You are in a new environment. I'm not. So we decided to talk about advice.
Yeah. That's a nice big topic we can discuss. And I think there's a lot of
interesting things about it because There are things that are contrary to our
intuitions about advice, but I started out here by asking chat GPT about
advice and the answers it gives, I think are sort of the first order things
you, everyone in might think about advice. So it's a good place maybe to start
to correct. Uh, I guess that's often a good use for chat GPT or other large
language models. They'll give you sort of the usual point of view that you can
start from. Okay. So here we go. People who tend to give advice, experts,
confident, helpers, and the extrovert. Okay. Okay. People who tend to ask for
advice, those who are inexperienced, open-minded, in crisis, or non-assertive.
Okay. It doesn't have a third category for people who don't believe in advice
I didn't ask about that. I asked who tends to follow advice as opposed to who
tends to ask for it. Right.
Okay, so who follows it?
Agreeable, conscientious, I trust, and those who sought it out. Those who seek
out the advice by following it. Makes sense. So these are all plausible usual
stories about advice. That is, you know, you ask for advice when you need it,
you listen when you need it, you ask people who know more, you ask when you
know less. So this is, in some sense, what we'd have to overcome to have a
deeper theory. Like, is there anything wrong with these? Does this really
summarize the asking, giving advice, or do we know something more about advice
such that this doesn't quite describe it?
I am now, I'm now just blanking on this woman's name, but a professor in the
business school, the Booth School at UChicago, we did an event together on
advice and she does, you know, sort of psychology research. And one, one, bit
of research that she did on advice was looking at advice givers and advice
takers in a like high school setting where some students advised others about
like how to get better grades or and they found that the students who received
this advice did not improve in their school performance, but the givers of the
advice improved. So
And so this was a causal result of the actual event of giving advice, not just
a correlation with the kind of people who tend to do it?
I don't think we know that. Okay. However, as I recall, it's an improvement.
Right. So possibly the people who are just better at living maybe feel freer
to give advice.
Maybe also just giving advice gives you confidence, makes you feel like you're
the sort of person who's going to succeed.
Or it expresses the confidence, you know, the confidence is there for whatever
reason, but this is showing the confidence.
But expressing your confidence can give you more confidence. Sure. So the
point is that the event of giving the advice could, for confident people, make
them more confident and even more successful.
Right. Although, I mean, we tend to want to distinguish, you know, confidence
is often a good sign because in fact, you have good reasons for your
confidence. If you just get more confidence without actually having good
reasons for it, then if that works, it's a little more puzzling. That is, we
might expect people to choose their levels of confidence to match their actual
levels of confidence or ability or potential or something. And then it makes
sense that if people are more confident, then in fact, that's revealing this
underlying better potential ability and they look, end up doing better. It's a
little more puzzling perhaps if just forcing yourself to be more confident
when you don't actually have good reasons for it would also make you do
better, but that's possible.
Right. I mean, I guess part of it is a question of if usually the teachers
will notice the people who are more confident and will think, well, more
confidence tends to go with being better and that's giving them better grades
and stuff, then maybe you can get an edge by just getting a little bit more
So your result that the people who listen to advice more weren't actually
benefiting, that's somewhat interesting and provocative. I guess there's an
experiment I want to mention here, a somewhat famous economist experiment from
roughly 10 years ago, where they just set up a website and said, come to our
website if you have a big decision, and we will randomly flip a coin and tell
you whether to make this change you're considering making or not. I'm not sure
they told them it would be random, but it was random. So they gave them advice
about their life situation and they said, make the change or not make the
change. This was a nice randomization so that they could see, well, was
changing a good idea? And then after a while they asked them, well, did you
make the change and were you better off? Apparently, on average, people who
were told to make the change did make the change more and were on average
happier later. They more, you know, approved of what they did. So this
suggests that maybe we're all a little too reluctant to make changes because
if you just randomly induce us to make a change on average, we like it.
But like there's also other explanations, right? Which is that if like making
a change is more of a pronounced decision than keeping things the same and
we're likely to look back on that in favorable ways. We're likely to think,
well, we made a decision that it was justified. And so it could just be that
effect. Like if you just took people and in all situations or whatever, who
made changes, you might just find that they tend to be happy with the changes
that they make.
Right. But this is at least some context. So I think our situation is we have
this very simple model of advice that people get advice when they need it and
people who give advice have it and it's useful. And in the context of that,
your advice is not to give or take advice.
No, I don't have advice, because that would be a contradiction, wouldn't it,
if I was advising people not to take advice?
Okay. Your stance is to not, your personal stance, that is, not to give or
I mean, for example, it might well be really good to give advice, because
you'll do better yourself, according to this experiment. So go give lots of
people advice, regardless of what the content of it is, and then you'll do
better. It's not obvious to me that you shouldn't give advice. It's that most
people don't tend to know what other people should do, I think. In giving
advice, they don't give the advice from a place of knowledge, and a lot of
advice is is oriented around, like in my experience, is oriented around
presenting oneself as the sort of person from whom one should take advice
from, right? So being a, you want the content of your advice to sound like it
comes from a respectable advice giver. And that does a lot to determine what
kinds of advice people give. And so I'm suspicious of the actual value of the
actual informational value contained in the advice.
