Paradox of Honesty
So here’s a paradox (I didn’t come to this myself, I got it from my husband):
You might think that honesty is a kind of terminal norm of communication. So
like you should just hold up your communication to the standard of it’s being
honest, which is to say, being as truthful as you can. That’s one horn of the
dilemma. The other horn is that every case of communication, in addition to
being a case of speech, is also a case of action, right?
So every act of communication, you’re doing something, and there’s only one
terminal norm of action, which is do what’s good, right?
Accomplish what’s good.
And, you know, if we accept the second horn, we have to say, you know, be
honest, if it happens to be good.
Right? But otherwise, be honest. And so, how can… So what’s your solution to
I will frame this in the context of the concept of norms.
That is, I will say, you know, typically we each just have a complicated set
of things we want, some of which we can articulate and some of which we can’t.
Other people have other things they want, some of which they can articulate,
some of which they can’t. And we often find that we are in conflict with each
other. And unless we coordinate, these conflicts will be destructive. And so
we often adopt norms as a way to adjudicate these conflicts to make us all
better off. Sometimes we embody these norms in laws and we make them more
But before there were laws there were norms. So a norm is a
description of a feature of behavior that you should or shouldn’t be doing
that is not just a good thing or a bad thing, but it’s a good or bad thing
that we choose together to call a norm and to celebrate or denigrate, and that
we are going to go to some degree to promote this norm by praising the people
who do the good thing or criticizing the people who do the bad thing. And
that’s what a norm is. And being honest is a norm.
What about the other side, doing good—is that also a norm?
It’s more than a norm, it’s more what we each might want to do if we were
feeling especially generous. It’s also maybe the goal of the norm system to
get people to do that. But the key idea might be that we can’t just set up a
norm where we say, Each of us should just do what we can to just make
everybody better off. That— In some sense, that would be the ideal norm: We
just say, Everybody should just do what they can. Sacrifice sometimes
personally if it will help us altogether, and we can define this altogether
goal. And that would be the ideal norm, but it’s hard to enforce. Because it’s
hard to tell in any one person and action whether what they’re doing is, in
fact, helping us all.
And so what we do is we make a compromise, a
trade-off, where we choose rules of thumb that are easier to identify whether
you did or didn’t follow the rule of thumb, so that we can then praise or
criticize you for following that, when we choose those rules such that on
average we think we’re all better off if we have the rule than if we don’t
have a rule. But we know, you know, it’s not always going to be the right
thing for our goals to follow the rule, and that’s just a cost.
So I hear you as just grabbing one of the horns of the dilemma, which is to
say, Well there’s this— I mean in a way there’s this norm, which is like a
heuristic, right, for doing good, and then…
Well, it’s more than a heuristic. It’s a collectively enforced heuristic.
We’re gonna make you follow this heuristic, even if you don’t think it’s the
right thing to do.
Right. But then there’s the actual goal of the whole norm system, right?
Which is like doing, you know, while…
Good for the rest of us.
…doing good for everyone, including oneself. And so in effect you think, like,
look, the second thing is the thing that’s really valuable in the end, but we
can’t sort of enforce that and enforce coordination around it. And we enforce
coordination around it by way of these collective and enforced heuristics
that, you know, like, are going to achieve more good even though in a given
case they might…
…achieve less good. So you suppose you know you’re in one of those cases,
right? Suppose you’re in a case where at least you believe that you can do
more good by being dishonest than by being honest. And that calculation
includes any penalty you think you’re likely to face for dishonesty.
What do you think you should do?
Well, so the easy— So these rules of thumb tend to have a number of standard
exceptions. For example, one of the exceptions is if you can convince us that
in fact clearly we’re all better off if you make this exception, then we allow
that exception. For example, we might say, Usually we can’t tell, that’s why
we have these rules, but if in this case you can show us and we are convinced
that we’re all better off here, then yes, we will allow the exception.
Of course if you could actually convince the person you wouldn’t have to be
Well that seems almost like— So maybe there’s a third party I’m being
dishonest to, for example…
So you know, the classic: The Nazis come to the door—“Are you hiding one?”
“No, I’m not hiding one.” But I’m really hiding someone. Right?
So that’s a classic, you know, reason why you might want to lie and be
So we might expect that some audience, when they hear about this later on,
they will praise your actions. Instead of criticizing they’ll say, Obviously
yes, we all agree that in this sort of situation you should have been
dishonest, that that was fine. But what if you don’t think you can convince
other people? Well now, yes, as you alluded, you’d have to do this
calculation: Well, if I do the good thing, then—and other people find out; I
mean, they might not—but if they do they might punish me and sanction me
through my reputation, including…
And that will have cost not just to me,
but to everyone, in the sense that I might, say, lose my reputation for being
honest and a good person, and then I might be listened to less in the future
and be taken less seriously. And that would be a cost not just for me but for
other people. So I’d want to weigh those costs.
But yes, I would think if you’ve— I mean it also matters just how much you
care about other people. And I’m going to take that as some feature of you
that’s not obvious. So I say all of us care about other people to some degree
(we care about ourselves a bit more), but weigh together, in this situation,
how much you do care about other people and how much you care about yourself.
And putting that together, if the net effect is positive from you, say,
telling the tr— lying in this case, then yeah, you should.
It’s weird to me that you’re giving me a normative assessment of what you
should do. But do you want to take it as fixed how much I value other
people—which might be very little, right?—instead of saying like, Well step
one, you should value other people a lot. And now I’ll tell you what you
should do given that you value other people a lot.
So why are you inclined to kind of insert the normativity only in that spot
but not in like— presuming you have views about how much you should value
Right. Okay, so you’ve given me the opening here to do a little dealism as I’m
explaining [laughs], because that’s the answer to your question. So many
people, even most people, are very comfortable, often, talking in moral terms.
And so I feel I should learn to talk in those terms too. But I don’t
necessarily always think in those terms, and so I have a translation in my
head between the terms I would rather think in terms of and the terms people
more often talk in terms of.
So I’m an economist, for example. And
economists have a simple standard way we evaluate policies called economic
welfare, and it’s a simple function of different people’s preferences. And
philosophers, sometimes, like yourself—though I don’t know you in particular
ever did—criticize economists by saying that we are neglecting a lot of
complexity in moral analysis. Our simple proxy for what’s good is way too
simple, and doesn’t include a whole bunch of other important things.
so this is a common criticism of economic analysis. And of course it’s
claiming that, you know, morality is complicated, and there is no simple
summary, and you’ll have to dive into this deep literature if you want to be
able to say what’s moral. Which of course gets in the way of our attempts to
do a lot of broad, ambitious analysis with relatively simple tools.
so having said all that, I would say an important role that I can play as an
economist, and that economists can play in the world, is to suggest to people
deals they can make. So if there’s a legislature trying to pass a bill, or a
board of directors trying to make a decision about a firm, or a church trying
to wrestle with a key controversy: Basically they will be looking to come up
with a deal that they can— enough of them can get behind. And when they do
that, they each have their different preferences, which include their concept
of morality, and how much they’re committed to it, and its realization in this
context. But fundamentally they have preferences, i.e., what they would choose
in various— given various options.
