00:00:00
You know, I'm just worried about
the future. I think we all focus on this point where this
singularity is going to happen,
00:00:10
when AI will really outsmart us.
And everybody thinks that this point will come from AI getting
better and better. But there is
00:00:17
an alternative story here of
humans becoming more and more relying— reliant on AI. And just
like we now have right now, an
00:00:25
air conditioner that can set our
temperature easily, and we can move from one place to another
without using any physical
00:00:33
activity, just like many of us
have lost something because of this maybe cultural or
technological evolution, we may
00:00:46
lose, as a species, something very
critical to our existence, which is our capacity to think.
00:00:53
Welcome to <i>The Ripple Effect</i>,
the podcast that takes you on a journey
00:00:58
through the minds of Wharton faculty. I'm your host, Dan Loney.
And in each episode,
00:01:02
we'll be diving deep into the inspiration
behind the groundbreaking research
00:01:06
that Wharton professors have conducted, and exploring how their findings
resonate with the world today.
00:01:13
We are certainly seeing how much
artificial intelligence is impacting our lives with the
advent of the technology and how
00:01:20
it is changing things like our
work. But how does it change our decision process? That's a
question that is addressed in
00:01:27
recent research, and pleasure to
have the authors of that research joining us here today.
Gideon Nave, who's an Associate
00:01:33
Professor of Marketing here at
the Wharton School, and Steven Shaw, who is a postdoctoral
researcher here at Wharton.
00:01:40
Gentlemen, great to have you
both with us. Thanks for your time today. Thanks for having us.
00:01:45
Let me start out, I guess, what was the genesis of wanting
to do this specific type of
00:01:53
research? Steven, I'll start
with you. Looking at the decision process. I mean, we see about it
from a business perspective, and
00:01:59
you know, the day to day impact
there, but the decision process is probably just as important.
00:02:05
Yeah. I mean, I
think, you know, we can observe things in nature. And just
observing how integrated AI has
00:02:13
become in our daily lives, we
sort of felt that the ability to actually outsource
thinking hadn't really been
00:02:23
studied itself. So it's
sort of a profound idea. A bit provocative, I would say, in the
paper, that— you know, with these
00:02:34
AI tools that are available,
they're so ingrained in our daily lives and decision
processes that we now have the
00:02:41
option or ability to actually
outsource thinking itself. Gideon, what was it that drew
your attention to this side of
00:02:49
artificial intelligence? You know, I've been studying
decision processes for a while,
00:02:54
and we know that there are
certain theories that account for how people make decisions
and how they use all the
00:03:03
sorts of types of decision
processes, from intuition to more deliberate analytical
thinking. And I think that, you
00:03:11
know, technology in the past was
integrated as a kind of a block that you can maybe offload some
of your cognition into, perform
00:03:21
certain tasks, like using a
calculator or using a GPS to navigate. But to me, it sounds
like the current theories of how
00:03:30
humans make judgments and
decisions must be going through some update once we have these
devices that really, as Steve
00:03:38
said, can really replace thinking
itself. And it's very important to start trying to account for
this theoretically and define
00:03:47
all sorts of tendencies and
constructs so we can really study it in the future. And Steven, I guess this is kind
of one of the next steps. When
00:03:57
you think about how artificial
intelligence is truly impacting our lives, we probably think
about people's reasoning, and
00:04:05
have thought about it off and on
for many decades. But obviously not in the perspective of how
artificial intelligence is going
00:04:12
to impact that process. Yeah, yeah. And that's, again,
one of the arguments that we
00:04:18
make in the paper is really, we
have these classic models. For example, the dual process model
that Giddy has talked about,
00:04:25
intuitions and deliberations.
But the reliance that we see, and the ability and the integration
of AI that we see in society now
00:04:35
just really changes the options
that we have for making decisions. And it does so so
dramatically that we need to
00:04:43
update our— many of our old
models in cognitive— cognitive science and psychology
and marketing.