So if we set a really low bar for what counts as advice, then if you walk into
an unknown building and ask someone for where the bathroom is, and they point
you somewhere in some sense to giving you advice, But it's pretty trivial. So
there's presumably those kinds of very simple situations where you ask where
the exit is or where the copy machine is or the coffee shop. Um, or, you know,
maybe you you're learning to ski and you're trying to figure out how to put on
the ski and some more experienced person, like could show you how to put on
the ski and you don't know how to do that. It seems like there's a whole range
of pretty straightforward situations where asking and getting advice isn't
Right. So that's why I think it's not a good way to carve up the advice
territory. So in an essay that I wrote about advice, I made a tripartite
distinction between advice and then two other things, instruction and like
coaching or mentoring. So the stuff you're describing, I would call that
instruction. So when somebody is giving you basically something like a
quasi-algorithmic procedure that is supposed to lead to a result that's really
well specified and can be well shared between the two of you, like here's how
you walk to the store, here's how you use this machine, et cetera, that's
instruction. And so I don't think that's advice. And I think people can give
each other instruction, no problem. All you have to do in order to give
someone else instruction is actually know how to do the thing that's
well-defined that you both want to do. And that people often have that kind of
knowledge. And I also think people rarely pretend to it when they don't have
it. I mean, I guess every once in a while, a malicious person tells you how to
get to the store when they don't know how to get to the store and they give
you false instructions. But basically, people are not going to tend to be able
to put themselves forward as being able to instruct someone on how to use the
copy machine if they actually don't know how to use the copy machine. And then
at the other end of the spectrum is coaching. And coaching might, unlike
instruction, it might be for a more abstractly specified goal, like being
better at tennis or something. And there may be no like, you know, even quasi
algorithmic procedure for getting better attendance might be like a lot of
different little things that you have to do. What distinguishes it from advice
is that there is a kind of close relationship between the coach and the
coachee and the coach has a kind of like intimate knowledge of the person that
they're coaching and thus the ability to continuously redirect their coaching
in the direction to improve the person. So what I take to be characteristic of
advice is that it is the goals are a little bit like the goals of coaching.
But the kind of contact that exists between the advice giver and taker is more
like what you typically find in instruction. So it's an attempt to combine
those two things. And that is, it doesn't rely on a continuous close personal
contact. And I think that's basically a mistake.
So there's a lot of long distance coaching or I don't know, ours is like
coaching, I guess you're disapproving. So a travel guide, an instruction
I think an instruction manual is instruction. A travel guide, I mean, I guess
I think like a paper, a book travel guide. So I think a lot of what's
contained in a travel guide is instructions. So like how to get from point A
to point B or how to navigate this thing or whatever. I actually think most of
it is going to be instruction. And then other things in it are just going to
be suggestions. Like here are some five things you might do. But I guess I
wouldn't a travel guide, I guess, is a combination of... How is a suggestion
not advice? Because they're not telling you to take it. Like if I say like, if
I say like, here are five things to do in Paris, like the guide tells you,
here are five things to do in Paris. I mean, maybe it's in the advice family,
Say I review a movie, right? And then I give it a higher than usual rating.
That would presumably tip you toward being more willing to see that movie, all
else equal. Right. That doesn't count as advice. It's certainly not
instruction. It's not a formula.
I agree. So my tripartite distinction didn't carve off advice from every other
possible category. I was just distinguishing it from instruction, because you
brought up examples of instruction, and then I think coaching is kind of the
other extreme. So I think that positive evaluations don't generally count as
advice. That is, we want to draw some kind of distinction between a positive
evaluation and advice. So if I write a movie review and I say, Barbie movie
was awesome. I don't think that's the same thing as telling someone, I think
you should see the Barbie movie. They're not like sort of so different that
they don't belong on a spectrum or something. But I guess I'm thinking of
advice as more like involving at least something like a request for advice and
then a fulfillment advice.
So you said the word positive there. Does that mean a negative review is
closer to advice?
No, I think neither of them is what I was thinking of as, let's say,
paradigmatic advice. Like, for example, suppose you made a really negative
review of a movie. And suppose somebody was like, yeah, I think that's
probably true, but I think I would enjoy all those negative features. And they
went and saw the movie. I don't think you would say to them, hey, you didn't
take my advice. And the reason you wouldn't say that is because you didn't
give the advice. You just said it was a bad movie.
I recently attended my colleague Brian Kaplan's annual gaming event, Kaplakon.
So I was recently in the mode of giving and receiving advice about playing
board games. That is, comments about, in particular kinds of situations, the
kind of strategies you might want to take or might want to avoid.
Oh, wait. So this is not about a choice between game. This is a choice of
strategies within a game, right?
For example. Yes. And in addition, of course, people recommend games and, and
I took their advice, games, but I'm struggling here to, to understand this
category of things that you don't want to do. Uh, cause I see it continue with
all sorts of things. So it seems like, you know, there can be a very specific
advice in a very specific situation, you know, don't, don't land on this
space. or generally in the early part of the game, let's focus on this element
of the game rather than that part of the game, or watch out for a move where
someone might do this, which would put another thing at risk. All of those
seem like advice to me, I guess.
To me, they seem more like instruction, because the person could probably
specify the strategy that they were following and the concrete result that
they were aiming at, like getting more of your pieces over here or something.
