Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Illusion of Transparency: Why No One Understands You

46 Post author: Eliezer_Yudkowsky 20 October 2007 11:49PM

In hindsight bias, people who know the outcome of a situation believe the outcome should have been easy to predict in advance.  Knowing the outcome, we reinterpret the situation in light of that outcome.  Even when warned, we can't de-interpret to empathize with someone who doesn't know what we know.

Closely related is the illusion of transparency:  We always know what we mean by our words, and so we expect others to know it too.  Reading our own writing, the intended interpretation falls easily into place, guided by our knowledge of what we really meant.  It's hard to empathize with someone who must interpret blindly, guided only by the words.

June recommends a restaurant to Mark; Mark dines there and discovers (a) unimpressive food and mediocre service (b) delicious food and impeccable service.  Then Mark leaves the following message on June's answering machine:  "June, I just finished dinner at the restaurant you recommended, and I must say, it was marvelous, just marvelous."  Keysar (1994) presented a group of subjects with scenario (a), and 59% thought that Mark's message was sarcastic and that Jane would perceive the sarcasm.  Among other subjects, told scenario (b), only 3% thought that Jane would perceive Mark's message as sarcastic.  Keysar and Barr (2002) seem to indicate that an actual voice message was played back to the subjects.

Keysar (1998) showed that if subjects were told that the restaurant was horrible but that Mark wanted to conceal his response, they believed June would not perceive sarcasm in the (same) message:

"They were just as likely to predict that she would perceive sarcasm when he attempted to conceal his negative experience as when he had a positive experience and was truly sincere.  So participants took Mark's communicative intention as transparent.  It was as if they assumed that June would perceive whatever intention Mark wanted her to perceive."

(The above wording is from Keysar and Barr 2002.)

"The goose hangs high" is an archaic English idiom which has passed out of use in modern language. Keysar and Bly (1995) told one group of subjects that "the goose hangs high" meant that the future looks good; another group of subjects learned that "the goose hangs high" meant the future looks gloomy. Subjects were then asked which of these two meanings an uninformed listener would be more likely to attribute to the idiom.  Each group thought that listeners would perceive the meaning presented as "standard".

(Other idioms tested included "come the uncle over someone", "to go by the board", and "to lay out in lavender".  Ah, English, such a lovely language.)

Keysar and Henly (2002)  tested the calibration of speakers:  Would speakers underestimate, overestimate, or correctly estimate how often listeners understood them?  Speakers were given ambiguous sentences ("The man is chasing a woman on a bicycle") and disambiguating pictures (a man running after a cycling woman), then asked the speakers to utter the words in front of addressees, then asked speakers to estimate how many addressees understood the intended meaning.  Speakers thought that they were understood in 72% of cases and were actually understood in 61% of cases.  When addressees did not understand, speakers thought they did in 46% of cases; when addressees did understand, speakers thought they did not in only 12% of cases.

Additional subjects who overheard the explanation showed no such bias, expecting listeners to understand in only 56% of cases.

As Keysar and Barr (2002) note, two days before Germany's attack on Poland, Chamberlain sent a letter intended to make it clear that Britain would fight if any invasion occurred.  The letter, phrased in polite dipliomatese, was heard by Hitler as conciliatory - and the tanks rolled.

Be not too quick to blame those who misunderstand your perfectly clear sentences, spoken or written.  Chances are, your words are more ambiguous than you think.


Keysar, B. (1994). The illusory transparency of intention: Linguistic perspective taking in text. Cognitive Psychology, 26, 165-208. 

Keysar, B. (1998). Language users as problem solvers: Just what ambiguity problem do they solve? In S. R. Fussell and R. J. Kreuz (Eds.), Social and Cognitive Psychological Approaches to Interpersonal Communication, 175-200. Hillsdale, NJ: Erlbaum.

Keysar, B., & Barr, D. J. (2002). Self anchoring in conversation: Why language users do not do what they "should". (2002). In T. Gilovich, D. W. Griffin, & D. Kahneman (Eds.), Heuristics and biases: The psychology of intuitive judgment. (pp. 150-166). Cambridge University Press.

Keysar, B., & Bly, B. (1995.) Intuitions of the transparency of idioms: Can one keep a secret by spilling the beans? Journal of Memory and Language, 34, pp. 89-109.

Keysar, B., & Henly, A. S. (2002). Speakers’ overestimation of their effectiveness. Psychological Science, 13, 207-212.

Comments (46)

Sort By: Old
Comment author: chesh 21 October 2007 12:05:32AM 4 points [-]

Google found the actual meanings of "the goose hangs high" (the future looks good), and the reason it means that (there is a goose in the larder, therefor you will not starve), and "to go by the board" (to be completely destroyed or overthrown), but not the reason behind it. What do "to come an uncle over someone" and "to lay out in lavender" mean?

