathy is talking about, should there be a no AI?
In a dictatorship, which I do not wish to live in,
I would demand that.
I think the students would be better off.
In the real world that we have, we can't.
And we're not really going to be able to stop kids
from having cell phones.
We're not going to be able to stop them from using chatbots.
And so I think we have to use other techniques.
One that I was briefly famous for was saying,
have your students write their papers with chat GPT
and then go through the answers and ask them,
where are the hallucinations?
How else could you write this essay
to make it more interesting and so forth?
What we need to do is to find techniques to live with the menace.
- Did you say the menace?
- The menace, I really do think it's a problem right now
for college students to have this too available.
I just want to say something about friction here.
There's actually two kinds of friction.
There's the friction of learning that you don't actually
understand something, which is kind of uncomfortable
and painful and I think is what Kathy is talking about.
And there's a friction of like going to the registrar's office,
wanting a copy of your transcript so you can send it
to somebody and they want too much money
and too many forms you have to fill out.
Like we should certainly be reducing the friction
that gets in the way of you studying.
But there is the friction of this is hard
and I got it wrong and I need to learn it now.
Like for me, the most important thing in graduate school
was learning the friction that comes from thinking
you have solved the problem and discovering that other people
can look at the same data and find another explanation.
And you've got to do like five more experiments
for you have anything convincing.
That is the friction that Kathy I think was talking about,
where you learn that you don't really know at all
and that you've got to work hard
and I'm afraid with these systems too easily in place,
the kids are not really learning that.
- And one point that Kathy made in her opening
was that there's a socioeconomic dimension to this
and that students who come from
more challenging socioeconomic backgrounds
are gonna be more susceptible to the urge to shortcut.
Offload cognitively, et cetera,
because they're not surrounded by people telling them
that's a bad idea.
Can you take on your response to that?
- Yeah, I'm not sure that that is true.
I mean, I think there are probably a lot of different places
where the inequities of the availability of the tools
will be problematic across socioeconomic levels and degrees
but I mean, part of it would be just access to the tools
when you have, for example, wealthier students
who can pay $200 a month or now $500 a month
to gain access to the best models and the best tools
and you know, students who can't
then we create a different kind of inequity.
You know, I think we have inequities in education
across and have for decades, centuries.
That are not, I don't know that are necessarily
any more exacerbated by this particular problem.
- Well, as I think the optimist about AI
and academia on the panel, how do you respond to that?
- To which part, sorry.
- To the fact that you're conceding that she's right
that there is an economic disparity.
- I think, so two things, one, I think it's incumbent upon us
to understand the challenge that we have in front of us
and to design for it and to not ban or police
or to suggest that we're gonna be able to keep everybody
away from things but to actually intentionally think
about the places where access is necessary,
where it actually creates equity to actually design
for the learning environments,
where it can take advantage of those tools
but also take advantage of the blue books
and the oral exams and the things that we're actually
returning to in an amazing way.
I mean, that we've abdicated our responsibility
around eloquence and oral capacity
and capabilities in higher education for a long time
but that used to be a fundamental component
in the 19th century of what an education would mean
in higher education.
We've lost a lot of those capacities
because we have moved more to a print culture
and we've created structures that require scale
and required us to produce and understand knowledge
in a particular kind of way.
So I think there are ways in which we need to design
for not only learning but design for those inequities.
- Kathy, you look like going to talk about that.
- I just wanna tell a story.
So I'm a good friend who teaches at UMass Boston
which is a school that you can basically get in
if you apply, if you're about, if you graduate from high school.
So a lot of those kids are from poor neighboring towns.
Their education hasn't been good.
I'm pushing back a little bit against Robby saying
these are adults, treat them like adults.
My teeth, of course, my friend in their English seminar
tells them not to use chat GPT, they do anyway.
They think that they haven't cheated
because they think that they just got a first draft
and then edited it.
The way she gets them to understand the divide
between what they think is true and what is true
is she asks them to read them out loud
and they can't pronounce the words.
They don't know what they mean.
And they didn't know that until that moment.
And she's not doing it to shame them.
She's trying to explain it to them.
You guys aren't learning by using chat GPT
and that's what we're talking about.
The friction, she caused the friction there.
Thanks, they goodness for her.
But that's a kind of friction
that most teachers don't have time to do.
(gentle music)
- Okay, let's go to some questions.
I have one request because we're at Johns Hopkins,
I would really like to give preference
for the first two questions to Hopkins student.
So if you're a Hopkins student in the front, go for it.
And one other person gets to move down.