So like advice, I guess, as I'm seeing it is something like someone asked me,
what should I do with my life? And and there, you know, the the direction of
your life or some broader kind of goal. are gonna be in some way shaped by it.
I think the example that I was thinking of when I wrote the piece are actually
famous people like Margaret Atwood being asked for advice for young writers,
like how are they gonna become great writers? And she says an amazing thing
when she's asked, she's like, well, I guess you should write every day. And
then she comments, she's like, well, I didn't do that, but it's probably a
good advice. She doesn't write every day.
Are we distinguishing advice on sort of ultimate goals versus conditional
advice given a goal? Because Atwood's case was if you want to be a writer.
So that sounds like a conditional advice. No, I think it's about the nature of
the goal. So I think that a goal like how do I get to the supermarket or how
do I get more of my pieces to the other side of the board or whatever. that
goal, it can be well specified. And that specification can be very well shared
between the two people. And there can be roughly algorithms for achieving that
goal. Go here, then go here, then go here. And so when all of those three
conditions are met, then I feel pretty convinced that it's possible for the
one person to transfer knowledge to the other person. But if the goal is
something vague, like how to be a good writer, if it's very unlikely that the
person being asked and the person giving share any clear conception of what a
good writer is, if there is no algorithm at all, not anything even wildly
close to an algorithm for how to achieve it, then we meet the conditions where
I become skeptical.
If I said, how do I become a well-published writer? Now it's a very clear
shared goal. And so now you would accept that as feasible advice.
Even that is not that clear. By well-published, do you mean I have a large
number of publications? Do you mean I publish things in a certain fashion? You
could easily specify those things. Sure. To the extent those were the issues.
So suppose someone said, how do I publish- How do I get into the New Yorker?
I want to get into the New Yorker.
Yeah. Good. Right. I think if you were to ask Margaret that with that
question, all of a sudden she would be like, I'm not sure what to say. because
she would realize that she didn't know the answer. Whereas if you keep the end
vague and be like, how do I become a good writer? Then she'll offer you
something like write every day. She won't say write every day if you ask, how
do I get into the New Yorker? I would predict.
This is sounding a bit like saying that when somebody asks you a very clear
question that they could check on, you're going to be careful with your
answer. But if they offer you a vague question in which you think they could
never check, then you feel more free to just make up stuff that sounds good
because nobody's ever going to check.
Basically, yes. That is, I think that, but we're forgetting the other extreme,
right? Because I actually do think people can usefully guide you with respect,
even to vague goals, like becoming a better writer, if they know you. So if,
like, you know, my students are often coming to me and saying, how can I
improve my writing? And I'll be like, well, here are the issues that I see you
as having over time. Here's how you developed. Here's something you can do.
And I have a kind of concrete knowledge of where their virtues and flaws are,
and I'm able to give them what I think of as good mentoring or coaching. But I
think if a total stranger comes to me and is like, how can I improve my
writing? My answer ought to be, I have no idea, rather than something like,
write every day. Don't be afraid to throw things away. Whatever. There's
generic advice that people give.
Okay. So it sounds like we're describing two reasons that make advice giving
difficult. Uh, and then noticing that if you face those difficulties, you
should probably realize your task is difficult. And then maybe your first
response would be to think it can't be done. So one reason that a task is
difficult is because the goal has been specified vaguely. So there's a lot of
potential goals, you know, that could fit the description of the goal. And
another reason, an obstacle to giving advice is that you might not know enough
about the context, including about this particular person and their styles.
Those are not independent, right? Sure. Not knowing enough about the context
and this person and their style is only a problem when the goal is fake. If
the goal is, how do I get to the end of the street? I don't really need to
know much about you in order to direct you to the end of the street, right? Or
if the goal is, how do I move my pieces to the other side of the board in this
board game? I don't need to know about you.
Well, I mean, if you were asking me about advice for shoes and you're just
another person on the internet I know nothing about, then I would face an
obstacle about getting you a new pair of shoes, even though the idea of new
shoes is pretty concrete.
It isn't that concrete. There's all kinds of different shoes, right? And you
don't know which kind I'm looking for.
Right. But that, that's the point, but I could know, I could know a bunch, I
could know your goal, but still not know other things about you. And then that
would be an obstacle to giving you advice about the shoes. I can say, I want a
comfortable shoes and I say, okay, but I need to know more about your feet and
where you're walking and things like that.
That's a good point because I actually see the role of the coach as somewhat
different from providing all that information. Say I wanted to get new shoes
and you're like a Mr. Shoe Expert online, what you could maybe do is ask me a
bunch of questions like, okay, what size feet do you have? What problems do
you have? Et cetera. And you ask me enough questions and then you're like,
okay, these are the shoes you should get. And you don't actually need to know
me. You don't need to have met me. You don't need to have interacted with me.
You don't need to have like a kind of extended experience with me, a
relationship with me. So I see coaching and mentoring as requiring a
relationship, not just knowledge of the person, but a connection to the
person. And so I think sometimes you can give sort of the instructional help.
the part of giving the instructional help, part of the content that you need
is certain information about the person. But that information doesn't have to
come to you through the channel if you're having a relationship with them. So
I would still see that as the instruction, the person who's telling you what
kind of shoes you should buy.