And standard economic tools of welfare
analysis are exactly well suited to suggest deals that they might all be
willing to adopt. And we economists, therefore, using our standard welfare
analysis in this sort of context, are telling people, Consider the following
deals as things that you might all prefer over other potential deals. And
that’s what I think of doing in many of these contexts, is thinking about what
would be the deal.
So in that context I have to take any one person’s
degree of altruism as a given. I have to say this person cares about other
people a lot, that person doesn’t. And when I’m thinking about deals they
could make together, I have to take that as a given and then suggest what
deals would be. So that’s the context of which, even about myself, I might
say, Well, depending on how much I care about other people.
Now in this
context, moral persuasion is certainly allowed. That is, you know, when people
get together they will be trying to persuade each other to have different
moral attitudes, and to perhaps subtly announce that if you violate their
moral standards, they will be complaining to some audience, etc. So there’ll
be moral persuasion.
And you could even, in the context of seeing the
deal we might propose, tell people that violates a moral principle and that
they should change their preferences and context of seeing the consequences of
their preferences acted out in this context. That’s all fine and legitimate,
but nevertheless, given any set of people and the current moral positions they
have and their current preferences, there is a set of deals that’s the, what
we call the Pareto frontier: The maximum best deals they could all get, given
that other people have different preferences. And that’s what economists do.
We offer that set of preferences.
So is the reason why you’re inclined to offer these deals to these people, one
of whom may happen to be yourself, is that you just happen to have a brute
preference for people making deals? And so that you’re satisfying that
preference of yours by offering these deals?
Well I think the world is better off when they can find better deals; that is,
all else equal, a world of people who make bad deals is a worse-off world.
That is, they’re all getting less than than they could from their point of
view. So I think on average the world’s better off. Now, in any one case I
might think that a particular deal is bad for the world.
But the fact that you think the world is better off if people make deals: That
by itself would motivate you. You would have to have a preference for the
world being better off, right?
Some nasty person might have a preference for the world being worse off.
All else equal. But again, that doesn’t make me full altruist. I still might
put more weight on myself relative to the rest of the world.
Right. So you sometimes might not offer deals even when you see them because
you see more advantage to yourself in not offering the deal.
If there were no cost to making exceptions. So this is related to the
exception of the two norms we talked about before, right? So I want to take a
simple stance to the world. I want to say, I’m an economist, and when I put my
economics hat on this is what I will do for you. And the more I have to
explain, “Except…,” and the more that’s complicated, the harder that stance
will be to take.
And is the wanting to take the simple stance—is that a brute preference?
No, that’s part of being able to be an expert in the world. So you know, if
there are many different kinds of experts, that for each kind of expert, that
kind of expert is more palatable, more attractive, if they present themselves
in a simple, understandable way, as to the kind of advice they will give.
Is that because the expert in some sense has to convey to the non-expert
information that the, in some sense, the non expert is not in a position to
evaluate, so that the expert has to be able to almost have like a cloak of
Well, it’s more that you might help— use different experts to check on each
So for example if a— Doctors might present themselves to you, as I think they
do, as “The advice I will give you will be the thing that promotes your
health, taking somewhat into account cost.” Okay, but I’m not trying to like
get you to marry my daughter or something, right? Like, that’s illegitimate as
a thing a doctor would be doing with their advice, right?
And so if in
fact some other doctor saw them recommending that you marry their daughter as
a solution to your medical problem, they could call them on that and complain
to the larger medical community that they have broken the understanding of
what it is to be a doctor, and therefore risk the reputation of doctors if it
became more widely known that this sort of thing was happening a lot.
So I just want to go back to honesty. One way you might think about honesty is
that it’s a kind of value of communication, where different arenas of human
life have different sort of forms of goodness that are the characteristic form
of goodness of that thing. So, like, art. Okay, let’s say art is supposed to
be beautiful. That’s controversial, but it doesn’t matter. It’s just an
Okay. So you might say, Look, beauty is the characteristic value that art is
supposed to have. You might say, you know, of certain kinds of entertainment,
its being entertaining or interesting or engaging to your attention might be
the characteristic value of that sort of thing, right? That is, you know, the
world is full of like heterogenous kinds of goods, right?
And so you might think that honesty isn’t just a rule that we follow while
talking, right? Any more than beauty is like a rule that we follow while
making art. We might follow lots of rules while talking. Right? Like we might
follow rules about, like, don’t spit in a person’s face, and don’t be nasty
and whatever, but that honesty is more intimately connected to
communication—that it is sort of what it is for the communication to be good
as communication in the way that someone might argue, though controversially,
that beauty is what it is for art to be good as art.
And that your
conception of, There’s just a norm that we agree on and coordinate around to
constrain people for the greatest possible benefit—that would be consistent
with we can have a norm like don’t talk too loud, right?
We do have such a norm. We have a norm for how far you’re supposed to stand in
front of a person. We have all these norms to coordinate communication.
Honesty seems different to me because those things don’t seem to me to be what
it is for the communication to be done well. Those just seem to be arbitrary
forms of coordination that we had to adopt so that we can get along. But the
honesty part is what it is for the communication to be good.
So it’s certainly true that norms vary in a number of interesting dimensions.
One of those would be how important is the problem the norm is addressing and
how badly could things go if the norm is violated. We do that with laws as
well, of course. And so in the context of art you could certainly say that,
you know, the norm of “don’t bother people” is less important than the norm of
being beautiful. Therefore offensive art is okay.
But they’re both things
we care about; we just care about the beauty more. So… And we also have norms
areas— we talked before on how enforceable they are, how easy it is for
someone to see, and then therefore what sort of evidence you would need to be
convinced there was a norm violation.
So I’m happy to admit that in many
different areas there are different issues to different degrees. So I would
certainly say in art, beauty is a bigger issue than convenience even, perhaps.
So if you had a really enormous painting and it was really beautiful, and you
might say, Well, that’s just inconvenient to put in our museum, people might
say, No, we’re fundamentally about art—we must find room for this piece of art
even if it’s inconvenient, because that’s a more fundamental value in our
So with respect to honesty, I’m happy to admit that, you know,
honesty could have— Dishonesty could have much larger consequences than some
other norms like spitting in the face. Spitting in the face is not good, but
it’s not that big a deal in the context of most conversations. But honesty can
be a really big deal, and so we might want to therefore, you know, pay a lot
more attention to honesty than the others.
See, it doesn’t seem true— Spitting in someone’s face is a pretty big deal,
and you’re likely to start a fight. And a given instance of dishonesty might
not even be such a big deal. It’s— so it’s—
But then spitting in the face won’t have spreading repercussions for the rest
of the community so much. Where dishonesty— If I tell you something dishonest
and you believe me, and you spread it, and then, you know, it could keep going
farther. And so there’s more— harder to prove all the risks. If somebody spits
in your face you’ll probably notice it; if you don’t notice it, it’s not a
problem. And so you can— Even if it’s a direct personal harm, it’s not
something the rest of us need to pay attention to with norms because you will
probably deal with it yourself, the pair of you.
Suppose you had to guess like, the, you know, average person, how many times
do they lie in a week? You just had to give a number. What number would you
I just happened to, yesterday, look at an old blog post of mine.
