00:00:00
Dan Loney: Well as our kids get
ready to head back to school, we need to look at the role that
artificial intelligence will be
00:00:07
playing in that learning
process. You are seeing quite a few school districts looking to
figure out how they are going to
00:00:15
incorporate it into their
processes this year and beyond. Pleasure to be joined by Hamsa
Bastani, who is an Associate
00:00:22
Professor of Operations,
Information and Decisions here at the Wharton School. Hamsa,
great to talk to you. How are
00:00:27
you today? Great. Thank you so
much for having me, Dan. You know, I mean, we have AI
conversations so much about so
00:00:34
many different things. This is
not a surprise that we're having one about education. But I guess
the question is just how it will
00:00:42
be included— not necessarily in
the college process. We're seeing that already. But in high
schools, in grade schools, what
00:00:49
are you starting to see and
hear from from some school districts? Yeah, I actually just got off a
conversation with some of
00:00:57
the teachers in in a high
school, and they were talking about how this year they're
going to start using Google
00:01:03
Suite in the class. So they'll
have Chromebooks for every student, and they're going to
have access to Gemini and other
00:01:10
tools like NotebookLM, which is
unfettered access to AI. One of the things I've talked about in
my research is that, you know,
00:01:17
giving students unfettered
access to AI can harm their learning, because they actually
tend to over-rely on those
00:01:22
tools. The number one question
they always ask is, what's the answer? So I'm asking them, you
know, why are you choosing
00:01:29
that instead of, you know,
something like Khan Academy or one of these ed tech tools that
maybe has better built
00:01:35
guardrails to support learning,
or a Socratic learning method? And they said that ultimately, you
know, our kids are going to go
00:01:41
into the workplace and use these
tools with unfettered access, and so that's what we're trying
to prepare them for.
00:01:47
But are the kids at this point
ready to have unfettered access? Like, have they— have they
learned enough about what AI is
00:01:55
and how it can impact
them, that we could just go ahead and let them, you know, kind of
get going with this?
00:02:02
I don't think so. So that's what
I worry about. So as a researcher, it's kind of an
exciting time. But as a parent
00:02:07
and an educator, it's not a great
time. I think it's very concerning. I mean, I don't
think anyone knows the answer,
00:02:13
but what we've seen from
at least our research studies is that students need
to build the fundamental skills
00:02:19
first and then they can
effectively use AI tools. And even then, you might have some
worries about de-skilling but
00:02:25
if you don't have the basic
fundamental skills down, then it's very difficult to even
assess the outputs that AI is
00:02:32
producing. And then it's also a
huge cognitive load for them, because they're trying to
process all this new
00:02:36
information, so they can't
actually check the outputs and collaborate effectively with
these tools. So I worry about
00:02:42
that. The other thing I worry
about is just taking away their sense of accomplishment, right?
One thing that we see is— and
00:02:48
this is so important for kids,
you know, that they have to feel that they have something to
contribute, that they did
00:02:53
something that they can be proud
of. And when you're just prompting an AI tool and
reporting back those results,
00:03:00
you don't have that. And that's
one of the reasons we see diminished motivation and
engagement.
00:03:05
Is there also an element of this,
that if they are going to do unfettered access, that they
feel this is the easiest way to
00:03:13
implement it. I'm speaking about
school districts. - Yeah.
- Without necessarily having to
00:03:18
make the investment on the teaching side, on the learning side— because we
know, with a lot of school
00:03:23
districts right now, budgets are
tight already, and it's hard to add new things to what they
already have in place.
00:03:31
Yeah, I think it's very much the
path of least resistance, right? It takes a lot of work to go
figure out, what are the right
00:03:39
guardrails, what are the right
pedagogical innovations that you can use to leverage these tools?