Comment author: gwern 27 October 2010 08:08:14PM *  2 points [-]
Comment author: UnclGhost 17 January 2011 07:05:05PM 1 point [-]

To "come the uncle over someone" means, according to this highly-trustworthy-looking site, to "overdo your privilege of reproving or castigating" someone.

Comment author: pure-awesome 31 May 2013 10:19:16PM 7 points [-]

As an example of illusion of transparency: On first reading, I interpretred your phrase 'highly-trustworthy-looking site' as sarcastic. Since it's a Webster's site, I'm going to guess that you were not intending to be sarcastic?

Comment author: UnclGhost 04 June 2013 12:04:42AM 2 points [-]

I think I did mean to be sarcastic, since it doesn't seem to be actually affiliated with the publishers of Webster's dictionary and the design of the site looks generally sketchy, but coming back to my comment now, you make a good point.

Comment author: pure-awesome 04 June 2013 09:45:55AM *  2 points [-]

Ok, thanks for clarifying. It actually makes a lot more sense for you to be sarcastic and I read it that way at first. I only got confused once I started considering the non-sarcastic possibility.

Comment author: arundelo 04 June 2013 01:27:11AM 5 points [-]

The name Webster's Dictionary may refer to any of the line of dictionaries first developed by Noah Webster in the early 19th century, and also to numerous unrelated dictionaries that added Webster's name just to share his prestige.

-- Wikipedia

Comment author: pure-awesome 04 June 2013 09:45:45AM 0 points [-]

Ah, good to know.

Comment author: Doug_S. 21 October 2007 12:18:26AM 4 points [-]

I've heard things like this before. This is one reason why it's a good idea to run your writing by someone else before you "publish"; you can get a better idea of whether what you wrote actually says what you wanted it to.

Comment author: Bruno_Coelho 04 June 2013 11:35:26AM 1 point [-]

This is academic habit, but vulnerable to group bias. Normally, you don't send drafts to experts who strong disagree with your statements, but close friends who wants to read what you write.

Comment author: Acad_Ronin 21 October 2007 03:27:27AM 1 point [-]

1) That's why Deidre (then Donald) McCloskey wrote in (I believe), "The Rhetoric of Economics" (JEL 1983), that, "If the reader says something is unclear, it is unclear." 2) The recent Rush Limbaugh imbroglio is consistent with the point about inferring meaning. The political leanings of the commentors guided their inerpretation of his words.

Comment author: douglas 21 October 2007 04:36:07AM 1 point [-]

Humor seems the most difficult. Once I taught a guy from Sweden English. One day he read an article in the newspaper to me and could tell me what it meant-- no problem. When we got to the comics page... the humor didn't translate at all. I still don't know why the difference was so marked.

Comment author: MrHen 09 February 2010 12:00:52AM 11 points [-]

Humor probably has a higher culture per word weight. (I made up this term.)

Comment author: bgrah449 09 February 2010 12:13:02AM 3 points [-]

It's because a key component of humor is someone's status being lowered, and someone just learning the culture won't be fluent in the status signals yet.

Comment author: Luke_A_Somers 18 October 2011 04:04:16AM 5 points [-]

… sometimes. This seems like a vastly overly broad requirement for humor. First off, there are puns. Some people don't find them to be humor, but I don't have to look far to find a non-pun exception - I fail to see a status lowering in today's Square Root of Minus Garfield, for instance, and even if it's not a masterpiece, it's somewhat funny.

I could find many other examples which only incidentally involved social status changes - Today's xkcd is a put-down, but the relevant status (dominance of chemical elements over Greek elements) is so drastic that no appreciable further change is going to occur. Furthermore, the same joke was made much better in Order of the Stick while eliminating the put-down element.

So it seems to me that saying it's 'a key component' had better not be along the lines of 'a key ingredient', but rather more along the lines of 'a principal component'.

Comment author: taryneast 28 December 2010 10:32:42PM 1 point [-]

Yes, and a heavier reliance on nuances and multiple-meanings of language that come with much greater familiarity.

Comment author: Jack 09 February 2010 12:12:17AM 16 points [-]

In fairness, newpaper comics rarely have humor to translate.

Comment author: Robin_Hanson2 21 October 2007 08:51:48AM 4 points [-]

Certain kinds of academics use this to justify a rhetorical technique: they simply say they do not understand what you say, and they keep repeating this for any statements that do not fit with their standard method. For example, in a community uses formal statistical techniques, they say they cannot understand any argument for a probabilistic conclusion that is not expressed as a formal statistical test.

Comment author: Scholar 21 October 2007 01:33:32PM 4 points [-]

Robin, could you give a bit more detail?