Talking about how in a five-minute conversation there would be several lies,
Okay, so that seems to refute the claim you just made that like, if you
violate the honesty norm the whole system is going to collapse. It seems like
we violate it all the time and the system doesn’t collapse.
Oh, sure. It’s just— There would be large instances that could be a risk. But
yes, most lies are relatively tiny.
Like I think, actually, honesty is interesting. You could think of a norm as
like— One way to assess a norm is like, How robust is it to violations, right?
Some norms, if you start violating them, it really starts to feel like that’s
actually not the norm anymore? Right? So I actually think beauty is an example
of this, right? And there was a kind of like outrage in the artistic community
in periods where the art moves away from the norm of beauty. And I think now
it’s actually quite controversial even to say that beauty is no regard, right?
Because that’s being the norm is in a way predicated on people following it
And honesty seems to me to be an area where that’s not
true, where people violate the norm all the time. Like every day, and
everybody knows that everybody violates the norm. And yet it’s such a profound
value to us that we hold on to the norm in the face. It’s extremely robust. So
what would explain that?
I’m not sure this is the answer your question because I have this thought in
my head that was responding to your previous question, which is that in
different areas of life we like have a whole bunch of things we care about,
but we also, I think—you know, to accept your point—we coordinate on what
we’re doing together there. Right? And so we might, say, decorate shoes. And
because we like to have decorated shoes and we’re going to a party tonight
where we have to use our decorated shoes, and in that context we might not
frame ourselves as artists, we might just be decorating shoes.
might do roughly the same thing but think of ourselves as artists. And now
that would influence what we’ve decided to do together and the norms that
might make sense. So if I’m offended by the way your shoes look if we’re—if
this is art, I might think I have more of an obligation to accept it. Whereas
if we’re going to the party tonight together wearing the shoes I’d think,
Those are ugly shoes, try again.
So in that sense, I could see a place
for making a distinction, you know, on—in terms of the priority of norms, and
the kind of norms, etc.—on what it is we’ve decided we are doing in a context
relative to other considerations. And so if you want to say that in some
contexts, what we are doing is more central to honesty, that’s the thing we
are saying that we’re doing together, then in that context honesty would be
But it seems— I mean it seems like at that point the paradox recurs. Like, how
can honesty be the thing that we’re doing together? Isn’t the thing that we’re
doing together always just doing what’s good? And if what’s honest isn’t
what’s good, then…
No! I mean, so the world is huge, we are a lot of different people, we have a
lot of different hours in our life. And so we specialize, right? Each of us
does different things. And at any one interaction we frame it as the kind of
thing we’re doing together. And that helps us be more effective because of the
great division of labor. If every relationship was simultaneously a
doctor-advising relationship, and an art relationship, and an econ-consulting
relationship, and an auto fixing relationship, then, you know, we’re just
trying to do the good but we’re not sure what we’re doing. And we don’t
necessarily coordinate well.
So to coordinate well we need to agree
together, roughly, on what we are doing in any one context, and have that be
the focus and priority of that interaction. You know, it could be— it’s open
to change. But we do need to choose what we’re doing, and then it won’t just
be doing the good, it’ll be doing the good in that context. It’ll be the kind
of good that you have agreed to do together.
So then let’s say you’re right, and that there are sort of going to be like,
We lie a lot in conversation but there’s maybe going to be some contexts in
which honesty is especially important…
For a trial, for example, where you are sworn to be honest and it’s not— You
know, and it’s considered much more of a problem if you’re dishonest when
testifying in court.
Right. So that’s one way to know. But normally if you want to know whether
you’re in the context in which that value is paramount, or like— It’s not
always made explicit. So like, are we in such a context now? Or how can one
tell whether one is in the honesty context or not? Because it seems a little
bit puzzling. Like if you were to tell me, We’re not in the honesty context
and I can’t believe anything you say—is that a thing you can say? I mean I
think you can say we’re in it, but can you say we’re not in it?
Well, I agree that in almost all conversations there is some implicit norm of
honesty default with an expectation of common violations. And if you’re going
to be unusually dishonest, you may need to flag that; i.e., actors on the
stage have flagged for themselves that they are unusually dishonest with each
other, they are playing these characters, and they are not themselves.
Do you think actors are being dishonest?
There’s a sense in which they are. But I mean, it’s less important—I mean
less, you know— Exactly what the word “honesty” means matters less to me than
what it should mean, maybe, or what the useful concept is. Similarly, I don’t
think I’m going to be very helpful in finding out what the actual honesty
norms are, right? To find out that you’d want to specialize in, you know,
surveys and talking to a lot of different people, and cultures and ages and
genders etc., to see what their attitude toward honesty is in different
And then you want to know what what their attitude— which of
those attitudes are norm-based versus just other-based, right? So but I think
we could have a more productive conversation about what should be the basis
for choosing the honesty norms: How would we know how to decide when we have
good honesty norms? That is, what are our standards for them?
Before we have that conversation I want to go back to this question of, Okay,
so you’re in the situation where you have these honesty norms where it’s
partly accepted that you’re not going to be totally honest. So let’s say we’re
in one of those situations right now, where I don’t have to be like 100%
honest, I only have to be like 80% honest.
It’s probably more like 95. But yeah.
Okay, 95. Alright. And then like it still seems to me that if you learned,
like, you know, that I told this little lie during this conversation, like,
you might—or at least I, if it were the other way around—I wouldn’t be like,
Well, that was within the 5%. I would be like annoyed that you lied. And so—
That’s difficult, because usually you can’t tell…
Oh sure, but I should expect, like, is it irrational? Because…
I mean, basically, you’re trying to estimate their rate of lying—
—but you always probably are with everybody. You don’t get that many signals,
you don’t actually catch them in a lie very often. And so when you do that’s a
big data point. And you’re wondering, you know, updating your estimate; I
mean, maybe they’re pretending to be 95%, maybe they are only 80. And you
know, seeing two lies in a row in five minutes that you caught you might
think, This person’s lying a lot more than most.
Right, I get that one thing that I can do is be adjusting my estimate of how
much they’re lying. But another thing I can do is be morally outraged at the
norm violation. Right? And what I’m saying is, it seems to me that it— If you
were right about how we only expect so much honesty, then when we encounter
lies, at least some of the time we shouldn’t— or much of the time, anyway, we
should be like, Well, that falls within my estimate. So it’s actually a
permissible amount. So they didn’t actually violate the norm, unless they want
to bump the [inaudible]
No, I mean… So in general, in law we have this concept of enforceability, and
we have many kinds of laws that are just quite commonly broken. Nevertheless
we accept that any one case where you find the violation should be enforced.
Like speeding, for example, right? A lot of speeding violations, but we still
say if, you know, they pull them over and they caught them speeding, then
yeah, that they should be prosecuted. So it’s quite possible to have a norm
that’s commonly violated. Nevertheless our norm is still that if a violation
is pointed out to us in a clear way, then we will apply the usual enforcement.
But I thought you were saying that we expect different degrees of honesty in
different contexts, right? And so I was trying to pick one of those contexts
in which we only expect 95% honesty. And if you only expect 95% and I give you
95%, which includes my lie that I told you, then I haven’t violated the norm.