And if you don't mind telling us
to lease your first name with your question.
- Hi, my name is Natalie.
I am a senior studying psychology political science.
And I'm currently, as a psychology major,
I'm currently taking a class on neurodiversity and learning.
And part of what we explore that class
is education and how we can make education
more inclusive and accessible.
And one thing I've been thinking about
particularly with your guys' conversation
is this idea of friction
and what kind of friction is necessary
versus what kind of friction
in pairs of learning process.
And I'm wondering, you know,
when it comes to traditional cheating versus AI cheating,
a lot of, I'm wondering if you guys could kind of explore
the question of what is creating that friction
that's causing students to want to be there
in the first place?
Like why do students feel so overwhelmed
that they feel the need to turn to these tools?
And how come, like, I guess what I'm trying to say
is students. - Are you saying is there a reason
that the students are drawn to doing this in the first place?
- Yeah, and it's tied to other issues like career funneling.
Like to what extent are we making education,
people answering to a set of arbitrary demands
versus really focusing on individualized learning
and clinical thinking?
- Let me take that Teddy,
because in your opening,
you alluded to thinking about a lot of these things.
So I wanted to take that question on first.
- Yeah, it's a great question.
Why are students doing this?
Why are we not as students?
Why are we not thinking about
what's maybe beneficial for our learning first?
Why are we taking shortcuts?
You know, at Georgetown,
and I imagine it's not too similar to here,
our students are oversubscribed often.
They have many different competing priorities.
They have internships, they're double majoring.
They are trying to. Many of them are working at the same time.
They have all of these things that are happening
that we've put pressure on them to find ways through,
of course, this is not a new problem.
And often what would happen maybe prior to having access
to these kinds of tools is making choices.
I'm not gonna go to that class.
I'm not gonna do as well on that.
I'm not gonna study as much in this space,
because this is my priority,
and this is the choice
because those competing priorities are too heavy.
So part of it is asking ourselves,
why are we putting so much on students?
Some of it's internal institutions,
some of it's society saying,
you're only gonna get a job if you double major,
and you have these internships over the summer,
and you do all of these things.
And so we're putting so much pressure
on young people to try to figure out their next step
and where they're going,
that I think it leads to a particular kind of
set of prioritization and choices.
And right now, one of the easy ways
of making those choices is to use these tools.
- Gary, are your thoughts on that?
- I'll just add that competition is another part of it, right?
I mean, let's say you're in a class
that's graded on a curve,
and your friends can get it output in 30 seconds,
and you think they won't really learn anything,
and you could learn a lot more if you spent three hours,
but then you don't have those three hours
for something else.
And so it's a matter of time management,
and it's a matter of competition and feeling like,
I don't want to be the chump.
Everybody else is doing it, I guess I should too.
- But it sounds like you're both agreeing
that this is not exactly AI's fault.
This is the system's fault.
- No, the point I was trying to make earlier,
and I don't know if I made it really well,
is that some of it's an interaction between
what AI happens to do in human foibles.
This weight is with a phone.
I mean, phones don't have any sort of intrinsic moral stance
or whatever, but they feed on our desire for approval
with the likes on Twitter or whatever,
and they feed on our desire
to see more information as fast as possible.
And so it just happens to be this kind of a bad fit
between the software that we've built for phones
and how human minds tend to work such that
it does kind of wind up exploiting people.
It wasn't like by design.
- Yeah, it was by design.
I'm sorry, it was absolutely by design.
- It was by design by people like Facebook,
but not people like Jobs, necessarily.
- I'm sorry, I can't go ahead then next question.
- So I just want to jump in and make one other comment
while I understand the pressures that students will deal with
because of competing priorities and so on,
but also behaves faculty members
to think differently about their pedagogy.
It's really, really important to design
the right kind of learning experiences
where the students are not put in the kind of pressure
that you are talking about
where it just makes it easy to type your prompt in
and get an answer and submit it as your own.
Having those proper learning experiences is critical,
so I've been urging and encouraging faculty colleagues
across and students to take joint responsibility.
The future with the AI can be bright
if we are collaborating partners, faculty members,
students and others.
- Next question, please.
Thank you very much for your question.
(audience applauds)
- Hello, my name's Leon
and I just had two questions about one about--
- Can you, I just can allow one question.
- Yeah, thanks. - Yeah, thanks.
Could you get a little bit closer to the mic?
Just a touch, thanks.
- I was wondering how you think AI will change
the length of a degree and also the significance
of the university as a physical place
because it allows more virtual education opportunities.
- Are you asking me?
- I'm looking at you.
- You're looking at me.