Well, if I'm going to summarize what I've heard from you as saying, the
process of giving advice can go wrong, Uh, and here are two factors that make
it harder. One is, you know, less about the person and another is, you know,
less about the context. Um, there are many other factors we can identify
presumably about. giving advice, that is just a problem that's harder. It'll
be harder to give advice about. There's probably just some kinds of situations
which are just easy in and of themselves. And therefore, you know, it's easier
to give advice because not too many things can go wrong. So it seems to, we
should just summarize and say, well, you're thinking of giving advice. But be
wary of these complications and difficulties. And if you see enough of them,
then you should be especially wary. But if you see very few of them, then be a
little more free. This is more of a continuum of, you know, be more or less
reluctant to give advice, depending on these factors. Isn't that a more
nuanced sort of advice to give about advice, which is be careful about things
that make advice giving hard.
I mean, that's the kind of advice that sounds like good advice. So that falls
into my category of how Robin would talk if he wanted people to trust him as
an advice giver. So it's a good paradigm. I don't think that, I don't like,
for instance, predict that people who receive that would do any better at
giving advice. So no, I don't think it, I don't trust it as good advice.
Here's how I would think about it. Instead of having a category advice, I
would say, look, people want to help people. That's why I give advice, right?
People want to be helpful. But people also want to believe that they can be
helpful even when they can't be helpful. And in particular, people love the
idea of helping other people at zero cost to themselves with almost no effort.
That's kind of what advice is, right? Like Margaret Atwood has a three-second
encounter with someone and she's like, look, I can improve your life in just
three seconds. And I'm here now being unfair to her because of course she's
put in this position. She probably would have preferred not to be. But the
point is, if you want to improve other people's lives, there are some ways of
doing it. You can give them instruction if you happen to know how to achieve
some goal, or you can actually have a relationship with someone and improve
their life, but it's high cost. You're going to have to put a lot of time and
effort into getting to know this person and working with them and coming to
learn how they work and how they operate. And so if what you were looking for
was the kind of helping of other people that would require nothing from me,
basically, my inclination is to think that's just mostly an illusion.
That sounds like a restating of what I said. Uh, that is, those are some of
the things that make advice giving difficulty and you just reviewed them
again. And so we can say, Hey, here are some obstacles to be aware of
regarding giving useful advice. So, um, yes, that, that seems to me another
restatement of the, here are some things to, to work, to watch out for, uh,
you know, In order to be valuable, you'll have to know enough about the
situation and that will require time and effort. And if you aren't willing to
put in the time of effort, then be wary about bothering because, you know, it
won't go so well.
People tend to really like giving advice. Not me, but most people. They, they
enjoy it. They positively enjoy it. Um, and, um, uh, when people get
unsolicited advice, they're often a bit annoyed by it. Um, I guess you would
explain that by the fact that like being a giver of advice is a positive
signal. Cause it means you're like confident expert, extroverted, et cetera.
Whereas being a recipient of the device is like a negative signal about you.
Is that how you explain that phenomenon?
Well, so we started with this very simple model of advice at the beginning
about, you know, you give advice when you have information, you get, you ask
for advice when you need information. And then I asked, you know, how should
we correct this simple model? And if you ask me how to correct the simple
model, my first correction will be in starting to do something like signaling.
I'd say, well, Similar to what you're saying, sometimes people want to seem a
certain way and giving advice will let them seem a way. And then they will
give advice in order to seem, give that appearance, even if the advice isn't
so useful for the other person. And that would be a reason to be wary of
certain kinds of advice. And to be wary of being the sort of person who would
fall into that habit. So yes, giving advice makes you look knowledgeable. Uh,
if somebody seems to follow your advice, that seems to make you look
influential, uh, you know, trusted, uh, respected. Um, and those are the
things people would like to seem, uh, it look even looks caring, apparently,
although as you say, many people, you know, on the receiving end are actually
liking it, but, um, people can present it in a way that they seem concerned
and caring for the other person. So these are all ways in which people are.
And on the other hand, we have to see that the recipient of the advice often
doesn't want to take advice for similar reasons. If they take advice, then
they are somehow accepting dominance or prestige higher of the other person
who's giving them advice. And so people often don't want to take advice
exactly because they don't want to create that unequal situation with the
advice giver. And so they might be. excessively inclined not to take the
advice for that reason, even if the other side is excessively inclined to give
the advice. So we can blame both sides in some sense via these sort of
signaling reasons. And of course, your needing advice may sort of suggest you
are less competent and less informed about, and less experienced about
something than you'd like to seem. That would also be reasons why you might
not want to be seen as taking advice
So has it ever happened that you received good advice, that is that you can
recall, where it has to fit my definition of advice, so it's not gonna be
instruction, it's not gonna be how to do some specifiable task, and it's not
gonna be coaching, it's not gonna be someone with whom you had a long
relationship who guided you through your use of the advice. Have you ever
received such advice that was good?
Well, so it's going to come down to these parameters, which I'm not sure how
to judge. So for example, early in my economics career here at Mason, my
colleague told me, if you spend too much time distracted from standard econ,
that'll be a problem for your getting tenure. You should not spend too much
time away from standard econ. They were somebody who knew me somewhat well,
but that advice wasn't really based on very specific knowledge about me. That
was pretty generic advice that you could give to us, to anyone.