00:04:52
Right. And so Gideon, it's no longer
kind of a dual process. It's a three-tiered process that we
probably have to bring into
00:05:00
play here, correct? Yeah, we added kind of
artificial cognition as a third model. And I think what's
interesting is not only
00:05:09
that we have this system and it
adds to it, but the existence of this system, just its mere
existence, changes how we engage
00:05:19
our other thinking modes. It
changes how we use intuition, it changes how we use deliberation,
and it also, surprisingly, also
00:05:28
changes how confident we are in
our responses, even ones that are not really critically
examined by ourselves. So there
00:05:38
are a lot of interesting effects
going on here. And some of them are maybe a bit dark, going to
the future, if we think that
00:05:47
these really are going to affect
the tendencies in behavior of humans nowadays. Steven, does— does having a dual
process now become— adding this
00:05:57
third component— does the dual
process run into any roadblocks along the way, as AI is becoming
a greater component of our lives?
00:06:07
Sure, yeah. And that's, you
know, one of the big things we show experimentally in the
paper. We call it cognitive
00:06:13
surrender. Basically, you know,
with access— just simply having the option of having AI
available for decision making—
00:06:23
people can surrender their
thought to AI and let it sort of think for them. So they're
basically subverting the whole
00:06:32
internal brain set of processes,
subverting and substituting system one, system two, and
instead, adopting answers from
00:06:44
what we call system three. And
as Giddy said, we saw— even, you know, when cognitive surrender
is engaged, people adopt those
00:06:54
answers and are more confident
in those answers. Do they rely on them more often,
then?
00:07:01
We saw— so in our experiments, we
saw— it was optional. You know, we just put ChatGPT in a
window. And they were doing some
00:07:09
sort of logic and reasoning
questions. We said, you can use AI if you want to, but you don't
have to. And so we saw over 50%
00:07:17
of the time, they consulted ChatGPT.
And once they consulted ChatGPT, adoption rates were
very high, even when AI was
00:07:27
incorrect, gave them the
incorrect answer, which we experimentally manipulated,
people adopted the answer over
00:07:34
80% of the time. Giddy, one of the things that I
know I have talked with many people about is the business
side of how bringing AI into the
00:07:45
business world is going to
impact so many different processes on a day-to-day basis
in companies. One of the things
00:07:51
that is brought up is how it
impacts labor and how people kind of interact with artificial
intelligence. It seems like part
00:08:00
of this research also can kind
of lead us down the path to better understand what that
relationship is going to be like
00:08:07
with companies and their
employees, and what's going to be expected from them as we move
forward because of the AI
00:08:14
component and the decision
process being impacted. Yeah. I mean, let's say it this way. If
we are completely surrendering
00:08:21
our thinking to AI, what value
do we bring to a company? It's not clear. So even this tendency
to surrender, maybe, something
00:08:30
that in the future, companies
would want to consider before they're hiring somebody. I don't
want to have some person that
00:08:37
just basically is giving me
what the AI can already give me. I can get this by myself.
And that also tells us, you
00:08:46
know, if we think of the
education of the next business leaders, obviously this is a
skill that we want to make sure
00:08:54
that people have and not lose,
the capacity to think critically, the capacity to be
able to check what the AI is
00:09:01
giving you. It's become more and
more important over time. So this is kind of a muscle— muscle
that we have, that hopefully we
00:09:11
are not going to lose over time. How much of a challenge, Steven,
potentially, is that? Because I
00:09:18
think at times we do feel like
AI, is this— is this beautiful benefit that we now have in our
lives. But we still need to have
00:09:28
that critical thinking as we
move forward on so many things, not only in our business lives,
but in our personal lives as well.
00:09:35
Yeah. And that's the key
question, right? I mean, how do we maintain critical thinking
skills in the age of AI? And you
00:09:44
said, you know, how fundamental
is that? Well, I mean, I think we're only at the beginning of
the age of AI, right? And this
00:09:51
technological integration is
just getting started. Right now, we are constrained by
communicating with the LLMs
00:10:00
through our phones or our
computers. And as those barriers reduce, right, that integration
is just going to become,
00:10:08
you know, stronger. What did you take away from
doing this research, Steven, that really resonated with you?