There's a lot of discussion right now about a three-year degree
and what that would look like.
I mean, you're up three-year degrees are common.
So I think one of the things that we value
in this country is formation and spending time.
I actually think one of the greatest assets
for higher education is time.
Our ability to spend time thinking and engaging
and stick with colleagues, with peers, with faculty,
with research, with people who came before us,
people are gonna come after us at the time
as such an amazing commodity and asset for higher education.
So losing that I think would be a challenge,
but there are, I think there are some momentum
in that direction.
But what it means then for higher education
to be viable into the future,
I think it's a second part of your question,
really depends on how you're doing.
how we make these choices right now?
Are we going to become insular and decide
that we have a particular value that speaks to what we
thought learning and knowledge was for the past 50 years,
maybe longer, though that's changed dramatically
since the internet, in fact, in many ways.
Or are we going to adapt to different ways
in which we're going to engage with tools, technologies,
but also information, information, construction,
development, knowledge production?
If we don't adapt to those things,
I think we're going to become very procured
and maybe obsolete pretty quickly.
I don't think there's anything magical about four years
in education.
I graduated college in three years.
I learned what I needed to know for my next step, which
was graduate school, about how to think and run experiments
and so forth.
I think that if AI could condense the learning experience
into three years rather than four, that'd be great.
I think it's just right now we're not quite getting
the things out of education in terms of critical thinking
and problem solving that I would like to see as get.
I don't think the particular amount of time is magical,
and maybe that will change someday.
Next question.
John, can I jump in on this one?
Do you mind if we go to another question?
Because we're a little short for time, thanks.
Hi, good evening.
My name is Guggan.
I'm a teaching professor with the Department
of Computer Science at Johns Hopkins.
I'm on the side of stunt.
So my question-- I have two points which I want to share
or raise with the enhanced side.
One concern that was raised by you is scale.
That modern education system causes a problem of scale.
There are lecture halls with a lot of students.
You don't have one to one tutor.
And AI can solve this problem.
What I want to say that 30 years ago, when I was a student,
I could solve it with a technology, which
was free at the time.
And that technology is free.
But now it's called Make a Friend.
That'll always be free.
Just chat with somebody in the class, make a friend, study--
I apologize.
I just need it.
We're so tight for time.
And I want to discourage comments.
But if you could just ask a question that comes out of your thought.
OK.
So there's no question on this.
The second question--
Do you mind if I pass to somebody else?
Think there's people waiting in--
I have a second question, which is not related to this?
If you can go straight to it right now with a question
mark at the end of it, I will allow it.
Oh, thank you.
OK, thank you.
It's tough to do.
My name is Patrick Roddy, and I'm an alumni.
Is, quote, "college learning the processes
and institutions of higher education?
Or is it the experience the student takes
and then uses to do the tasks which make life better?"
Thank you.
I'll take that to Kathy.
Yeah, I mean, thank you for asking that.
That's kind of one of my points.
By the way, perfectly for his question.
Yeah.
Nice and short, sir.
I appreciate that.
Like 10 years ago, I have a blog.
I blogged incessantly about my fears
that mathematics as an industry would be over
because there's these massive online open courses
that anybody can take.
And oh, my God, all these calculus teachers that I loved,
they're going to lose their job.
It didn't happen.
Why didn't it happen?
Because people don't take massive online open courses
to completion.
I just looked up the statistics.
Less than 10% of people finish it.
And the people that do finish it make friends who
are also taking that course, which
was not at all the model when it was first introduced.
It was basically like, oh, everybody
has access to education online.
It's free.
So everybody will just learn calculus through the vapor.
That's not what happens, right?
It's about an experience you have with your teachers,
with your mentors, with your friends.
And that's a critical part of learning.
Comment from the other side, and if not,
I'll go on to another question.
I'm keeping humanity.
Keeping the human at the center is necessary for everything.
I mean, I think one of the problems we're running into right
now is we're thinking that we have to keep human beings
at the center because they can do things that AI can't do.
I think that's fundamentally flawed.
You mentioned the popes in cyclical.
Human beings have inherent dignity.
Our relationships to others, we should be investing in that.
I don't think the tools necessarily matter.
It means that we are committed as a society, as a species,
to making sure that we care for each other,
that we recognize every individual's human dignity.
In the process, I think that you'd outline for our education.
I think it's both, actually.
But I think your question is important.
Next, please.
All right, good afternoon.
My name is Dwight Hargrave, and I have a quick question.
I'm a K-12 educator.
And my concern is, has the horse already left the barn
with the issue we were talking about this evening?