Yeah, I think, I guess I would say that counts probably as instruction, but
it's a bit of a, it's, it's, it's a, it's not, it's not a clear case. So it's,
it's somewhere between maybe instruction. Okay. And you think that was good
Yes. For my purposes.
And you took it.
Right. Or I took it somewhat. Perhaps I squeaked by, but should have taken it
more, but it was a trade-off there, but yes.
Right. I mean, I guess I'm inclined to think that sounds a lot like if you
want to go to the supermarket, take this road. And it sounds different from if
you want to become a better tennis player or a better writer, you know, do
this. Where, like, for instance, your colleague wasn't saying that you should
do standard econ or that you should care that much about getting tenure. They
were just saying, like, if you want to achieve this result, take this
strategy. But like giving someone advice in a game.
I'm puzzled about tennis as your example, because it seems to me most people
who play tennis agree on the criteria of a better tennis player as someone who
wins their tennis game.
When I was a kid, like 10, 11, 12, my sister, she was good at tennis and she
took tennis lessons. I was not good at tennis. I did not take tennis lessons.
I routinely beat her in tennis. The way that I beat her was I played in a way
which was so antithetical to how her coaches were explaining how you should
play, like hitting the ball super high up in the air and all kinds of wads,
that it just made her so angry that she couldn't focus and she would lose to
me. So I think I could just, I could beat her, but I wasn't a better tennis
player than her. I think being a good tennis player, like, yes, it's gonna be
correlated with winning matches, but quite often the better player will lose
in a given match. And I think if you ask like a professional tennis player,
what is it to be a good tennis player? They will give you a long and kind of
articulate and refined answer that I cannot, as a non-tennis player, give you.
But that's the sort of stuff coaches know.
But nevertheless, I stick with my claim. People who give advice on being a
better tennis player almost always agree on the meaning of advice that will
help you win more tennis games. So according to that, it's all instruction
because they all agree on the end.
No, so there's two elements of it. So first of all, I don't actually think
that there is a shared conception of the end. Say you're an expert tennis
player and I'm a beginner. I, as a beginner, don't know what a good tennis
player is. For instance, it may be that you, as an expert, have an
understanding of how the good tennis player holds their back. I don't know
anything about that. I don't know that that's part of being a good tennis
player. So we don't share. The point is not, is there a clear conception? Is
it shared between the person who's getting and the person who's giving?
We only knew the conception of what the goal was, not how to achieve the goal.
So how to hold the racket is part of how to achieve the goal, not what the
goal is. The goal is just to win the game. So we all agree about the goal. We
just don't agree about how to achieve it.
I mean, okay. So I think this is an interesting, although somewhat tangential
question that in wanting to be a good tennis player, I guess I don't think
that's the equivalent of just wanting to win a bunch of games. For instance,
If I could bring it about that I won every tennis game I ever played, but I
did it by, you know, causing subtle irritations in my opponents in the way
that I did with my sister, that would be somebody who had the goal of wanting
to win their tennis games, but not by means of being a good player. Most
people who want to win a tennis, they want to win by being a good player. And
if you ask them, what does being a good player mean? Now there's like a long
and rich account that the coach can give you and I can't.
I think you're, you're taking, you know, you're really picking on a really
small difference here. And I can't believe that this is going to make the
difference here. There's going to be lots of small differences all the time.
When you advise someone about go to the grocery store and you tell them a
path, there's going to be all sorts of little details about them that you're
not taking into account. And you can say, oh, that's not really an instruction
because you couldn't take into account their limp or maybe the fact that
they're, you know, they get blinded by the light. If they go in a certain
direction, the lights in their eyes. I mean, there's always going to be all
these little details.
Well, the difference is whether the details matter. So like in tennis, for
instance, really small details about like how you hold the racket and like
whether you bend your knees and whatever, all those details might really
matter to being a good tennis player. Whereas, you know, what exact path you
take down the street, how many steps you take along the sidewalk, etc. doesn't
matter to getting to the supermarket. And so you can instruct someone on how
to get to the supermarket without speaking to those issues. If you happen to
know this person is disabled, for instance. and they can't go on any sidewalk
except the kind with a cutout, then you would be giving them poor
instructions. Unless you could give them instructions of how to use the
streets that have the cutout. But, you know, that's going to be all of the
details are sometimes relevant and sometimes not. And you can give them
instruction if you know all the ones that are relevant.
still seems to me what we mainly want to do here is identify the reasons why
this simple process might be going wrong for giving advice.
I think you're saying basically what there is is instruction and that works
and then we can just, advice is basically like instruction only it sometimes
goes wrong. And I just think no, instruction has its own weird funny case and
advice is really not very much like that. That is what's happening in advice
is really like more like one person trying to steer another person's life in a
substantive way. And that really what advice is, is sort of coaching or
mentoring that tries to squeeze itself into the box of instruction. And that's
why it just doesn't work.
Right. You can advise tennis players you don't know about how basics of
tennis. That's, you know, what you'd find in an instructional tennis video,
for example, for beginners.