Either— maybe even something that
00:10:19
you didn't even expect you were
going to see play out? How readily people were willing
to cognitively surrender.That
00:10:26
was pretty shocking. And how
well the experiments worked, you know, based on our theoretical
contribution of this tri-system
00:10:36
theory of cognition. Gideon, what about you? You know, I'm just worried about the
future. I think we all focus on this point
00:10:46
where this singularity is going
to happen, when AI will really outsmart us. And everybody
thinks that this point will come
00:10:53
from AI getting better and
better. But there is an alternative story here of humans
becoming more and more relying—
00:11:01
reliant on AI. And just like we
now have, right now, an air conditioner that can set our
temperature easily, and we can
00:11:08
move from one place to another
without using any physical activity, just like many of us
have lost something because of
00:11:19
this, maybe, cultural or
technological evolution we may lose, as a species, something very
critical to our existence, which
00:11:31
is our capacity to think. But going back to something you
said a little while ago, then
00:11:36
the learning process for people
in general is going to change, because we have this component,
and it is going to be a part of
00:11:46
almost everything we do moving
forward. We may or we may not. You know,
typically, we know that the
00:11:53
technology— technological
development moves much faster than any policy change. It's
difficult for policy and
00:12:03
educational systems to respond
very quickly to what is happening here. I don't know
what going to be the
00:12:10
answer. And of course, there are
a lot of different perspectives and different disagreements.
What we suggested is provocative
00:12:18
to some degree. It's not obvious
that everybody would favor it. Even, you know, you can say, "I
want freedom. Let the market
00:12:26
solve itself." And with freedom,
sometimes you get outcomes that are, at the end of the day,
determined by power imbalances.
00:12:37
I don't know where we are
headed here. We are, as academics— are here to maybe try
to blow the whistle on it. We
00:12:45
can't do much more than that. Steve, having done this research
now, is there a next logical
00:12:52
step that maybe you would like
to take as you maybe delve deeper into this further?
00:12:58
Sure. Yeah. I mean, so in, we—
one of the part of— the part of the title of the paper is "the rise
of cognitive surrender," right?
00:13:06
So we now know that this
phenomenon exists. And the question is, when is it
adaptive? When is it good to
00:13:13
outsource your thought to AI,
right? AI gives us access to super intelligence. There are
many instances where, you know,
00:13:20
turning off thought can be a
good thing. But in a lot of high stakes contexts— education,
healthcare, right? We don't want
00:13:29
that to be happening. And so how
do we fight the rise of cognitive surrender in those
contexts? Is it on the user
00:13:36
side, on the human side, through
AI literacy and training, or is it on the UX design side by
putting in different types of
00:13:45
prompting or roadblocks that
make sure or try to induce critical thinking, you know, as
those decisions are being made?
00:13:53
Well, I guess for either one
of you, that brings up— not that this is specifically in
your research, but Gideon, it
00:14:00
brings up kind of the regulatory
side of this, of just how much of a role regulatory may have in
a lot of these developments as
00:14:09
we move forward. We're going to see, at the end of
the day, how the software
00:14:13
companies or the AI companies
will respond to this, how policymakers will respond to
this, how educational
00:14:19
institutions will respond to
this. I think what we have, but that is quite nice in this
paper, is that we have a very
00:14:25
clear method of measuring
cognitive surrender. And now we can try to take into the lab all
sorts of interventions and see
00:14:36
how we can maybe move it around.
And that's going to be quite a useful research tool.
00:14:45
Gentlemen, I appreciate the
time. Good luck with the research as you move forward.
All the best.
00:14:50
Thanks for having us.
- Thanks, Dan. Thank you. Gideon Nave,
Associate Professor
00:14:54
of Marketing here at the Wharton
School, and Stephen Shaw, postdoctoral researcher here at
Wharton.
00:15:00
Thank you for listening to <i>The Ripple Effect</i>. We hope
you found this episode
00:15:03
informative and engaging. Don't
forget to subscribe and leave us a review so that we can continue
to bring you the best insight
00:15:10
from the Wharton School.