I think that's a great question.
You add the fact that open Chinese models
are basically free, not that compute is free.
But just think about the actual market of AI.
It's not going away.
It's not going to be, unless the internet goes away,
which would be fine, which would be fine.
AI is going to be out there for students.
And basically, what you have is at scale,
a dumbification machine that makes it harder to teach kids.
It makes it harder to create friction.
And then you have individuals who have to go around teachers,
parents, friends, and make sure that the friction is still there.
So it's actually giving us more work than it's giving us
more work to educate the next generation.
I think the way to address that is institution-like Johns
Hopkins and Michigan have a responsibility
to work with our K-12 partners and partner with them
and provide them the resources.
At Michigan, we have been routinely doing that every year,
working with K-12 teachers.
And sharing ideas with each other and ways to enhance
and continue to maintain our critical thinking skills
in the age of AI.
Because as Gary points out, it has left the bond.
So it's not something we can ignore.
So the question is, in the age of AI,
how do we ensure that our critical thinking skills
are maintained, supported, enhanced?
And people are aware of the consequences that has to happen.
And that requires partnership, including the AI.
The mic is yours.
- Good evening.
My name is Savannah Robinson.
I'm a student at Western High School.
And my question for you was having acknowledged
that students are so overwhelmed with work constantly.
How do you advise is the best way to go about studying
or not necessarily studying, but keeping that hope
and what you refer to as friction
without feeling the need to use AI.
- I have sympathy.
My son wants to be a mathematician
and for all sorts of reasons, that's a terrible idea right now.
But how does he keep himself when he's really stuck
from asking AI?
When I was his age, there was no place for me to look
at for the answer to a homework.
I mean, there probably was.
It wasn't a new problem.
I wasn't doing original research.
I didn't know where to look.
So it's irresistible.
And so I feel sympathy for you.
I would say that the best thing is to go with your friends,
have a peer group where you guys all agree,
I'm not going to use AI if you don't use AI.
Let's have an anti-AI group.
And that would probably work better
than just trying to do it individually.
- Do you all fundamentally feel that
students should be trusted to make their own decisions
about this or should there be?
- It's not fair to put the pressure on them to be,
they're not responsible for this mess, right?
So to be like, oh, you're responsible.
It's sort of like free cocaine.
And just don't take it, you know, like, it's not fair.
- Well, Ravi, you've got a couple of times
that we're talking about adults here.
- Yeah, I mean, in case of young women,
she was from high school.
So I do want to address that there.
The idea there is, you have to think about
in terms of prioritization and also work with your counselors
and teachers to find out what classes you're taking,
what activities you have that's overwhelming you
and pay attention to mental health and wellness balance.
That's really, really critical among all the young people today.
And that requires getting partnership with teachers,
counselors, parents, working together.
When we are talking about K through 12 students,
when I'm talking about college students,
is where I'm referring to them as adults
and working with them and trusting them
and taking them into confidence.
Sure, if they make a mistake, that is a teachable movement,
working with them to show that this is a teachable movement.
And, you know, we had a situation where recently,
a pro host is in trouble because of his use of AI
in one of the institutions.
And I see that as a teachable movement.
And so this is where we need to have conversations
about kids and support them.
Next question, please.
And thanks for staying coming from high school
to join the program tonight.
We really appreciate that.
Joel Furchis, writer for Medium.
So the difference between textbooks and AI
is that textbooks can't talk back to you.
There's no interaction that occurs there.
All the AI on the market right now is from big companies, right?
Universities have the ability to produce technology as well.
So what would you think about universities producing a type of AI
which serves as sort of a so-cratic interlocutor
instead of an asecophant?
Eddie, I'd like you to take that question first.
And I'll come to you, Robbie, because you--
There are plenty of folks who are creating chatbots
that actually work in that way.
I don't think you have to actually reproduce an LLM
or you have to have your own model to do that.
I mean, I think there are different ways of creating front-ends
and setting up agents to do that kind of work,
depending on what you're trying to accomplish.
So that is happening.
The AI tutorial model, Robbie mentioned
the Harvard Physics Studio course
where one of the professors saw 60% improvements
across multiple years using AI.
But that is an AI tool that is designed
and set up to be interactive.
It's not to give answers.
It really is designed to push, which I think is part--
one thing I would say to the answer to the last question.
I wanted to. work is about endurance, it's about building your capacity to learn, it's not something
you can just walk into and do, you have the friction matters, being able to encounter
friction matters, AI tools can help to generate that friction, it depends on how we use them,
it depends on how we think about them, I don't think they're the only solution, I take
Kathy's point that we can't always use AI to solve the AI problem, but I think sometimes
we can, and I think both of those things can be true.
