« (Moral) Truth in Fiction? | Main | Informers and Persuaders »

February 09, 2009


It depends somewhat who you are trying to reach. The young don't have so many social preconceptions. I suspect that Eliezer is going for the Doug Hofstadter position of engaging/inspiring to the academically young.

On the other hand he is aiming for the mathematically talented youth, who may have different tastes again.

One problem with your argument is that many recipients of "expert" arguments are in fact inclined to correlate their beliefs with what they perceive as the expert's belief, rather than scoring one's beliefs against the real world. See Eliezer on "guessing the teacher's password". This may well be a valid rationale for avoiding any form of persuasion in such contexts, if not keeping scientific arguments entirely secret.

I call BS. I think your choice of classification buckets is fundamentally flawed.

I'm taking the position that the meaning of a communication is its effect on the listener, and therefore interpret your buckets not just as buckets for effects, but buckets for communications - as it seems you do yourself, when you talk about "choosing what to read".

What you call "Persuasion" may simply be a collection of facts - i.e. Info - biased to support a particular position. Or to describe the other side of the coin, Info may actually be Persuasion - and in fact, if it's drawn up by a human, it is almost *always* Persuasion. The shadows cast by facts unrevealed create polarity in every missive that pretends to neutrality. I assert that the only Info that is free of bias is one which is Random within the domain.

Moreover, what you are trying to describe - communication free of artifice - is itself artifice of a more subtle kind, the one that presents an appearance of impartiality.

The staid presentation style comes from the belief that values are not objective, so if you are trying to impart only truth, you must leave all values out, and be a robot.

But when you are arguing about values to begin with (as has been happening here lately), it would be a contradiction to assume such a posture -- you would be saying with your arguments that "X is right," but with your posture that nothing is right.

"It reminds us that propaganda in some form or other lurks in every book, that every work of art has a meaning and a purpose — a political, social and religious purpose — and that our aesthetic judgements are always coloured by our prejudices and beliefs."

-- Orwell, Frontiers of Art and Propaganda

Will, the young mostly play the same game, even if they don't know it as well.

Barry and frelkins, I make no claim that human communication can be free of persuasion. But it can have more or less of it, and we do have signals of which case applies.

anonymos, yes not all readers of experts are experts or truth-seekers.

Ian, this all applies even to info that illuminates values.

@Ian C.

"trying to impart only truth, you must leave all values out, and be a robot"

This confuses me. Is not truth a value, and the urgent idea that it must be imparted also a value? Are not genuine truth-seekers often some of the most passionate (even if quiet - there can be a kind of passion that is so deep it does not roil the surface) non-robotic people?

"Ian, this all applies even to info that illuminates values."

Don't you see a potential problem in that though -- in saying "This (something) is good and right and true," and saying it like it robot? The audience will not believe you believe it, because if you did, you could not say it unemotionally.


I agree. Truth seekers can be very passionate, the idea that we're all Vulcans is an unfair stereotype.

I was not advocating for the unemotional style, but merely trying to identify the underlying belief that causes academics to use that style.


" I make no claim that human communication can be free of persuasion"

This is a subtle reply. But in offering it, it appears you make no distinction at all between "persuasion" and "propaganda." Is this your intent?

A logical argument can be persuasive. But it need not be propaganda, altho' clever propaganda will enlist both logic & the truth at certain points to heighten its seeming objective persuasion. I would not so conflate persuasion and propaganda, and this makes me ask you flat out, Robin, what you think propaganda is.

Robin, your post nicely negates itself. You have a list of things to avoid that you repeat three times. To illustrate your point you tell a story about having to write in a single column format. You "lament" this signaling equilibrium but claim not to use emotion. To top it off, you present your argument in a way that I would consider eloquent.

I think that you are so good at persuasion that you don't see yourself using it. If you removed all forms of persuasion from your writing, it would be nearly impossible to relate to. A good writer uses enough persuasion to maintain the attention of the audience with out manipulating the audience.

The most persuasive format is a clear logically argued document.
The status signalling that makes academese unreadable except under threat makes it
less persuasive because the message is hidden in the chaff.