If the norm were to be 95% honest, but that’s not the norm. The norm is to be
100% honest, but it’s violated 5% of the time.
But I thought you said that there were contexts in which we’re doing honesty
together, right? And like in the law it’s like serious, maybe in the law it’s
Right? But now, you know, we’re not— this isn’t— We’re not in a law court,
we’re just doing a podcast. So maybe it’s 95%. So that’s our context. Doesn’t
that mean that that’s a different norm that we’re applying? We’re applying the
norm “be 95% honest.” Such that if I were only 10% honest you would have a
real claim against me: “Agnes, you’re lying too much.”
So take speeding, right?
There’s a norm not to speed, but the degree of severity of the violation is
considered to depend on context.
So in that sense you could say the norm is different. Like if it’s raining
heavily, and people can’t see very much, then speeding is a bigger violation,
whereas if the air is clear and there aren’t very many cars, then speeding is
less of a violation, etc., right? So we have a way we vary our speeding norm
by context even if we might enforce any one case that we see as a violation.
As we still have the rule, Okay, you were speeding! Then that’s breaking. We
might say, Well, we’ll punish you less in this context for speeding than in
that context. So the norm can vary by context without necessarily giving an
excuse differently in different contexts, although that also applies.
Right. Okay, so you might say, for example, if my lie was small enough— like
suppose you asked me how I was doing and I was like, “Good,” but actually I
was like not doing well, right?
I could even see that on your face, perhaps, but…
And then I’m— And then you’re like, Okay, but that’s like speeding five miles
an hour over the limit, that’s fine. I’m just gonna let that one go. Right? So
that’s your— Is that your thought, that we are enforcing these norms, and
there are these little lies which even if we catch the person in the little
lie we don’t get outraged. And then not getting outraged is like the not—
either small fine or no fine.
I’m not even sure outrage is binary either. So again the legal analogy, I
think, helps us show the range of dimensions that are possible with respect to
laws, and most of those dimensions are also possible with respect to norms. So
laws can vary in not just, you know, where the line is drawn, say, for
speeding, but, say, the size of the punishment by different contexts. We can
also vary how much we pay for enforcers to be out there looking for them. We
may also vary how much we will trust any one witness who claims to have seen
it. There’s just all these different dimensions.
And then yes, of course,
typically the enforcers have a budget and they choose priorities. And they
quite often do—the police quite often do say, Jaywalking, you know, it’s just
not a priority. Unless somebody really puts it in our face we’re not going to—
Even if I see it as I drive by I’m not even going to do anything unless
somebody makes me, right?
And that’s also the kind of thing we do with
norms. We, you know, we know that when we see a norm violation that we could
complain about it. And we would be entitled to complain, and, you know, if it
were the following characteristics, another would feel obligated to
acknowledge that we— It was a violation, and to join in sanctioning it to some
But for many more violations we see we don’t actually bother to
complain. And this outrage in your head is some sort of indication inside you
of how inclined you are to complain. You know, if the outrage bubbles to a
high enough level you just feel compelled to complain, but if it’s a low level
of outrage, then it’ll be, How busy am I? And you know, Do I have a beef with
this person I might complain about? And you know, you have all these
different, you know, factors that you include in deciding subconsciously how
outraged to feel and whether to complain about the violation.
So there’s a different way to see honesty. Like the way you’re seeing it is
you’re kind of taking for granted the concept of coordinating with people…
…like that we coordinate with other people, but then that coordination—
We use norms to do so.
—has to be governed by norms. So the norms…
Well, it doesn’t—I mean, there are many ways to coordinate. It’s one of the
ways we coordinate.
Okay, but, right: So norms come in as a means of kind of like regulating this
coordinating behavior. But you might think— So like, I think— I connect
honesty up with concepts like trust and betrayal, right? Where there are many
norm violations that someone could do. Like spitting in my face would be a
norm violation and might be very upsetting. And it might be a betrayal of
trust under the relevant context. But you know— but like certainly something
like speeding, it’s like you might be breaking a rule, but I just might not
care that much, right?
But if— Especially in some contexts, lying, what
it seems to do is to shake the very possibility of coordination. That is,
you’re helping yourself to the idea of coordination, like here’s this thing
coordination, right? But you might think that part of what honesty is is it
sort of describes a kind of coordination. That is, if I’m honest with you I’m
like giving you access to my mind, right? And I’m sort of like letting you
into my mind. And so then when we communicate, we are in effect coordinating
our minds. If I’m dishonest with you then I’m having you coordinate sort of
with someone else who I’m not, right? With like an image that I want you to—
you know, and I’m in some sense then, not even— I’m like almost manipulating
your mind, right? I’m like trying to give you these beliefs, sort of, right?
I’m trying to give you these beliefs that are not the beliefs that I have.
Right? And so you might just think honesty is a kind of name for coordination.
It’s how we coordinate, it’s not that we coord— You know…
No, I mean, so for example, one way we coordinate is trade. Right? And so
imagine the typical bazaar scenario: You’re walking through a bazaar and
there’s a rug you find intriguing. And you pause and you stare at the rug, and
the proprietor says, This rug, for you today, discounted. I offer you, you
know, x price. And you say, No, it’s not worth that much. I would bet most
people like pay y for this. Right? And then you go back and forth in price
Now everybody knows that that’s usually somewhat dishonest,
in the sense that, “The lowest I will take was x” is not true, because a
minute later you will take x minus 100. And vice versa. But it’s part of the
usual routine of negotiating and trade, is to make offers and counteroffers
that you sometimes pretend are your final offer. And you coordinate
successfully in the sense that you often do make the trade and buy the rug.
you both know that the other one is going to be trying to get the best deal.
And part of their trying to get the best deal is to not admit quite as much
how much they want it, or the cost of giving it up. That’s a context in which
a very standard, accepted degree of dishonesty is part of the coordination.
Right. So maybe one way to think about it is we could take coordination and
evaluate it on a metric. And the metric would be, kind of, closeness. And I’m
not sure exactly how to sort of— how to make this precise, but it would be
something like, there are forms of coordination that are very loose, right?
And then there are ones that are tighter. And the loose ones—it’s like, you
know, there could be a thing, for instance, where there’s like a location, and
people just leave stuff there when they don’t need it, and then other people
take stuff when they need it, right?
Exactly. That would be very low coordination.
Very low. And like a market is tighter than that, right?
But so I think of honesty as maybe marking sort of that innermost boundary of
extremely tight coordination. Right? And that’s why it’s in a way right to say
honesty is coordination, but also true to say that there are forms of
coordination that are dishonest, but those are the looser ones.
I might more just frame the question in terms of what facilitates cooperation,
and that includes multiple forms. And when we get to that level, you know,
much of, say, negotiation… Negotiation is one form of coordination, right?
Where we make offers to each other and then make a choice. And in negotiation
over deals, a reputation for honesty, and a habit of honesty, and a belief in
honesty, will help it— help you be easier to make good deals.