Right. So I think that there is some instruction that's possible for tennis,
for example, suppose that all I want to do it like suppose that I just keep
hitting the ball into the net or something. I have no idea this actually
happens with tennis. Right. And then like someone can a professional tennis
coach kind of like the the shoe advisor, the case that we imagine, can be
like, ask me some questions and then be like, I know what you need to do. You
need to tilt your racket more at some angle or whatever. And then I could try
that, it could work. That would be a case where there's some piece of the good
tennis player thing we can carve off that's concrete enough and the road to
which is algorithmic enough that it's then susceptible to instruction. So I
don't want to deny you can instruct people about tenants. You just have to
narrow the focus.
So tomorrow night, I give my standard introductory graduate student lecture to
a graduate student class, and part of my standard introductory lecture is
about overall strategy of being a grad student in the world of academia. And
so in that context, I will be giving them general advice about the world of
grad students that isn't specific to them because I don't know them yet. It
isn't algorithmic at all. It's sort of general strategies. So are you telling
me this won't work?
Right. That's what I'm telling you. Yes.
Even though you don't know what they are yet.
Right. I'm skeptical. That is my initial opening. Okay, Kabir, you're some
kind of advice for me. You're the counterexample to everything I've ever
experienced in my life up to this point. That's totally possible. I'm eager to
hear what is this brilliant advice. But if you're like every other instance of
this phenomenon I've ever encountered before, yeah, it's a situation in which
both people make each other feel good by creating the illusion that one person
is helping the other, but actually neither knows each other well enough to
really be of any substance with the system. But tell me, so what's your
So one kind of thing is just to identify common failure modes. like common
ways in which smart, capable grad students would nevertheless fail the goal of
getting a PhD or getting an academic job afterwards. I can identify failure
modes that they don't know about and that I've seen, and then that can be a
warning to them about things to watch out for. That seems to me useful advice.
Okay, let me try a thought experiment. Say we could somehow do, this thought
experiment involves imagining an experiment. Say we could somehow, you know,
look at every student of yours, every graduate student who would ever receive
your advice about how we should conduct themselves in grad school, and compare
them to a control group of econ graduate students who had not received
comparable advice. And then, And then the question is, are your students gonna
do better? And now, here's my question to you. How much money would you bet on
the thought that this group of students that received your advice will just do
better in grad school than the students who didn't? Would you be willing to
bet money on that, and how much would you be willing to bet on that?
I'd bet that they would do somewhat better. Now, you'd need a large sample to
see a small degree in difference, but if we had a large enough sample, I'd bet
at least four to one.
Okay, so you only expect a small difference. So you don't expect your advice
to do very much.
A small difference is worth giving, given what's at stake. So I don't think
it's a waste of time.
Right. No, but so like, I mean, in a way, right, if you think about, say,
avoiding failure modes, right? So you give the students a bunch of different
failure, possible failures, and then you identify them so that they won't fail
in those ways. In a way, you might think that's incredibly useful. I mean,
they can avoid those failures. You're going to see massive differences because
all these other students are going to be failing and your students won't be
failing. But that's not actually what you predict, right? You don't predict
that the result of your telling them about these failure modes is going to be
this like massive difference in them not failing in ways that other people
That's what I said. Yeah. So I thought they are modest differences worth
imparting, but not enormous ones.
Well, no. So, so, so what I'm doing is, um, you can't, you take a look right
now. We're recording. Sorry. Okay. Hold on. Okay. Um, uh, so, um, um, the, um,
What I'm saying is, compare your prediction about how much of an effect this
advice would have with something like the content of the advice, right? So the
content of the advice is, here are some common failures and here's how to
avoid them. You might think that the students who hadn't been told about the
common failures, they never knew about these common failures, right? They
would just be failing all the time. And the students who had been told about
them, they'd be able to avoid these common failures. And so you might have
thought if advice worked the way one would think, then it would be a huge
difference. But even you don't expect that telling people about these things
is going to make much of a difference.
I just don't understand why you think anything that would be helpful has to be
hugely helpful. I think some things can be modestly helpful and still be
Right. I mean, like saying that it's only modestly helpful, it sort of like
lowers the burden and like, sure, it could be. But I mean, what I was trying
to draw your attention to is that the inefficacy, the way that the advice
doesn't work is already built into your view. You already sort of agree that
advice doesn't really work. You just think maybe it works a little bit enough
for me to get it.
I don't think it works overwhelmingly in a very simple way, but I think it
works in a complicated way that's helpful. So just to be clear, they could
have heard the advice from someone else, counterfactually. I might not be the
only possible source.
No, no, but I ruled that out in my setup, right?
Most of them will probably forget it and they won't remember it very well.
Are you there? Yes. Okay. Secondly, third, there's going to be a fair bit of a
matching problem for them to recognize the situations I'm describing in their
lives, and they may well fail to recognize the failure mode situations when
they're looking at their scenarios. So they might not actually see. And of
course, these are correlated things. These are things that are typically not
worth the risk, but sometimes worth the risk. So there's always a judgment how
much of an exceptional situation they're in in order to do something
exceptional. Every failure mode is basically a kind of a chance people take
that most of them shouldn't be taking, but some of them should.
So these are all reasons to understand why it could be complicated for them to
apply the advice and for, you know, and of course, many of the situations may
just never occur for them.