And I second frelkins, that you seem to be conflating persuasion and propaganda.
Purely informative writing exists, but not much out of engineering and scientific
handbooks and textbooks.


"your post nicely negates itself"

This should not surprise fellow members of Hanson's cult. Robin's beautiful persuasiveness against persuasion is what one would expect from such! This is the irony that reveals wisdom.

frelkins and bill, I defined persuasion to be causing a correlation between beliefs and author-desired beliefs, over and above any correlation between beliefs and the world state, and I defined propaganda to be persuasion-heavy communication.

When I was in U of Florida's engineering college, they more or less taught us to write how Robin describes an engineer's writings. The class was called "technical writing", and had a prof who was especially despised (one of his favorite things seemed to be students using an exclamation mark; one use would cause the student to fail the course entirely). Of course they did not teach us to signal our own competence via overly-complex arguments, only to convey information accurately.

One problem seems to be that people are loath to read things they consider "boring". We can more easily teach a moral with an engaging story than with a published journal article on the philosophy of ethics. I suspect, but have no evidence, that this phenomenon is due to most people lacking the incentive to seriously delve into a subject. Engineers and many scientists have incentives to wade through arcane writings, while Joe-six-pack does not. Still, people must learn more from "funner" sources, or teachers and profs would not try so hard (well, some try hard) to make material fun and interesting. Is this a grey area between info and propaganda, or is there another (fun/boring) dimension not listed here?

The amount of technical information (or "tech", as its sometimes called online) in our culture is (to me) an alarmingly small portion of overall media. This wouldn't seem odd before the Internet, but nowadays primary, technical sources are only a mouse click away.

Robin: High status writings in ethical philosophy, whether by Peter Unger, Parfit, or whoever, typically include stories within the pages of academic articles or texts. The standards in a blog are surely somewhat more lax.

Most of what's in Robin's post seems fair enough to me - especially with Pearson's point added about appealing to mathematical youth (versus who else?) and Vassar's point about academicians of high status being able to use more informal arguments (think Richard Feynman).

I suspect that part of my informal style is just from having read many academicians of high status, rather than actually going through academia and being socialized by lower-status academicians. So to me illustrating Lob's Theorem with smiley-faces seems like a natural thing to do - why, I'm sure Raymond Smullyan would do the same - but if I was going to pursue an academic-status-based career, I wouldn't be able to get away with it without being Raymond Smullyan. I wouldn't seem SERIOUS.

The main point on which I seem to disagree with Robin is how much of the divergence between dull technical style and lively writing is in fact greater propaganda-nature on the part of writing, and how much is entirely signaling. Back in the world of professional writers, dull, monotonous-sounding language is usually considered a good way of hiding something. Consider Orwell's Politics and the English Language, and "I'm going to imprison Mr. Jennings for years without trial" versus "Unreliable elements were subjected to an alternative justice process."

It is not at all clear to me that a journal article full of overly complex math and passive-voice sentences, is less capable of persuading people apart from info, than a well-written short story. It's just a less dramatic form of noninformative persuasion.

Robin, I feel that you are incorrectly, picturing the global intellectual climate. The majority of people are not even concerned with being rational. They are concerned with being entertained. To even breach this apathy to reason, you have to start with an entertaining concept to draw interest. You cited religious displays and their characteristic language. Religions have a large corpus of, to their view, well thought out logic that the people those displays convince can then go to and parse if they are so inclined. Both styles are necessary, one to broaden and seed the platform of rationalists and the other to create correct, coherent models of behavior. Rationality has very few entertainers pushing it to the wider public.

I will not address whether persuasion should be used for the verification of systems of belief. That is the question Eliezer is talking about though, not whether persuasion will make him less respected. I don't think that matters to him.


Authors who want to be seen as minimizing the propaganda element of their communications avoid using flashy styles, eloquent language, or compelling stories, even when such things would make it easier for readers to assimilate the presented info.
I don't have an impression that academic papers are deliberately decolored. Where helpful, metaphors and examples are usually used, but overall the aim is clarity and speed of communication. The problem is that there usually is no point in adding embellishments, when facts can be stated plainly, or authors are bad writers and dry complexity comes from their inability to express the idea concisely.