So that’s a
general feature of negotiation. If both of you can just be honest about what
you want, what you don’t want, then you can more quickly figure out the deal
that gives you, you know, the most for each of you. So, and that’s good,
right? So that’s a sense in which honesty helps coordination. It’s not the
only thing that helps coordination, though—we can identify a number of other
features in these contexts that will promote coordination. And there are other
kinds of coordination. But in other kinds of coordination, honesty could still
help with coordination.
So I have like a general rule in how I think about reputation for— Anytime a
reputation for x is valuable, that’s going to be because, independently and in
the first instance, x is valuable, right? So saying like a reputation for x is
good is explanatorily posterior to some other thing, where you explain why x
itself is good.
But honesty facilitates core— deals, deal making.
Well, so I want to say, you know: Coordination, deals, cooperation, all of
that is… You could view it in two ways: You could have the number of deals,
right, so like people can make more deals, right? And that’s the thing you’re
into, them making more deals…
We would want to do a weighted sum of the value of the deals times the number,
Okay. But there’s another— I’m looking at it from another angle. It’s like,
imagine there’s another dimension to the problem in addition to the dimensions
that you’re looking at. And my dimension is how tight is the coordination. So
like, if you look at circumstances in which dishonesty is viewed as especially
betrayal, like, and violation of trust and whatever, it’s going to be saying
intimate relationships, right?
So those tend, I think, tend to be the context in which people view dishonesty
as especially destructive. Now…
The tail of the distribution is there. I mean, in most relationships, there
are conversations they have on elevators, plenty of very low-scale— some of
Yes, absolutely. And so even there, there’s going to be maybe certain topics
on which dishonesty about those topics is critically problematic. But the
point is, what you have— In an intimate relationship—what is an intimate
relationship? It’s a form of coordination, right? It’s a form of
cooperation—but it’s, you know, there’s a difference between the way you
cooperate with your child or your spouse on the one hand, and the way that you
cooperate with buying or selling something.
And the way that I’m trying
to describe that difference is that you— it’s more tight with the spouse or
the child. And so because you’re doing a different kind of coordination,
right? One that is tighter, which is to say, one that in some sense it like
more directly involves like your mind coordinating with their mind. Honesty,
like, is the… It sort of describes that form of tight coordination. And then
as people do the looser forms of coordination they can leverage off of that,
and be like, Oh, you’re a trustworthy business associate because you treat me
almost as though I were your wife, and you’re gonna be that honest to me,
So that’s your reputation, right? The reputation for honesty, in a
way, is like a reputation for taking this very tight form of coordination
that’s appropriate in intimate relationships, and actually like, you know,
having that be like— apply more generally.
I agree that we can spread, you know, relationships on a spectrum, tight or
loose, sort of deep, intimate, you know, strong relationships on one end, and
loose, distant, temporary relationships, say, on the other end of the
spectrum. And that one of the, you know, classic descriptions of the modern
world is that we have fewer deep relationships and more distant relationships.
And you know, that’s the market economy, for example: More trade with arm’s
length relationships. And in the past, you know, your cobbler was your friend,
and you knew them, and the, you know, etc.
And you know we also, like,
people are nostalgic for times of war sometimes after the time of war, because
“We all felt so close back then.” And they felt that there was just a lot at
stake in their relationships, and they felt bonded to each other because they
had this common enemy. And so there’s certainly a lot of interesting things to
say about to what degree do we want relationships to be how close, and whether
we’re getting enough of that today.
And it’s certainly true that for any
one thing you might say you could get away with lying about it more easily to
a stranger than to a close intimate. They will, you know— And that you are
therefore less likely to lie about any one thing merely because they would
know different, right? At the bazaar, selling the rug, you could tell them
that you were a— You could tell them that you were a computer programmer who
doesn’t have much use for rugs and not be honest that you’re a dentist and
plan on putting it in your entrance lobby, you know, at which point they might
want to charge you for that.
But with your wife you can’t lie to them and
tell them that you’re a programmer instead of a dentist because they will know
otherwise, right? So it’s just true as a nature of the fact that relationships
are close that we just know a lot more about each other. And therefore a lot
of things, it would just not make sense to lie about. But then there are going
to be things that a boundary of things where they might not be so sure, and
then that would be the issue of how honest you are now. It’s not clear to me
that in fact, taking that into account, closer relationships are more honest.
You know, so…
Yeah, good. I think you’re right about that. So okay, so maybe let’s bring in
a third thing here. I think you’re right that this spectrum doesn’t quite
track what I want it to track. And there is one context in which—let’s say
maybe more than one context—which is a kind of conceit that this paradox of
honesty doesn’t exist. Which is to say that telling the truth and doing the
good are just gonna come together.
I think the legal context is one
context where there’s some conceit of that kind, but maybe intellectual
activity, intellectual inquiry, like what we’re doing right now—like, I think
that it would just be weird if I were lying to you. Like it’s a little hard to
imagine that I can achieve some good by lying to you in this context.
Well, I think you just need to spend a little longer, then you’ll be able to
imagine these things! [They laugh] Nothing’s that hard!
Right. I mean one can certainly imagine it in the law case too, right? It’s
not unimaginable, but it’s— What I want to say is it’s a kind of, you know— So
we have these norms. I think the norms of honesty are extremely strong in
intellectual pursuits, right? Like if you’re writing a physics paper or
something, you’re not supposed to like make up false physics.
The lip service for them is strong, whether the practice is strong is a
Okay. Well so you were just saying like, Are people actually more honest in
intimate relationships? Do you think people are more honest in, you know,
their published research or something?
So I mean, actually I would say probably the main factor affecting how honest
people are is how likely they are to get caught if they’re dishonest. So— And
that’s also true for intellectuals. So intellectuals will be honest,
especially carefully, on things that if they are dishonest they will get
caught likely, and where the consequences would seem to be large. So among
intellectuals if you sort of date the first draft of your paper August and it
was really September, we might not get very upset about that. But as other
important parts of the paper—maybe the data set that was all made up—then that
would be a much bigger deal.
But academics are dishonest, as we know, in
many ways that violate their norms. Like, as you know, there’s this recent
replication crisis and, say, p-hacking crisis. And the p-hacking crisis is
about how much search you do to find a particular statistical analysis that
apparently gives statistical significance. And the norm would be first of all
that you don’t do any search, you just define your, you know, procedure up
front, and you just report it. Or if you do search you report the search and
adjust your p values for the degree of search you did. And— But that’s very
hard to monitor and enforce.
And so we see from the overall academic
literature that there is, in fact, a lot of p-hacking going on. A lot of
people search among specifications and then report the p for the specification
as if they didn’t do any search. So— and there are a number of other ways that
the literature on academic behavior finds widespread dishonesty when it’s hard
to catch and enforce. So there are often interesting that— You can look at an
average of behavior and find on average the level of honesty, and not be able
to identify any one case which one is being honest. And that’s these ways
where we can tell how honest people are being in cases where each one of them
doesn’t get caught.
That’s interesting. So people are, I mean, so, you know, I guess one thing is
it’s not going to be a function of how likely people are to get caught, it’s
going to be a function of how— that is, how many lies we catch is going to be
a function of how good people are at predicting, right, whether or not. So
maybe also intellectuals are, you know, they might be good at— they’re good at
predicting whether or not they’re going to get caught, right?