Gary, are you, are your concerns a swage by the idea that schools can develop an AI that
is responsible and serious?
I don't think we can't really do that yet, well, I think that someday we may be able to
do that, I think really good educational software would start with a model of how the child's
sort of student's mind works, what they know, what they don't know, what they still need
to learn, and I can at least envision that, but I have trouble envisioning
the current technology, which I've been very critical of for many years, being able to
do that well, I just, I don't see that coming out of the large language model.
Thank you for your question, I just want to let you know, we can take two more questions
and then we're not going to have time, and it would be really helpful if your question
can be really tight and the responses can be really tight, please.
Yes, hi everyone, I'm Tanika Alexander, I'm an alum at the University of Baltimore,
and I just accepted the yes to the alumni panel today.
My question is, what's the quality, the quality equivalent in an AI world, like who would
make sure that it just doesn't work only just for students who don't have the support?
Good.
I mean, part of the problem right now is it's the Wild West, like the things just go out,
we have no regulation around any of it, like you could imagine a more thoughtful response
from the federal government than we've seen.
I mean, I actually audit algorithms, technology and education that aren't AI, and they don't
ask that question either.
They often don't ask the question, does this work for kids with autism, does this work
for kids with learning disabilities, does this work for different kinds of learners?
You know, so it's, I fundamentally a problem in education, especially with technology.
Next question, please, thank you for your question, even I'm Lonnie.
So it's probably true that I am among the generation of students that are going to
be the test subjects.
You spoke about universities and internet.
These are mediums that have drawn us to progress our history.
But might or might not make this just another incremental innovation characteristic of human
progress.
Yeah, I think that's an interesting question and it's something that's come up in a lot
of the base.
We've done a big technology that every technology scares people, television before that radio
than the internet, then social media, now AI.
Is this just one more period that sounds rough and we're going to get through it fairly
easily or not?
I think that first you're obvious to the other side.
So I think you raised a very good point.
Obviously, we have been with internet for, you know, early 60s when internet was developed
and we still haven't resolved all the opportunities and challenges that internet brings.
So AI is, the difference with AI is the agility with which the changes are occurring.
Gary talked about how models are coming, it's out wild west and models keep coming every
few days at University of Michigan, we have taken the steps to develop our own harnesses
to provide the right kind of guard rails to also provide learning opportunities and
experiences.
And I think that's the approach if we were to take with AI tools.
I think our future is going to be fine.
Last word from the side.
I'll just say that like expanding the question a little bit, like kids aren't done.
Like the kids that are in college today are just as intelligent as my generation or
anybody else's generation.
And they are going to figure out that this isn't working for them in one way or another.
And one way they are going to find it out is that they're going to have trouble getting
a job because AI has invaded the job market.
And that's another part of the experience of being in college is also the experience of
trying to get a job afterwards.
It's almost impossible in the context of AI.
So I do think there's like an urgent need for this conversation.
That's one of the reasons I'm glad to be here tonight.
I have this conversation about what's the future of AI college learning with respect in
the context of AI?
But what's actually the future of college in the context of AI?
Thank you very much.
And that's a wrap on questions and answers.
Thank you everybody who asked questions.
And now we move into our closing round and our closing round.
The debaters once again stand up and make a much prefer statement, closing statement.
One more time to try to persuade you to vote for their side in the argument.
You have two minutes to make the case that generative AI will stunt college learning.
I think the last question made the most important point, which is that all the students who are
dealing with this right now, they're all guinea pigs in an experiment they didn't sign
up for.
Sign up for.
They are just part of this.
They have to go along.
They didn't sign a consent form to be in an experiment or anything like that.
Because it was one of the most peaceful debates I've ever been in, we hardly disagreed
about anything.
And I think Robbie made one of them, I think it was Robbie made one of the most important
points, which is we need to think differently about pedagogy now.
And I think we all agree on that.
But the reason we need to think differently about it is because right now the way that
it's working out is it is causing a lot of harm to a lot of students.
Maybe that's the one thing we might disagree, but probably it's a matter of proportion.
I think right now the AI that we have and the ways that we're using it is causing a lot
of harm to a lot of students and that's why we need to think differently about pedagogy.
Companies like OpenAI and Throp, I have dumped a mess on our feet and we're trying to do
the best we can.
The students are trying to do the best they can, the schools, the deans, everybody's trying
to do the best they can.