Comment author: Noumenon 22 October 2007 03:53:08AM 2 points [-]

I'd have to hear the voice mail message to believe what you're saying. In my daily life, when one of my friends says, "That's great. Just great" it's totally clear whether he's being sarcastic or not. Because he intends to sound sarcastic, and changes his inflection.

Now in this case the message was the same, so the inflection didn't carry any content. But the subjects were told Mark was leaving a message for his friend. So it's a reasonable assumption that Mark is using an inflection that his friend will recognize as sarcasm. I would assume that unless, as in the second case, they told me he was trying not to show sarcasm.

Comment author: army1987 12 April 2012 03:53:23PM *  3 points [-]

http://languagelog.ldc.upenn.edu/nll/?p=2868

(Yes, I know I'm replying to a 2007 comment with a link to a 2010 blog post.)

Comment author: thomblake 12 April 2012 04:04:32PM 4 points [-]

Yes, I know I'm replying to a 2007 comment with a link to a 2010 blog post.

Totally reasonable. Unless the tone was "you'd know this if you'd only bothered to read the 2010 post first".

Which I observed once as a comment to a post by Eliezer linking to a much-later post by Eliezer.

Comment author: TheOtherDave 12 April 2012 04:23:19PM 8 points [-]

This only seems unreasonable to us because we haven't yet read the 2015 post about acausal reference.

Comment author: thomblake 12 April 2012 04:27:53PM 1 point [-]

Well then that commentor should have linked to that 2015 post.

Comment author: TheOtherDave 12 April 2012 04:49:55PM 5 points [-]

Sadly, the AHTML markup syntax won't be available until 2017.

Comment author: dlthomas 12 April 2012 05:17:39PM 1 point [-]

Hmm, with the proper content-based indexing...

Comment author: Eliezer_Yudkowsky 22 October 2007 04:19:15AM 8 points [-]

Noumenon, one of the experiments (I think Keysar 1994) showed that, when people read in text that Mark left a voice message (that is, the experimental subjects themselves did not hear a spoken message, only read about Mark leaving one); or alternatively read in text that Mark left a written note; then they were equally likely to judge that June would identify sarcasm. In other words, the modality of text versus speech made no difference to their expectations.

Comment author: Noumenon 22 October 2007 07:03:55AM 0 points [-]

Okay, then, that's just weird. To me there's obviously not enough information in a text message to tell sarcasm that way. Even a voice message wouldn't be enough for someone who didn't know Mark, like the restaurant manager.

I can't really imagine myself in this scenario any more. My takeaway is that some people are dumb about text messages (or were in 1994 before the smiley).

Comment author: Nick_Tarleton 22 October 2007 01:37:15PM 1 point [-]

If they were told Mark left a voice message, it would be rational to assume June would identify sarcasm, because it would be rational to assume that Mark left it in a sarcastic tone (or not) if he intended sarcasm (or not). Though the written note case shows that's not what's really going on, unfortunately.

Comment author: Noumenon 23 October 2007 05:35:15PM 2 points [-]

I believe the reason for saying the opposite of what you mean is so that you and your friend can say, "We're clever. Others would not perceive our nuanced tone and would stupidly think we meant what we said." In the bush, don't people sometimes make the sound for "snake" when there is no snake, just so they can laugh and mock the person who foolishly believed there was a snake?

Comment author: Whit 23 October 2007 06:21:55PM 2 points [-]

Chesh, lay out in lavender means "displayed in the best possible light" according to http://www.phrases.org.uk/meanings/222725.html. If you google "out in lavender" or "laid out in lavender", you get more hits.

Comment author: Marty2 23 October 2007 07:58:08PM -1 points [-]

Dear Whit, yes,when there is no real danger like at a party the youngsters may play everyone knowns it is harmless. Never in the bush this would be very bad manners to frighten for a laugh beside what pleasure is gained by such unkind act to me that is a foolish game and being mean is not showing cleverness. Marty

Comment author: TGGP4 24 October 2007 01:46:30AM 0 points [-]

Marty, where specifically are you from?

Comment author: Mark2 24 October 2007 04:30:40AM 7 points [-]

It should have been perfectly obvious to Julie that I was being (a) sarcastic, (b) sincere.

Comment author: Noumenon 24 October 2007 08:18:40AM 2 points [-]

Talk about clever -- Mark illustrates this entire post in one sentence! He points out all these readings:

It should have been perfectly obvious to Julie that I was being sarcastic. (sarcastic) It should have been perfectly obvious to Julie that I was being sarcastic. (sincere) It should have been perfectly obvious to Julie that I was being sincere. (sarcastic) It should have been perfectly obvious to Julie that I was being sincere. (sincere)

and the point of the post is that you really can't tell from the text which way he means it or how he really felt.

Comment author: Cynical_prof 09 November 2007 07:31:13PM 2 points [-]

From Dictionary of Phrase and Fable:

Don’t come the uncle over me. Do not overdo your privilege of reproving or castigating me.