And, you know, like Ethan
00:03:45
Mollick and I've also been
talking about like, it's actually difficult that that's
an open research question. It
00:03:50
takes a lot of exploration,
creativity and work, and school teachers are not paid enough to
do that work. They have no
00:03:58
dedicated time or training to do
that. Administrators, like you said, they have, you know,
limited budgets. Just giving
00:04:04
everybody access to, you know,
one of these Google Gemini Suite or OpenAI subscriptions is kind
of a cheap investment, right?
00:04:13
And so it's kind of easier and
it's less likely to be criticized. So I think it's kind
of just going down the path of
00:04:19
least resistance. And you see
teachers reacting the same way. They're very much going back to—
a bunch of them are like, "Can we
00:04:24
find good plagiarism detectors?
Can we just have students write in their blue notebooks?" Also
things that are not going to
00:04:30
work. And I think, you know,
without creating the time and space to explore and understand
this better, this is what you can get.
00:04:37
But something you said a little
bit ago, really, let me circle back to it, is— there is
something about when a kid, a
00:04:44
child, comes up with an idea by
themselves, on their own. It becomes very successful in the
classroom. That that's the
00:04:53
similar formula that they need
to implement when they're out in the workplace as well. We can't
just rely on AI in the
00:05:01
workplace, because it's — for me,
I mean, it's like acting like a crutch. It's— it's— it will help
you along, but it won't serve
00:05:10
the larger good. I 100% agree
with you. I mean, also, if you could rely on the AI, then why
do we need humans? We're going
00:05:17
to end up firing our entire
workforce. So we need to train people to add value above and
beyond these AI outputs. And one
00:05:26
of the things that really
bothered me was, my colleagues and I ran a study on
chess, right? We hear AI has been
00:05:32
much better than students for a
long time, but people still play chess for the love of the game.
And one of the things we found
00:05:37
is that if we gave students— we
followed them for about three months. We gave them different
AI tutoring platforms. If we
00:05:43
gave them the ability to click a
button that lets you figure out the answer, or give you the next
move without having to think
00:05:49
through it, they kind of started
getting addicted to using that button. And they found chess
less fun, because when you
00:05:55
figure out a really complicated
move that, you know, gets you a piece or wins you the game,
that's when you feel great, even
00:06:01
though it's a lot of work and a
lot of struggle to get there. And so by illuminating this
struggle, it's very difficult to
00:06:07
say no to that, because we don't
like the struggle. We're also removing the sense of
accomplishment.
00:06:12
So then is the expectation— and
I'll say this just from my perspective of just talking with
you here for a few minutes— that
00:06:19
while schools will implement AI,
that this will still be a learning process for the
schools, to some degree, to
00:06:26
understand its uses, its
implementation, how it will impact students, you know, as we
move forward? It may not be, you
00:06:35
know, totally 100% just put it
in and don't worry about it ever again. That there will have to
be some watch and some
00:06:43
some research on just how it
kind of focuses in to what students deal with in schools?
00:06:49
My perspective is probably a
little bit more pessimistic than that. I guess— okay, so
historically, we've had policies
00:06:56
that we've invested lots of
money in, like one laptop per child, or the LA iPad, where we
just threw money and technology
00:07:02
at kids and hope they would
learn better. And they have not worked, and sometimes they even
have negative effects. The other
00:07:08
especially pernicious thing, I
think, about generative AI, is that now kids are not seeking
their teachers for help, so
00:07:14
we're losing the human
connection. So we don't even know what's going on with our
students. They don't come to
00:07:19
office hours anymore. They don't
ask for questions. They don't— parents tell me that, you know,
their kids don't come and ask
00:07:24
them for help with their
math homework anymore, because they can just ask ChatGPT. So
how would we even know what's
00:07:29
going on with our kids? No one's—
no one's talking to them, and that combination, I really
worry about.
00:07:36
Hamsa, great to
talk with you today. Thanks very much for your time. Thank you, Dan.
00:07:41
Thank you. Hamza Bastani, who's an Associate Professor of
Operations, Information and
00:07:45
Decisions here at the Wharton
School.