But the reality is that the natural thing for this system to do is to hijack children's
experience with the friction that allows them to do the critical thinking, learn the
medical condition that they need, and so for now it's not a great thing.
Thank you.
Thank you, Gary.
[Applause]
Eddie, you're up from the opposite side.
The floor is yours.
When writing as a technology came into being, people were concerned about its effects on
our ability to memorize, when the printing press came on, people were concerned about people
having access to information.
When the internet came about, I had colleagues, I was in graduate school, commercial internet,
not the 1960s or internet, but colleagues who were concerned that this was going to provide
students too much information and they were going to be able to curate that information.
So we needed to cut back on all of these things.
We're hearing similar kinds of arguments with AI in part because there are things that
we value that came before these technologies and we want to hang on to what those are.
And part of our challenge right now is not to say, or think about, what is AI doing to
the things that we valued, but what are they doing to the things that we need to value
as we move forward, how are we going to adapt to this moment, how are we going to adjust
to this moment.
There are going to be some things that are probably going to get lost, but there will also
be things that will be gained as there have been with all of these other technologies that
have allowed for what we call cognitive offloading in particular kinds of ways or that cognitive
struggle.
So I think we're at a moment right now where we have to make some choices about how we
design for that future, how we think about where we're going.
And if we don't, then I'm just going to go sit over at that other table because I think
that is exactly right.
It will continue to stunt our learning.
But I think as we also heard, we can stunt our own learning quite well on our own.
We don't need the technologies to do that.
We make choices all the time that inhibit our learning.
We make choices to use technologies in ways that don't enhance our learning.
So it's also incumbent upon us to decide how we're going to engage with these tools, how
we're going to think about them, how we're going to make them part of our practice, and
how we're also going to invest in the things that we care about and we value.
And in that, I think that's the only way we're going to imagine and design our future.
Kathy.
Kathy, it's your turn, Kathy O'Neill.
I'm so glad I came today.
This has been a really fascinating-- I'm going to look at the clock this time.
It just brings up-- it brings up when you ask, like, what will happen with college education,
the context of AI?
What will happen?
I mean, that's a big question.
And it involves politics, actually.
It involves the future of capitalism.
What will happen to AI companies?
Like, if capitalism goes on the way it is, they're still going to be around, right?
What is college for?
I mean, who is college for?
That's what I keep wondering.
And I kind of think these are the same questions.
The reason I say that is because an educator friend of mine from California mentioned that
in California, community college is free, it's two years, and there's small classrooms.
And it solves the problem, ideally, of having people of all classes needing to be educated.
I think we all agree that high school isn't enough anymore, and it's class blind.
And ideally, with small classrooms, you could actually overcome the problem of AI because
you have personal attention, you have the opportunity for personal attention.
I might sound like I'm saying, here's the way we should do this.
I just don't think that's likely.
I actually don't think it's a political bitch.
big ask for us to do that.
It's a political, big ask for us to say,
we're going to ignore the very loud noise
of the big tech companies and all their money,
and we're going to prioritize education for everyone,
but that's what I think we would have to do.
I just don't think it's gonna happen,
but I want it to happen.
Thank you.
(audience applauding)
- Ravi Pense, floor is yours.
Thank you, we had a really wonderful conversation so far.
So I'm going to end it with a very quick story.
Todd Holland is a neurosurgeon at University of Michigan,
and Todd and his team, leveraging responsible AI,
developed a technique which allows them
when they operate on patients with malignant tumors,
instead of sending patients back
with 25% of the tumor remaining
to now send patients back
with less than 4% of the tumor remaining.
That's the power of responsible AI.
That is the power of saving lives.
So the question in front of us is,
do we want to be a Todd Holland
and make a difference in save lives
by using responsible AI?
Or do we want to bury our head in the sand
and not leverage this tool that is in front of us?
But we have to approach it as a student of AI.
I consider myself as student of AI.
I believe that everybody should be a student of AI,
not an expert in AI.
Don't consider myself that at all.
So the key point here is, as we approach this technology,
and we have heard a lot of different conversations around it,
it is really important to think of it as a partnership
between faculty members, students,
other colleagues and staff members,
and work collaboratively where we can engage with this technology,
and actually it will give us the power
to shape this technology as opposed to outsourcing that power
to people in Silicon Valley.
It has to be the institutions like University of Michigan,
the Johns Hopkins and other great institutions
working together demanding the right kind of responsible AI,
which will make difference in each of our lives every day.
I started with a Sanskrit phrase,
so I'll conclude with one, Vidya Dadati Vinayan.
It means knowledge gives us humility.