Phrasefinder says,

Laid out in lavender means prepared for burial. The phrase has also been used to mean 'show something in the best possible light'.

Comment author: gwern 27 October 2010 08:02:21PM 0 points [-]

The City in Slang: New York Life and Popular Speech by Irving Lewis Allen says that to lay out in lavender also means to pawn some good: http://books.google.com/books?id=j41z0yeKbeIC&pg=PA159&lpg=PA159

Comment author: HonoreDB 17 February 2011 04:14:03AM *  4 points [-]

The Mark/June experiments seem flawed in that Mark and June have a pre-existing relationship that the subjects aren't privy to. The 1994 paper you link says repeatedly that information not available to June shouldn't affect our expectation of June's interpretation of Mark's message. But that's bad Bayesianism. The other examples seem unproblematic in that we know that the subjects were being overconfident.

(below lies belaboring of this point)

Suppose we re-did the experiment, word for word, except the message went as follows:

"June, i akabad de junt lo restaurante ou você rekomendad, e devad dicher ou flo derigimedero, simplesima derigimedero."

The correct assumption to make is that Mark is communicating successfully, since we have no other usable information. I would assign low probability that Mark is accidentally miscommunicating, and very low probability that June doesn't speak whatever language that is.

Sarcasm between friends can be a language unto itself. My brother and I each have a vocal tone that sounds sincere to everyone else, but sarcastic to each other. I had a similar tone, my "lying voice," that I used to signal to my college girlfriend that I was lying but that no casual acquaintance knew the meaning of. Similarly, the choice and repetition of the word marvelous signals sincerity to some pairs, signals sarcasm to others, and is ambiguous to others.

As near as I can tell from the 1994 paper, Keysar is simply missing the evidential chain: Mark's experience at the restaurant is evidence of his intention. Mark's intention plus his note or voicemail together constitute evidence of Mark's model of June. Mark's model of June is evidence of what June is actually like, and what June is actually like is evidence of how she will interpret the note.

The paper does edge tantalizingly close to acknowledging this when talking about Grice, but it still insists on calling this entirely reasonable kind of deduction an illusion.

Comment author: TheOtherDave 17 February 2011 04:11:16PM 0 points [-]

A complete digression: it delights me that although I have no idea what language that is, I am fairly sure I understand every word in it except "derigimedoro," which can fairly easily be classified by context. (Or, well, it could be given tone of voice, anyway.)

Which leads me to suspect it is probably Portuguese, with which I often have that relationship, although probably not Brazilian Portuguese, which I have slightly more familiarity with.

Comment author: rampal 23 February 2011 07:47:05PM 0 points [-]

I think it would be "de regime de ouro" or in other words, a golden diet...

Comment author: TheOtherDave 23 February 2011 08:15:04PM 0 points [-]

That's a marvelous phrase.

Comment author: Peterdjones 01 November 2011 08:57:59PM 2 points [-]

"go by the board" is still in use where I come from

Comment author: gwern 20 January 2012 09:13:38PM 3 points [-]

Then Mark leaves the following message on June's answering machine

A recent version of this for email: http://faculty.chicagobooth.edu/nicholas.epley/Krugeretal05.pdf

Without the benefit of paralinguistic cues such as gesture, emphasis, and intonation, it can be difficult to convey emotion and tone over electronic mail (e-mail). Five experiments suggest that this limitation is often underappreciated, such that people tend to believe that they can communicate over e-mail more effectively than they actually can. Studies 4 and 5 further suggest that this overconfidence is born of egocentrism, the inherent difficulty of detaching oneself from one’s own perspective when evaluating the perspective of someone else. Because e-mail communicators “hear” a statement differently depending on whether they intend to be, say, sarcastic or funny, it can be difficult to appreciate that their electronic audience may not.

Comment author: NickiH 02 February 2012 06:52:49PM 3 points [-]

This is an interesting read. Specifically, their work suggests what could be a potentially very useful way of reducing miscommunication.

One of the experiments the authors ran tried to reduce the overconfidence they saw in predicting whether people would understand or not. They asked people to write sarcastic sentances, and then read them back, out loud, in a tone of voice which made them sound completely serious. They also did serious read in a sarcastic way. They found that people were then less confident that the email would be understood in the way it was intended, because they had changed the way they "heard" it in their head.

I propose that this could be useful in the following way: if you write an email, read it aloud to yourself in the opposite tone of voice. If you are still confident that it will be taken the way you originally thought it should, it's probably safe to send. But if you can now see how it might be misunderstood, redraft it. Repeat until you feel ready to send.

There are times in my past when this advice would have been very useful to me.

Comment author: Kevin92 12 May 2013 02:57:26AM 1 point [-]

I believe this article illustrates the greatest cause of conflict on the internet.