Right. So why wouldn’t they know they won’t get caught more often?
But you know, so do you think— One thing you might care about is whether
you’re going to get caught, and another thing you might care about is whether
this whole phenomenon will be caught even if you aren’t individually. But so
your thought is people don’t care about the second thing, they just care
Right. Because we find it relatively easy to see overall levels of dishonesty
when we can’t tell individual dishonesty.
Right. I mean, so I think that, you know, as I see it, this thing— this
p-hacking thing is called a crisis for a reason, right? And if you did a study
of married couples or something and you discovered like, you know, there’s
like 20% lying or something, I don’t think we called it “the marital crisis.”
So I still think that that is…
I think that’s kind of hype and misleading bluster. I wouldn’t call it a
crisis. In fact, I don’t think they care very much. So let me offer you
concrete evidence on the related replication crisis. So the replication crisis
is that people often do studies, and through p-hacking in part, but in other
ways, to produce a result that won’t replicate. That is if you try it again
you won’t get the same answer. And of course it’s seen as a problem that we
have so many results that don’t replicate.
One of the things you could do
as a journal to not publish papers that won’t replicate is you could do a
survey or a prediction market where you ask people, Will this paper replicate?
And it turns out people can judge that pretty well. People who look at a paper
and have some, you know, experience in replication literature can judge
whether individual papers will or won’t replicate. So journals have it within
their power to reject papers that are much more [un]likely to replicate. And
one way they could do this is to, when they have a submission to a paper, put
that paper to a prediction market, say—ask, Will this paper replicate? And
then if they get an especially low number, say, No thank you, this paper won’t
So I have been somewhat on the periphery involved with the
project, which not only had prediction markets on papers and whether they
would replicate, which showed that in fact it’s relatively easy to predict if
papers will replicate. But at the beginning of this project, you know, I had
the group try to approach journals saying, Would you like to be part of this
project such that you would tell your people who submit to your journal that
they will be part of this project? So their paper will be put up for
evaluation, and the score that we get about the chance of your paper
replicating will go into our evaluation of your paper. No journals were
Right. So I mean, the way to think about this, right, is that the whole
journal system and more largely the academic system is caught in the sort of
crosshairs of the honesty paradox, right? Which is that— Let’s just take at
the level of the individual researcher, the individual researchers incentives
are that they have to produce some surprising results, right, in order to get
published, in order to, you know, get a job, keep their job, etc, right?
so in order to do good as they see it, that’s what they have to do, right?
Whereas in order to tell the truth, right, they’d have to do a different
thing. Right? So we have— There are like pragmatic facts about the way that
the system works, where the desire to do good is going to pull you in the
opposite—at least for yourself—is going to pull you in the opposite direction
from the desire to tell the truth.
Do you think there are any contexts—
So what we can conclude there is, at least as it stands, academia is caught in
this, and the attempts to correct it are perhaps getting caught in the same
trap, right? Because there’s no reason to think that attempts to correct it
aren’t going to like— Once these things are aligned then they would be very
open to suggestions like yours, right? But right now they’re not because the
things— You’re like the same position as an individual researcher.
This is the concept of institutions and institution reform. And this has been
a specialty of mine for many decades. And I’m an economist, and then, you
know, studied this academically, etc. So the key idea is to say that in any
particular area of life, if people have some usual degree of selfishness, say,
or bias toward themselves, then their local incentives produce a net behavior
that can be lamentable for the whole system, and recommending to any one
person that they sacrifice themselves for the benefit of improving the system,
you know, falls on deaf ears. Unless, you know, you could say, We’re trying to
coordinate to a new system. And so that’s the question about institution
reform: How do we ever coordinate to change systems?
Right, but it looks like the person that you’re going to talk to at the
journal is still part of the same system, the same incentives.
Exactly. So in all different areas of life the question is, you know, Who is
the best audience for suggesting reform? And what is the best kind of pitch?
Or what kind of homework should you have done before you make the pitch? That
is, what should they have a right to expect from you in your pitch to make a
Do you think— Suppose they’re unwilling to make the change to their system? Do
you think that that shows that you screwed up? That is, that you didn’t make
the right pitch or you didn’t approach the right person?
Well so in one fundamental sense, we might say, there’s an ultimate customer.
And if we think like an institution is all for the customer, then if we make
our pitch to the customer and the customer doesn’t want it, then yes, there’s
more of a sense in which we didn’t understand the customer. So for example
think about medicine. We say, What’s medicine for? You might say, Well,
medicine’s for the doctors to have a nice office and a hospital, right? For us
to watch medical TV shows. We might say, No no no—medicine is for the
That’s— So it’s all for the patients, right? And if
something’s going wrong, you don’t go recommend it to the doctors per se, or
to the hospital or to the health insurance company. You might try that. Then
you might be able to succeed. But if they say no you might think, The customer
is the ultimate person I should be making my pitch to. Can I convince the
customer to make a switch?
So for example, you know, this happened with,
say, HMOs. Right? We had the ordinary fee-for-service doctor relationship long
ago. And then there was a long period of producing the innovation of HMOs,
Health Maintenance Organizations. And the difference was you pay us per month
and then we do everything for you. And you know, if we have to do a lot this
month, we lose, and we do a little this month, we win. But that’s going to be
the new deal.
And if you suggested that to any one doctor in 1900 they
might say, Well I don’t like it, I think I’ll make more money under the
fee-for-service. But somebody was willing to create an HMO, a new institution,
then try to attract customers. And as soon as they did that the doctors used
the government to try to prevent them from doing it, because they coordinated
to make it illegal and then to put a lot of legal roadblocks.
actually World War Two that really got HMOs going, with Kaiser at the
shipyards in Oakland. Because the, you know, the government needed ships, and
whatever it took to get those workers happy to make the ships, they were
willing to do it. And that they wanted an HMO in there—the union wanted an HMO
for their health plan, and that was okay.
But anyway that’s an example of
going to the customer. And over time HMOs have become much more common. And
many customers became convinced it was a better deal. Like me: I use an HMO.
But many other customers have not been convinced, and you might say that’s a
problem, whatever your theory is that says HMOs are a better deal for
customers. The fact that most customers don’t choose HMOs means that your
model isn’t fully right. You’re missing something.
My view is that medicine is not for the patients. My view is medicine is for
the sake of health. But I think we have to save that for another conversation
because I want— We only have a little bit of time left and I want to go back
to the honesty thing. So what we’ve determined is that there are these like,
two, let’s say, motivations or incentives, that pull apart the desire to do
good and the desire to speak the truth. And that like in many contexts they’re
going to pull apart, and even in intimate relationships they pull apart, and
even in academic or intellectual contexts they pull apart. Do you think
there’s any context in which they don’t? That is, in which one can…
I think I want to resist the framing here. I would say, you know, more
generally, the way we economists think about it is there’s your individual
interest, and there’s a collective interest, and those are pulling apart. And
in this context that happens to be via the mechanism of being dishonest. But
it’s not fundamentally about honesty as the conflict; it’s about the
collective good. That is, everybody who does p-hacking can look and see that
on average we’re all worse off because we do p-hacking. But they each have an
individual incentive still to do p-hacking.