The power of AI working together
with everyone involved in this room and beyond
will allow us to be humble and approach future with optimism.
Thank you.
[APPLAUSE]
Thank you, Robin.
All right, that is a wrap on the argument portion
of the evening.
We're going to have the result of the yes and no vote,
the before and after vote in just a few minutes.
I'd like to ask the four debaters to come to Downstage
with me, and let's just have a little bit of a chat.
I found the discussion of cognitive offload really interesting,
and I always try not to make anything about the moderator,
but I will share a thought across my mind.
I remember the first time I used a GPS back in 2006.
It wasn't part of my phone.
It was a separate unit.
I was out in California shooting a story for ABC News.
I had to drive about 65, 70 miles.
I didn't know where I was going.
The first time I ever had one of these garment devices
put it in the visor, and it told me turn left, turn right,
turn left, go for five miles.
And I followed it, and I got to where I was going,
and I thought that was really, really cool.
And all of a sudden, it glitched, it had stopped working.
And I realized, I had no idea whatsoever
how to go back because I had stopped paying attention to it.
And as I was listening to the conversation tonight,
and you were bringing up this notion of non-friction, et cetera,
which you all fundamentally agreed with,
it struck me that I had an early experience of that kind of thing.
And that's why John Muir, to be intentional
about using that garment and other technology,
but also not forgetting the triptych from AAA.
You had that experience long before the garment.
Writing is the ultimate cognitive offloading.
We've been doing this since we learned to write.
This is something we take thoughts.
We put them down, and we put them somewhere else.
We transfer those thoughts.
We give them to people.
We do things with thoughts that are no longer
ones that we hang on to in the same kind of way.
We don't process in the same kind of way.
Writing is an incredibly powerful cognitive
offloading disagreement here.
I think that writing is the way that I actually learn things.
So it's true. Part of writing is maybe
to stick a piece to this side, and maybe that's what you're talking about.
But for me, the way to engage something and to understand it is to write.
There's a famous line I can't quite summon about.
The best way to learn something is to teach it.
Even better way is to write it for a mass audience,
which I do kind of every day in my sub-stack.
That's not offloading.
That's actually forcing me to engage more deeply.
And that's what I think the LLMs are not doing.
For some people, that's really true.
There's a lot of evidence for not everybody that that's true.
And thinking is not always a writing process.
So yeah, the people who write, the people who think about writing,
who care about it, who think that they think to write,
think that everyone does.
So in this case, I agree with you.
I actually think better than I write.
And so that's one occasion that I get in, and I actually agree.
I'm going to disagree with all of you.
The other students who are here are bringing up
the results of the vote.
Before we do that, I just want to say one more time.
Eddie, thank you so much for stepping in the way you did.
That was incredible.
[APPLAUSE]
So I understand you are part of the Hopkins political union
or forum, the union.
The union.
Can you tell me what that is?
Yeah, so I'm actually thrilled to be joined on the stage
by my leadership team, Brittany and Chris.
And as well as Jackson from the Student Government
Association, we have actually been
to every Hopkins forum debate.
So we're first, first of all, very grateful for you
and the team at open a debate for bringing
these important conversations to our campus.
The Hopkins political union is a civic debate
and discussion society that formed as an ongoing partnership
between the Johns Hopkins College Democrats
and Republicans to bring together students and faculty
across the political spectrum to disagree just like today,
openly, authentically, and respectfully
on the issues that matter most to us.
And that's why we are excited to repeat this debate
with among our students and our faculty
as a partnership between the Asana Fagora Institute,
the Hopkins political union, and the Student
Government Association.
And we know that this issue affects all students
and, as well as their future employers, by the way,
and we're just grateful for the opportunity to share that.
Thank you so much.
Thank you for that.
And I'll say thank you.
And you can return to your seats.
Thank you very much.
Again, the question is does a generative AI
enhance or stunt college learning before the--
you heard the arguments, 26.4% said,
will enhance 38.0% said stunt, 35%, 0.6% said undecided.
After the vote, 39.5% said enhanced, 40.3% said stunt,
20.2% said undecided.
Decide that moved up the most was enhanced, up 13.1%.
So the enhanced side actually prevailed in the debate
in terms of moving the audience.
So thank you.
And I just want to encourage you all to shake hands.
[APPLAUSE]
Thank you so much to Johns Hopkins, to our debaters,
to the Asana Fagora Institute, and to all of you
for attending, for all of the questions.
And thank you, especially to our debaters,
and everybody, have a good evening.
[APPLAUSE]
And now it's time your turn to weigh in on this one.