Unless you have undergone a remarkable change in philosophy in the past few months, your statement that you "tend to avoid stories" is apparently unreliable. You and I had a back-and-forth about the story you used in this post, which I felt was gratuitous, detracted from the statistical evidence you were presenting, and was at best hearsay and at worst false. You defended the story and its use in your post vigorously. In a personal email to me, you claimed that, based on the comments, your readers found it persuasive. On my blog, you responded:

I fully acknowledge that stories told by friends can be in error. Do you claim no one should repeat such a story? I passed on a story from a personal contact I trust. This doesn’t guarantee its accuracy, but suggested to me it was more likely true than not, and it was relevant to my post.

Apparently, when you use stories (even uncorroborated ones) to persuade, you consider it efficacious. When others do so, you consider it an overstep in an effort to promote their beliefs. I think that you, as do most of us, use the variety of methods at your disposal to persuade, and defend those methods when called on it.

Grant, yes, tech writing teaches a non-nonsense style.

Michael, yes, academic ethics makes use of a few short stylized stories amid lots of analysis.

retired, I understand that you were not happy with my post of 16 months ago, but continuing to harp on it in comments on unrelated posts is inappropriate. I will delete any further such comments.

Eliezer, yes high status academics can signal that status by showing they are accepted even when they violate common standards. Style feature F could not be credible signal attribute A unless people with A actually found it easier to use F. I said some features, such as overly-complex math, seem plausibly signals of expertize, and some, such as two-column double-spacing, seem plausibly signals of conformity to arbitrary conventions. But I also gave a long list of features that seem to me plausible signals of info-over-persuasion reader effects. I agree it is a shame we don't have more research showing these connections, but it is hard for me to respond to your skepticism unless you say more specifically which features you think signal what instead.


"I defined propaganda to be persuasion-heavy communication"

Again, Robin, I worry that you are defining propaganda far too broadly here. By your definition, Euclid is propaganda - the entire book is filled with proofs to persuade you through logic that certain propositions are true. Surely you don't mean anything nearly this wide.

What concerns me here is that propaganda has clear definition, a political one, which places it in the realm of Orwell's concern. By diluting its meaning, you remove the sting of the term. Propaganda clearly is that which distorts facts and truth, through selective presentation. It often includes an emotional and graphic component, for a mass political end.

The point of propaganda is that it unfairly presents information with an intention to manipulate. It is possible to be highly rhetorically and logically persuasive without selectivity, unfairness, or manipulation.

frelkins, you forget that Robin "defined persuasion to be causing a correlation between beliefs and author-desired beliefs, over and above any correlation between beliefs and the world state". On this technical meaning, Euclid's proofs are not very "persuasive" at all. They're pure info.

On this technical meaning, Euclid's proofs are not very "persuasive" at all.

That can be confusing. I suggest using a term other than "persuasive" to denote something that Euclid's proofs are not.


All different (correct) proofs should lead to reader to the same (correct) belief that is in accordance with the world state. So you're saying that there's no room for persuasion.

But rhetorically, a proof is addressed to a skeptical reader who has no a priori reason to believe that a proposition is correct. In effect, it shoves the reader away from the right belief and then leads him/her ineluctably back to the right belief. So there can be varying types and magnitudes of persuasion. And this device seems to be helpful, because we can get away with making coarser distinctions and can use more available paths to the truth.

Robin's definition doesn't seem to be coherent over these two ways of understanding proof.

Anyways, it would be more correct to say that Euclid's proofs have minimal information beyond the geometric axioms and the allowed rules of deduction.

I'm open to using some other word than "persuasion" for communication effects on beliefs that correlate with author-desires, above and beyond correlating with the world. Nothing comes to my mind though.

Eliezer, yes high status academics can signal that status by showing they are accepted even when they violate common standards.

Robin, this seems a bit too cynical as a view of high-status academics. Yes Richard Feynman cared a lot more about his status than he wanted others to think. But I think it would be fair to say that Feynman genuinely cared about educating people, too. The world of high-status academics is not utterly dark and gloomy. Feynman using down-to-earth language and examples for physics problems - making himself much more accessible - reflects the fact that he's a better writer and can do that, not just signaling status.