So that seems— Like I get how you could carve things up that way in terms of
if you’re doing institution reform, that’s going to be what you’re interested
in. But I’m carving things up differently in that I’m interested in cases
where you’re going to say what’s false because you think it does more good. In
some of those cases it might do more good overall—like maybe in intimate
relationships, people actually do more good overall by saying what’s false.
I would still say that’s a context that is not optimized for truth telling,
right? If in fact you’re regularly going to be in a situation where you can do
more good by saying what’s false, then you have an, I would say, a non-ideal
communicative context. And so academia too is going to be a non-ideal
communicative context— at least certain kinds of research, right? And in
academia there’s an additional issue, which is maybe that society is also
worse off. But I’m less interested in that. I’m just interested in the pulling
apart between honesty and the good.
Just to connect them together: I might say, you know, in general if you see an
equilibrium where people are being dishonest, I think it’s reasonable to infer
there’s an alternative equilibrium where they were honest which would be at
least as good. And therefore, you know, it’s showing you, Hey, there’s
something better we could do. I think that is roughly right. And in fact
there’s a theorem in incentive design—mechanism design—about basically, for
any equilibrium, there’s an equivalent truth-telling equilibrium where you
tell the truth.
But that’s an abstract formal result. It may not be so
relevant to individual conversations. But I definitely think that it’s a very
strong heuristic and a powerful heuristic that an equilibrium where people are
lying is probably worse. There’s probably some other in principle improvement
where they could tell the truth. But the question is, Can we figure out how to
Okay wait, I want to challenge you to actually produce that in a given case.
So take my— I’m wearing a very pretty dress that has flying fish on it. And so
I ask you, Do you like my dress? And now the equilibrium that we live in, in
my experience, is if I ask you that question, you have to say you like it?
Right? So that’s a kind of lying equilibrium about, Do you like people’s
clothing. So tell me, how do we switch to the truth-teller?
Well in the context of these usual game theories, they’re not including the
effect I think you’re alluding to here. But you know, in a sense we might
interpret my saying “I like your dress” as meaning that your dress is
acceptable. Because— If that’s what it usually actually means. That is, this
word “like” isn’t, you know, its meaning isn’t written in stone. And we could
then… And I think people often do reinterpret these words in these contexts.
No, that’s not what I mean. Like, I mean, like, Is it pretty? Like is it
especially nice? And I want everyone to tell me that all my dresses are
especially nice. And I think you can pretty much get people to do that by
being like, Don’t you think this is an especially nice dress? or something.
And they’ll be like, “Yeah.” Certainly, like, my friends would probably say
like, “Yeah”—in some moods, they would say no.
So in that sort of context I think what you fundamentally want is a relative
ranking. So you know— So for example, think of letters of recommendation for
students or even for professors, right?
Yeah. So we know that people exaggerate. Right?
So, and then the game is, you know, where does it stand relative to the usual
distribution of exaggeration? So when you read a letter like that, you
basically, if they say that this person is the, you know, top 1% of students
I’ve ever seen, you think, What does that really mean? Maybe it means 10%. And
you’re adjusting in your mind the exaggerated claims to calibrate them to a
more realistic distribution of claims. Now if we had a system with letters
where they actually had a budget of these rankings, and they had to like post
all of the people they ever gave a letter of recommendation to and put them in
an overall order, then they wouldn’t be able to lie about that order.
And then when you saw 10%, you know, it really was 10%. It didn’t
mean 30%. And then, you know, we’d all be roughly in the same situation and
getting the same information out of the letter. So the fact is, if my saying
that dress was pretty, I was forced to have this budget of how often, you
know, I said that, and where it was in the ranking, then I might have to
choose, right? If I tell you your dress is at the 80th percentile level today,
then the next person I tell I’ll have to say it’s a 10th— 20th percentile. And
I’m asking, you know, am I willing to pay that cost later to especially rank
you high at the mall?
Right. So—and just as an aside—of all the kinds of texts that I read regularly
in my life, the kind that I have the hardest time extracting information from
are letters of recommendation. And you would think that like, you know, after
like a decade as a professor, I would get good at this. I don’t seem to have
improved. Like basically I can’t get anything out of them. Like I think I’m
unusually bad at like— There’s something that’s being said, and—
This is one of the costs of lying, is that it’s harder to calibrate what
things mean. So I mean I’ve heard people interpreting letters of
recommendation, and what they do is they have a very field-specific,
context-specific interpretation strategy where they know that this person has
said these things, and they heard from somebody else that this other person
has those things. And that means they need to use a lot of context to be able
to interpret it. And that’s a cost of this ambiguous kind of language.
that’s in general a cost of ambiguous language. That is, there’s a sense in
which if we were all just very clear and direct in our language, then we would
all have a lower cost of understanding; that our habit of being vague on
purpose, often, in many things we say, puts us all at a disadvantage, because
now all of us have trouble figuring out what these vague things mean. There
aren’t clear definitions of vague words.
Right. But I think— So going back to the dress case, I think, like, if you
were like, you know, That’s in the top like 30% of dresses that I’ve seen this
month or something, like I wouldn’t be that happy with that. Because, like, I
want that— Maybe I want this dress to be more special than that or something.
Whereas like—and I might be okay with it, right? I might be. But like then if
you were like, you know, Maybe it’s only in the top 50%. Right? Versus if I’m
just like, Do you like my dress, and you’re like, Oh, it’s very pretty, I’m
happy with that whole encounter.
So what I’m wondering is, you told me
that we were going to be able to have another equilibrium that’s the
truth-telling equilibrium that captures everything I wanted out of the lying
equilibrium, in which I’m happy because people tell me my dress is pretty.
And, but it seems to me that the one you’re proposing, with a limited budget
of praise words, it’s not gonna make me happy.
So just to be clear, these usual game-theory models I’m referring to are
assuming what we call rational agents who have a set of preferences, and then
their beliefs are just instrumental in producing actions that help them
achieve their preferences. Now…
Are you saying I’m not a rational agent? [Laughs]
I’m saying that we have a preference for a certain kind of dishonesty. And
that is that it seems more plausible that we just more directly have the
preference for dishonesty here.
And if we do have that preference then the theorem doesn’t hold anymore.
Exactly. Of course.
And then this is part of how we’re built, and this is a key thing to
understand about people. So people are hypocritical and in some ways they want
to be hypocritical, and a world that produces the same outcomes without the
hypocrisy will be less happy because they less have the pretense that they
want. They— People want pretense of various sorts in many ways. They want to
pretend to themselves that they are prettier than they are, and they want to
pretend to themselves that they’re more honest than they are. And many of our
institutions are vague and hard to enforce exactly, to allow these sorts of
And— But so going back to— Do you— Like, so there are different contexts,
right, in different spheres of human life, and honesty is going to be
differentially important in those different contexts. If you had to pick like,
is there some context—and you can be as specific as you want, right?—in which
that preference for dishonesty is at a minimum, what would it be?
Well, my preference would have to go along with other people sharing this
preference. Otherwise it’s not going to work.
Right, so you might— Right, I’m not talking about you. I’m talking about what
context, right? And then who might sort of be in that context.