After hearing the full program, we
want to know which side of the debate you are listeners
fall on.
Go cast your vote now by going to open to debate.org/vote
or scan the QR code on your screen.
We will also drop the voting link in the episode notes
in description.
[MUSIC PLAYING]
This debate was produced in partnership
with the Stavros Nyarkas Foundation
Agora Institute at Johns Hopkins University.
As a nonprofit organization, working to bring together
people with conflicting views to engage in rigorous civil
debate, open to debate is made possible by listeners
like you, by the Rosencrantz Foundation,
and by supporters of open to debate.
Robert Rosencrantz is our chairman.
Our CEO is Leah Mathau.
Andrew Lipson is head of production.
Michelle Debrisini is head of marketing.
Alexis Pengrozi is our executive producer,
Katie Perry produced this episode.
The open to debate team also includes Gabriella Mayer,
Gabriella Yana-Cherley, Linda Lee, Scotchikovi,
and Julia Serkin.
Damon Wittemore makes this episode.
Our theme music is by Alex Clement.
And I'm John Donovan.
We'll see you next time on Open to Debate.
[MUSIC PLAYING]
Whoever says saving money is hard doesn't
know that new customers who bundle and save with progressive
save hundreds.
Because switching your home in auto insurance
is better than cutting down on the things you love,
like those ice lattes that help you get ready to face the day.
So switch your home in auto to progressive
and see if you could save and keep sipping away.
Get your quote today at progressive.com.
Progressive casualty insurance company, affiliates,
and other insurers, not available in all states.
[MUSIC PLAYING]
Podcast Summary
Key Points:
Generative AI in education is sparking debate over whether it enhances or stunts learning, with strong arguments on both sides.
Critics like Gary Marcus and Kathy O'Neill argue that AI promotes cognitive surrender and undermines critical thinking by enabling passive learning.
Supporters like Ravi Pensey and Eddie Maloney believe AI can enhance access to information and support diverse learning needs when used responsibly.
A core concern is that AI removes the necessary friction of learning—struggle, error, and effort—that fosters deep understanding and growth.
There is a significant socioeconomic dimension, as students from disadvantaged backgrounds may be more likely to rely on AI to cope with academic pressure.
The debate highlights the need for AI literacy, ethical pedagogy, and institutional reforms such as blue-book exams or in-person assessments to preserve learning integrity.
Faculty responsibility is emphasized—teachers must design experiences that balance AI use with critical thinking development.
Ultimately, the impact of AI depends not on the technology itself, but on how it is integrated into education with human-centered, thoughtful design.
Summary:
The debate centers on whether generative AI enhances or stunts college learning. Critics argue that widespread AI use leads to cognitive surrender, where students passively rely on prompts to complete assignments, avoiding critical thinking and deep engagement. They highlight that AI undermines the essential friction of learning—such as struggle, error, and effort—that fosters understanding and growth.
This is especially problematic for students from disadvantaged backgrounds who face greater academic pressure and fewer resources. Supporters counter that AI can democratize access to information, support diverse learning styles, and improve educational equity when used responsibly. They emphasize the importance of AI literacy, teacher-led pedagogical design, and student agency in learning.
The discussion reveals that AI doesn’t inherently harm or help learning, but outcomes depend on how it’s implemented—such as through structured classroom practices, oral exams, or human-centered teaching. Ultimately, both sides agree that education must evolve to maintain intellectual rigor, foster critical thinking, and preserve the human elements of learning, including mentorship and meaningful interaction. The conversation underscores the need for institutional reform, not technological solutions, to ensure that AI serves as a tool for empowerment rather than a shortcut to academic success.
FAQs
The debate centers on whether AI enhances learning through personalized support and accessibility, or stunts it by promoting passive consumption and reducing critical thinking.
Passive use of AI can lead to cognitive surrender, where students rely on prompts instead of engaging in deep thinking, potentially weakening critical and problem-solving skills.
Yes, AI can provide alternative formats of information and support, making learning more accessible for students with physical or cognitive disabilities.
Faculty are responsible for designing learning experiences that guide students to use AI appropriately, fostering critical thinking and avoiding overreliance on the technology.
Yes, studies and experiments show that when AI is integrated thoughtfully into learning environments, students can learn faster and perform better than in traditional settings.
AI reduces the friction of learning by making tasks easier, which may remove the discomfort and struggle that help students develop deep understanding and problem-solving skills.
Chat with AI
Loading...
Pro features
Go deeper with this episode
Unlock creator-grade tools that turn any transcript into show notes and subtitle files.