Yeah. It's also true that like giving based on the first day of class, it's a
good way to establish yourself as like an authority who's dominant, who's in
charge, who like knows more than they do. So that could be like an explanation
of why you're giving it too. But you want to predict, so according to your own
signaling theory, that's like part of what's going on here, right? Right. But
as long as there's a little bit of a system.
That can be a benefit. direct benefit itself. That is, they might want to be
reassured that their professor is knowledgeable and cares about them. In some
sense, I am signaling a concern about them, but I'm interested in them and
their careers. I'm not just wanting speeches up there.
Yeah, this is, for me, the really galling thing about advice is the deceptive
signal of care. That is, I think advice signals that I don't care enough about
you to spend more than a second spewing a few words at you. And I'll never
check up on whether this actually helped you. But I want to kind of make it
look like I care about you. And so I respect you little enough to think that
this will do. That's the signal it sends to me.
It's pretty unfair. That is, if this is the first day of class, I'm not
telling them I'll never talk to them about these things again. I'm inviting
them to think about the topic.
How often do you talk to them about these things again?
Often they come and meet with me and then they ask follow-up questions about
their options and career, and then we can discuss that. But having introduced
just the framing of these issues, in some sense is inviting them to continue
So I think if we view your speech as the opening gambit of a mentoring move,
where you're saying, I'm presenting myself as a possible mentor to you. And
the way you convey that signal is by giving them some useless advice. I'm sort
of open to that. I'm skeptical that that's the main role that it plays. But
maybe that plays enough of that role. That would justify, from my point of
view, if what this were were really just the invitation that they come and
initiate a relationship with you and then in the context of that relationship,
you could actually help them.
but I don't grant that this is useless to the bikes. I think there are some
simple things you can tell people about many situations. I mean, I think this
is very common, not just for a professor, grad students, a new coach and a
variety of people that is. You often start with some basic, simple framing of
the new situation they're in and you give them some very simple warnings and
then later on you go into more detail. So even like to think of a guide to
hiking the Grand Canyon or something, right? The very first 10 minutes about
the Grand Canyon might start to say, people don't realize just how easily they
could fall and here's how many people have died here in the last 20 years,
right? I mean, that would be part of an initial introductory speech by a guy
to the Grand Canyon who's going to lead a group there. And in the process of
being a guide after, I don't know, a week or something, they would have gotten
into a lot of details. But at the very beginning, they're getting some very
broad picture framing and advice, which, you know, makes, usually makes sense.
So there aren't that many. Very simple, big piece of advice you can give, but
there are some. I mean, I think like in an acting class, a similar thing might
happen, right? The acting coach might talk about some very simple failure
modes for actors, sort of thing you might expect to get out of this class
you're not going to get, the sort of kind of effort you're going to have to
put in that you might not have thought you'd have to put in, right? And this
is just a very common introductory mode for any sort of a class is to sort of
I agree that it's common. But why is it useful? Well, I, like, so, um, I mean,
say, say the canyon by saying, um, look, like 80 people have died here this
year or whatever, you know, watch out or something. Um, like, is that going to
make people safer? Or like, what if there's people who just, that makes them
way more freaked out and they're like, so nervous the whole time that that
makes them more accident prone. It certainly seems possible. Or like when you
tell people about here's like a common failure mode, Does that lead them to
avoid it, or does that lead them to be so confident that they can now avoid
that failure mode that they are actually likely to fall into it? I don't know
the answer to these questions. They seem to me like empirical questions, and I
just don't have that much of a guess about which way it would go.
Okay, but notice that these sorts of uncertainties about the value of advice
also apply to all the other cases where you approve. including the coach who
knows you for years, they still face those uncertainties. And even in
instruction, there's going to be uncertainties about what exact situation the
person you're instructing is in.
Right. But so what I'm saying is the uncertainties are just really big. So the
coach make predictions about how you're going, how you specifically are going
to react to certain pieces of information in order to give you, because they
know you. So they know whether you're the sort of nervous person who, if they
say, eight people died here this year, that's going to freak you out so much,
it'll make you less safe. Or you're the sort of person who will take that in
stride, or you're the sort of person who doesn't even need to hear it. The
coach does that.
They guess that, they know more, but they still have a situation.
They have like a reasonable basis on which to proceed. They just know more,
but there's wrong.
But you can be wrong in all these cases. I don't see why, you know, the very
simple initial advice needs to be, have so much more radical uncertainty about
its value than any other more context particular advice.
I guess like the, um, the very simple initial advice, like for writing, it's
gotta be like write every day. That's something that everyone says. So that
must be, um, like, I mean, maybe like, let's say, you know, that you're say
51% more likely to become better if you write everything. Maybe there's some
tiny percent by which you're likely to improve by following the advice. But,
you know, if I sort of introspect, like, I might, I might become more friendly
to advice like that if it were presented with the appropriate level of
confidence, which should be unbelievably low. Right. So if you gave a speech
to your students and you're like, look, there's only a small chance this will
help any of you and for several of you, it's going to make your lives work. or
to take my advice. I'm not sure who that is, but I think it's fewer than 50%,
obviously. And I don't really know you. So you said all that. And you're like,
here's some things that have a tiny chance of maybe being better than a
percent of helping you. Here you go. And you presented it that way. And you're
like, but I don't know that this will benefit you. And I don't know that this
is what you should do. Maybe I'd be maybe I'd be friendly, but that's actually
not how people give us our advice.