Raymond Smullyan's books aren't just signaling that he's high-status enough to write logic puzzles in informal language. He's using informal language because he wants you to understand his logic puzzles, and he's high-status enough to get away with using informal language.

"Manipulation" seems closer to the idea in certain key ways. Also, "beguilement" and "seduction".

It's also interesting to consider what a medium (e.g., complex academic text vs. fiction) signals about the the author's apparent beliefs about readers.

In my first research internship some years ago, when I was asked to present my progress to the research group, I tried to present the material tutoring-style, with questions (ten-second exercises) so listeners could check their comprehension and could actually get fluent with the formalism I was using. I was (correctly) told that I shouldn't use such formats as they are insulting to the listeners. A tutorial style exposition implies that the speaker is in charge of the "correct" answers and is competent to structure the listeners' learning, while standard academic presentations present research to be looked over and commented on by listeners, with a presumption that the listeners will understand.

Perhaps some of the annoyance at Eliezer's use of fiction stems from readers' insult at the view such fiction signals of their abilities or of their non-peerhood with Eliezer (and not from concern about fiction's effects on readers). If readers can instead claim to be members of a blog filled with math and complex, passive-voice sentences, they can signal that they can read such material.

Part of Smullyan's brilliance in "The Tao is Silent" is that he makes clear that he is not taking himself too seriously, though he thinks his conclusions rigorously correct. He succintly states his own inclinations, and then engages the reader to discover for himself or offer up a counterargument. Feynman's writings share that quality. Some academics are more pedagogically gifted than others.

The variable persuasiveness is unnecessary (plus causing boring confusion). Presumably all communication is aimed at moving our beliefs toward what the author would have them be, to as much of a degree as the author can produce. Otherwise the author wouldn't bother. The question is whether the author would like us to believe true things or other things. I agree that communication style signals this.

A communicator can signal that they are persuading us of truth by actions like writing clearly partly because doing so makes critique easier. This doesn't mean that listeners are much more likely to reason for themselves, or should. It signals to listeners that the communicator is willing to put more at stake for their claims. This is same as other situations where information transparency allows common trust without everyone having to read the information.

Communication not aimed at truth *is* often more persuasive, but that's not due to differences in how much the author would like to persuade. Persuasive techniques undermine audiences' abilities to reason, as well as not including clear information. Thus to use them is to make critique harder, so what's at stake lower. Also persuasive techniques are more desirable to audiences who don't want to reason, so using them says you aren't expecting an audience interested in truth. My point is that persuasiveness is part of the signaling, not the feature beneath it.

Robin makes two explicit (and correct) points:
(1) people (ie EY) should signal
(2) people (ie RH) should read signals
but there is a subtext (pretty explicit in the comments, on other posts and here) that RH is penalizing EY for failing to signal. Since RH has a lot more information about EY than the signal, it is irrational to make much use of it. So it looks a lot less like the explicit theory of the local game and a lot more like cartel enforcement.

The way that signals start out as rational equilibria and calcify into hazing and arbitrary ways of limiting the size of the in-group is a very important fact about the world that I don't understand well. One theory is that people are unwilling to deal with multiple sources of information, particularly incomplete information, so simple tests that apply to everyone take over.

Best...Comments thread...Ever!

Eliezer, humans evolved tendencies to take the actions that signal high status and when possible to consciously believe that such actions were done for some other more noble purposes.

Anna, yes one is wise not to insult high status audiences.

Katja, informative authors often have little idea what conclusions readers will draw from info provided.

Douglas, my topic is not what I can infer about Eliezer from his total corpus of writings.

All, I'd think it more important to understand typical author and reader behavior than to focus so on a few outliers like Smullyan or Feynman.

Robin, yes of course it is wise not to insult high-status audiences. That wasn't my point; I was wondering if we should be paying attention to readers' signaling about themselves, and to authors' signaling about their beliefs about their readers, as we try to understand writing conventions. In particular, I was wondering if the desire to signal OB readers' (intelligence/seriousness/analytic skill/peerhood) is playing some role in the dislike of Eliezer's fiction.