So I mean, I would think that we have some things we say that are especially
important, where a lot is at stake.
And you think in those contexts? Because you might think you might be very
inclined to lie those contexts, to make sure that—
Right, but I would like there to be a community at least which shares a strong
knowing that at least in this community—and this community specializes in
these important topics, say a community of intellectuals—so for example, say,
forms of government. Like it’s important that we have a good form of
government. Right? So people who specialize in understanding different forms
of government—who would be in a position to expertly recommend different, even
new, forms of government—that’s a really important choice.
And it would
be great if those people in that community had a norm of being especially
honest about the most special, important things they talk to each other about
there. And even better if outsiders believed they had that norm, so that they
could in fact make a recommendation on a really important thing, and have
people believe they were in fact being very honest about their recommendation,
so that we could all then believe what they say, and do what they say, and be
Which of— Suppose you had to choose between them actually having the norm and
them being believed to have the norm. Which is more important?
Well, the conjunction is the important thing. If you believe they’re honest
and they’re not, then they won’t be telling you what the actual best forms of
government is and you’ll be adopting the recommendations that aren’t best, and
that won’t be very good at all.
But presumably you would at some points be willing to trade off some amount of
intra-group honesty for extra-group reputation, because otherwise they’re not
going to make their thing happen.
It’s a matter of like who the audiences are that they should be pitching to,
and we just need a big enough audience to believe them. We don’t need
everybody to believe them. So for example imagine there are startup cities in
the world who are in the market for a new form of government. We want those
people especially to believe the honest recommendations about forms of
government, because they are especially people to take action on.
general, like, who do we need to believe doctors if doctors are telling the
truth? Well we need sick people that believe doctors. We don’t really need
well people to believe them if they aren’t taking any action based on what the
doctors say. So lots of ordinary well people not believing doctors is just
In this platonic dialogue called the Gorgias this orator named Gorgias says,
You know sometimes my brother who’s a doctor and I, we both go to the patient,
because I’m an orator. I’m much better at persuading the patient to take
painful medicine than my brother who’s a doctor. Right? So you know, it might
be the case that there is a trade-off. You have two values here, right? You
have the value of this group, an intellectually honest way of pursuing this
end, and then you want them to have a reputation among at least some people,
And so it’s that package that you really want, right?
So that highlights the importance of our institutions of advice: That is, our
institutions of advice are in some sense our most important, most meta
institutions, because all the other institutions are only going to be valuable
if they can be fed through the institution of advice. If they produce insight
but then nobody will believe them, or the important people who take action
don’t believe them, then they’re not very useful. So yes. And in fact I have
tried to center my attention on the institution of advice.
So that would be, then, the place where you would most want to both have—and
to have a reputation for—honesty, is in the institution of advice.
Yes. Absolutely. Even more than forms of government, right? If I can create an
institution of advice that’s reliable, then that could even create incentives
for people who, say, specialize in forms of government to actually be honest,
because the institution of advice will make them be honest.
But you might think that there’s going to be a problem in centering— your
attempting to center the honesty there, because if they ask, Who are the kinds
of people who want to give advice, right?
Not everyone does.
But I would say wanting to give advice that is actually then taken, right, is
sort of close to wanting to rule. Right? That is, you want the order as you
see it in your mind to be the order of how things are. That’s quite different
from wanting to know the truth about things. So like, it could be coincidental
that there is somebody who wants to know the truth, and he also wants the
order in his mind to inform the world. But those are two independent facts
that just could coincidentally come together.
Let’s just clearly define what we want, and I think it encompasses all these
effects. I mean, we want a good institution for advice. That good institution
for advice needs to therefore meet a number of criteria here. So it first of
all needs to give advice on the topics of advice desired. Whoever the
customers are for advice—if it’s advising them on shoestring color and they
don’t care about shoestring color, then that’s a failure of this institution
of advice—it should be advising people on the things they care about.
it should be capable of convincing the people who are there advising that it
in fact is good advice: accurate, high status, whatever else they want from
the advice. Next, given the kind of advice that customers want, and that they
want accurate, informed advice, it needs to go induce other people who might
be able to learn about those topics to go bother to do that effort and to
participate in the institution somehow, so that their advice is included. And
presumably there are many people who participate, so the advice then needs to
be merged and aggregated into some summary forms and then given to the
customers who want the advice.
So if there are people reluctant to give
advice, then this institution faces that problem. And that’s— A good
institution will be an institution that finds a way to overcome that problem.
And if there are people who are dishonest and wanting to influence the advice,
which there are because they have agendas, this institution needs to be
properly skeptical about those people. And again, that’s the problem faced by
the institution of advice.
And I mean presumably you would want the institution also to grow such people,
right? To cultivate…
Of course. You know, on the longer term we would want to train them. And we
want to induce the training of them, and the development of them, the
selection of them; and then how they allocate their time between learning in
general and learning about a specific problem, with learning and talking with
their colleagues to, you know— All of those things we want to induce, you
know, the right trade-offs in those things in order to produce the best
Do you think there’s any tension in trying to raise or grow people who are
both passionately motivated to seek the truth on the one hand, and equally,
independently, passionately motivated to make the world match an image of the
world in their mind?
Just think about all the other professions and all the other specialties that
people have in jobs. You know, there’s just a wide variety of the ideal person
for that job. And typically it doesn’t have that much to do with how committed
those people are to the overall purpose of the job. But this is a fact about
people and jobs.
Though this is a very important job, too.
Nevertheless it’s still true about a wide range of jobs. And I’m not even sure
I see a correlation between their sort of emotional attachment to the larger
goal of the job, and they’re being good at the job and fitting well in the
role that they are assigned in the industry or the profession they’re in. So…
So it’s almost really the institution that’s giving the advice, not the
Yes! Exactly. Precisely! That is the way we, you know, think about
institutions—the fundamental choice in the world, in some sense, a more
fundamental choice is institutions, because once you have institutions, they
last. They get stuck, as we talked about. It can be hard to change them. And
then people come and go, you know, relationships come and go, lots of other
things come and go. And that thing stays. And that’s the really high-leverage
That is, so many people, what they want out of politics, say, is
to get their person in charge, right? But their person won’t stay in charge
that long. And so you know, we people who do analysis of local institutions,
we say, No, the fundamental choice is the presidential system or a
parliamentary system, or there’s six-year terms or twelve-year terms. You
know, what are people are allowed to say when they advert— you know, they're
soliciting people to vote for them, those are sort of the institutional
features of the system that will last and have these bigger effects.
then individual politicians will adapt to the rules of the system. And you
know, getting your person in charge isn’t necessarily that big a deal. In a
sense, part of the system is, How do you know who is your person? And people
game that, basically, in order to win. And so, you know, the system is the
Okay, I think we should stop. But before we stop let me ask you: Do you feel
like this— Do you feel like we’ve solved this paradox? Or do you think there
was just no paradox to start with?
Well, quite often paradoxes result from some sort of discrete framing. And
then if you choose a more continuous framing, then you see a trade-off. And
that’s a common way we economists resolve paradoxes.
Good! That’s a good ending.
All right. Until we meet again!