I'm happy to support, you know, complaining about overconfidence and
overestimation in all contexts, but It seems to me, again, the key point here
is to well calibrate the connection between the usefulness of your advice and
all the obstacles to it being useful. And, um, so here's another example that,
um, to me frames the, you know, issue of advice. Um, it was in a book about a
manager management and power, I think, and a manager, um, You know, I was at a
high level where regularly many people were asking them to have meetings to
get advice from them. This is just a thing that happens at a certain high
level of management. Many lower level people regularly ask to meet with you to
ask for your advice. Then this person announced their retirement. Now they
were going to retire for another six years. And the day after this
announcement, they were just as knowledgeable about this organization as the
day before. But as soon as they announced their retirement, all of a sudden,
all these people wanting to ask for their advice dried up. Nobody wanted to
ask them for advice. Right. So that suggests that this advice relationship was
less about the advice itself and more about other things. that if they were
implicitly asking maybe for support for their choices or support for their
Or just showing deference in some way that they think would be rewarded and
like a show of respect.
Right. So there's all sorts of those things being mixed up with advice. And
that's one of the harder things. So analyzing advice is we're tempted to take
this simple surface function of advice, i.e. giving information, and we're
tempted not to focus on all these other social functions of advice.
Maybe here's a way to think about it. So like with conversation, right,
there's the, um, uh, there's what we say we're doing in conversation, which is
like sharing information. And then there's what you think we're actually
doing, like in your book, the health of the brain, which is, um, uh, showing
each other our toolkit of skills. I don't agree with that, but in any case,
that's what you think we're doing in conversation. And so you think that a
relatively small percent of, um, what's happening in conversation is the
actual information transfer, right? Right. Okay. Like, so how small, like 5%
I mean, it depends on which categories of things we're averaging over.
Okay. Suppose we just have to average over all conversations that purport to
So, you know, the case where it's lowest might be sort of chitchat. But if you
sit down- I don't think it's the lowest. Okay. Maybe I would think, say you
sit down at an ice rink and you try to put on your ice skates and there's a
person there to advise you about how to put on the ice skates.
I think it's going to be relatively high there, over 50%. That is, there are
some particular contexts where you can infer that the main reason for the
conversation is some information, but even then it's not 100%, it's maybe 50%.
Right. Instruction, as I would call it. So if we go to advice, right? And now
we go to the kinds of contexts that I'm going to think of as canonical. Two
people who don't know each other. One person is telling them how to do
something where they don't have a shared conception of what the thing is and
there's no algorithm for achieving it. Like you telling your students how to
succeed in grad school. In that kind of context, the amount of information
being transferred, the useful information, what percent of the interaction is
that, is useful information transferred?
I find it hard to delimit this category that to you is a clear category and to
me, I'm pretty fuzzy about it, but I'll guess 20%, 30%.
Okay, so that's, I mean, in a way, you sort of agree that most of what's going
on in advice giving isn't anybody helping anybody else.
Well, via the advice, there are, the other things could be helping.
Right, right, right, right. Like getting yourself recognized as an authority
and all that, sure. And then, so then if we say only 20% of it, and then we
also look at like, in that 20%, how valuable is the actual information being
given, right? And it's, you know, it's not going to be like, really good,
because we just don't, these are things we already don't know about. Then,
like, I think we could sort of compromise on the view. I'd be able to. Um,
people can't really help each other very much by telling each other how to do
vaguely specified things where there's no algorithm and they don't know each
other well, but they pretend like they can do all of those things.
So I have some advice for people who ask goodbye. So. You're allowed to have
advice. And I'm often in a situation where somebody junior wants to initiate a
conversation with me. Yes. And they often think the safest form of
conversation, the conversation they choose is an advice framing. They ask me
for advice, right? Yes. And I want to tell them that's not usually a very
That is, I think they feel like that's very deferential and respectful. And I
want to tell whoever that is listening in that role, be a little less
deferential and respectful. Pick a topic and discuss it. that you know this
person's interested in and you're allowed to disagree with them. And, you
know, it would be more interesting to talk to you and this person in that
mode. You know, I might remember something about you. I might see your ability
to reason and think. That would be a potential result of this top, whatever
you pick that we talk about. If you just say, how should I become a professor
or how should I learn to write or some sort of generic thing like that, first
of all, I don't know that much about it. I don't usually have good answers.
And secondly, I'm not learning much about you in the fact that you're asking
Maybe what you're saying is like, look, this is the form of deference I'd
really prefer. I don't like the kind you're doing. I like this other. It's
actually not very differential to me if you're just saying, you know, whereas
like my kind of deference, like we'll really get you somewhere with the same,
the goals that you have.
Well, I don't want to call it, there is a matter of deference in the sense I'm
saying this is what I prefer so that you hear it. If you give me what I
prefer, that's a kind of deference. But I'm saying there's also another
function, which is if we just talk together about anything, we will learn a
bit more about how the other person thinks. And that's just much more valuable
for both of us than, you know, generic advice.
Okay. We should stop there anyway. We're, we're finally agreeing about
Yes, I agree. I accept that advice that we should stop. All right. Bye.