Robin, informative authors at the very least want audience to conclude the info they are giving is true. How does whether they can guess audience members' next steps of inference from it change the analysis?

Robin, my understanding of the conventional view in evolutionary psychology is that we have evolved to signal certain attributes by actually having them - for example, to signal strong friendship by actually being strong friends with someone. The idea that our actual, evolutionary motives are lurking in our subconscious and actively steering things is not conventional evolutionary psychology.

The idea that our strong friendship might, for some odd reason, start to fade when our friend can no longer be of service to us, would be more conventional evolutionary psychology - but that's a new hidden circuit activating that the thinker doesn't know about; it's not a subconscious motivation that was at work all along.

Yes, we've evolved a general faculty to (a) in general, be influenced by 'unvirtuous' desires such as desire for high status; and (b) in general, put the best possible rosy cast on our actions. But to suppose that every trace of virtue is just status-seeking by trying to appear virtuous, is simply too cynical; it does not seem psychologically realistic.

I note that the rules may be different in economics than in evolutionary psychology. I can see why, in economics, you would try to put the burden of proof on someone claiming that a behavior is genuinely noble. But in evolutionary psychology, the burden of proof is on whoever claims the more complex adaptation. If there's a selection pressure to signal quality X, then one simple adaptation that does the trick will be to actually have quality X. If this seems on its face to be psychologically realistic for humans, then the default is to say, "This would seem to explain why we have quality X!", and maybe try to invent an experimental demonstration by reasoning about the ancestral conditions that should activate or maintain quality X. A more complicated conscious-plus-subconscious hypothesis that enabled even more sophisticated behavior, would bear the burden of further demonstration in a way that distinguished it from the simpler hypothesis. It would also prompt the question of why attention to the signal hadn't faded over evolutionary time.

Economic actors routinely invent quite complicated and intelligent plans on the spur of the moment; evolution tends to be relatively more parsimonious.

I'm expanding my comment into a post.

Eliezer, surely how friendly I act toward anyone one person at any one time is the result of far more than just some stable "is he my friend?" label stored somewhere in my brain. And surely whatever are the relevant internal states, many of the processes that change those states are influenced by indicators of how much it is in my genetic interest to act friendly to this sort of person in these sort of situations. If you agree with these claims, I don't see what you think you disagree with me about.

I read an interview with a teenager who made about a million dollars by buying lousy penny stocks, and then sending out spam emails saying something like


He experimented with it a lot, and said that though he
initially thought that using grammar and punctuation
would be more persuasive, he got better results using all caps.

This indicates that most people don't divide domains into
emotional ones in which persuasion is appropriate, and
informational ones in which reason is appropriate.
Rather, those domains Robin singles out as "persuasion"
domains are simply those domains in which more people operate,
and so they are tailored to a style that is persuasive to
most people - a style Robin calls propaganda.


Eliezer wrote: "The idea that our actual, evolutionary motives are lurking in our subconscious and actively steering things is not conventional evolutionary psychology."

AFAIK, this is not just conventional evolutionary psychology; it is the whole of conventional evolutionary psychology,
and of psychology in general. Psychology usually
proceeds as if there were no such thing as consciousness.

Phil Goetz,
could you link to (or otherwise cite) that interview?

@Robin, Eliezer:

I am enjoying this exchange. It recaps debates I'm always having with myself as a writer.

I think both of you would enjoy the book "Clear and Simple as the Truth", by Francis-Noel Thomas and Mark Turner. It is one of the best books I've read on writing. It is about standard English prose style. However, unlike most writing manuals, which just vaguely exhort you to be clear, Clear and Simple observes that standard style is a style of "disguised assertion" and that the appearance of clarity is always kind of illusion. It discusses the history of this style its unspoken assumptions about the structure of the subject mater and the nature of argument, the settings in which it is appropriate, etc.

It is also a short book. To me at least, this signals quality -- I suppose because it shows someone is not trying to signal quality by writing an unnecessarily long book!

Douglas - I think it was in Wired magazine a few years ago.

The comments to this entry are closed.

Less Wrong (sister site)

May 2009

Sun Mon Tue Wed Thu Fri Sat
          1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30