My Approach to Non-Literal Communication
This is a linkpost for this post on my blog. It’s primarily intended for my non-rationalist acquaintances, so regular LessWrong readers will likely be familiar with most of the concepts I mention. I’m cross-posting it here because there’s enough overlap in content that I think some LessWrongers might find it useful, and because I’d be interested in getting feedback on my approach to daily communication.
Imagine a society much like ours, except that people always speak literally. No implicature, no hyperbole.
When someone asks “does this dress make me look fat”, they get back a “yes” or “no”, and can use that information to decide on what to wear.
People in this society are still subject to normal human emotions, so hearing that they look bad in a dress makes many of the askers sad. Some of the answerers feel sympathetic and don’t want to be the cause of the asker’s distress. So they start saying “no, it doesn’t make you look fat”, even if that’s not true.
This behavior spreads and becomes normalized. Askers start asking the question not because they want to know the answer, but because they want to know how much the answerer cares about their feelings. Only a rude, uncaring person would make them feel bad by saying “you look fat in that dress”. So answering “no” is a signal that they care about the asker enough to figure out what they’re actually asking. The question has turned from one about the facts of the dress to one about the relationship between the speakers. It’s better than asking “do you care about my feelings?” directly, because that question has far too obvious an answer; no one would ever answer it wrong, so it’s not a credible signal of care.
One day, someone wants to know if a dress makes them look fat. They’re going to an important gala and want to make their best impression, so they need to know the real answer. They know that they can’t just ask “does this dress make me look fat?”, since those words actually mean “do you care about my feelings?”. So they include a request that the answerer answer honestly. The answerer does so, the asker figures out what to wear, and then has a great time at the gala.
Other people start doing this too whenever they want to know the real answer. Now there are two types of people: The ones who want validation and the ones who want the answer. But now the difference is evident; if one person asks “does this dress make me look fat?”, everyone else knows that they’re the first kind of person. This has all the same problems as explicitly asking “do you care about my feelings?”. So they start adding on the “be honest” request as well, and the cycle repeats.
Every time someone wants an honest answer, they have to be more explicit than whatever method of asking is currently accepted in that society. You get people saying stuff like “Ok, I know that people always say this and it’s not meant to be taken literally, but I promise I’m asking honestly. Please don’t just give me the ‘socially correct’ answer, I swear I’m not going to be offended.”, they get an honest answer, and then they get huffy about it. “Well I didn’t want you to be that honest.”
These self-sustaining loops of ever higher and higher abstraction erode the ability of people to usefully communicate. In our world this doesn’t continue infinitely, since there are other forces pushing back, but they can still go surprisingly high. The result is a spectrum of ways to ask, where the more elaborate ways have an increased probability of honesty. This forces extra effort onto people who want honesty, and if they answer they get back is the socially acceptable one, they can never be sure that it’s entirely true.
I frequently find myself wanting to ask people for honest feedback about something I did. But many other people also ask for honest feedback, and would get upset at the answer if it’s “too honest”. The feedback-givers have no good way to know what’s going to happen, so it’s in their best interest to only give positive feedback and avoid the chance of social repercussions. As a result I can’t trust the positive feedback I get to be a fair representation of what the other person thinks of my actions. Only for negative feedback can I be certain it’s heartfelt.
A similar problem occurs when I notice something someone else could have done better. I don’t want to force the feedback onto them, since that can easily lead to bad feeling. But if I ask “Hey, do you want some feedback about what just happened?”, they can’t say no without risking being seen as unreceptive or defensive. So I have to play convoluted social signaling games, where I incrementally edge closer and closer to the subject while giving them a chance to back out of the conversation in a way that allows them plausible deniability they aren’t trying to avoid feedback.
This dynamic is everywhere. If I ask someone a potentially-awkward question and say “it’s ok if you don’t want to answer”, they might be worried that my statement was a platitude and I actually would be offended if they don’t respond. Or if I offer to pay a friend to do me a favor, they might think that I don’t mean it and feel that they have to either do the favor for free or decline to help me at all. Etc.
This makes it really difficult to have a conversation about any topic that comes with strong social expectations, and accurately getting ideas across requires a lot of set-up work to dispel those expectations.
Humans naturally operate on many different levels of meaning, switching back and forth as the situation dictates. Often this is due to the previous described dynamics of needing to say something while having a fallback plan if it’s received differently from what you intended. Sometimes it’s also used for humor, as in a double entendre.
The vast majority of multi-level communication happens seamlessly. Context clues and shared prior information make it obvious which meaning is meant, and the recipient interprets it in the intended way.
But sometimes miscommunications happen. Someone says something that’s supposed to mean one thing, but it gets interpreted differently. Or people confirm to incorrectly-perceived social pressure and end up doing something they don’t want to do. (And it’s probably more often than you expect.) Consider how frequent it is for “obvious sarcasm” to be taken seriously. This can lead to friction between people who tend to operate on different communication levels.
For example, I attend a lot of professional conference presentations. If the presenter says something incorrect, I often speak up to correct them. My thought process is “I want everyone here to have accurate information”. (Education is, after all, the primary purpose of having the conference in the first place.) But sometimes this is interpreted as “Isaac is trying to undermine the speaker and make them look bad.” People assume that I must have some sinister underlying motive driving my behavior and I couldn’t possibly just care about preventing the spread of misinformation.
It can be extremely difficult to get people to understand that no, I actually meant what I said. Consider the following conversation I had with an acquaintance who’s in charge of a group of people:
Me: “Do you consider it inappropriate for people in your group to do [thing]? If I see someone in your group doing [thing], should I contact you about it?”
Them: “[Other group] does not allow their members to do [thing]. Why do you ask?”
Me: “I want to know what to tell members of your group about your stance on [thing] if they ask me.”
Them: “[Other group] has significant overlap with my group and their rules are very clear. Does that answer your question?”
Me: “Not really, no. I was already aware that [other group] doesn’t allow [thing], and that I can talk to their leaders about it. My question is about your position, whether I should be talking to you about [thing] happening among members of your group, and whether you plan to encourage your members to not do [thing].”
Them: “Let me ask you a question: Do you think that members of my group doing [thing] is a significant problem? I’m not aware of it happening in large numbers. If you think it’s a significant problem, I can talk to [other group]’s leaders and see if they want to take action about it.”
Me: “I do think it’s a significant problem, but that’s not very relevant to my question. I was not requesting that you take action about [thing], I was asking what your current stance on it is so that I know how to answer questions about it from your group’s members.”
Them: “I appreciate your willingness to bring up potential problems regarding [thing]. I will discuss this with [other group] next week.”
I was trying to find out exactly what I asked: What should I be telling people your policy is? But they seemed incapable of understanding that that really was what I wanted to know, and seemed to think I was trying to push them into changing their policy. We later had a meta-conversation about some of our previous conversations, and it went something like this:
Them: “You tend to speak very literally, and myself and other people tend to read more into what you’re saying than you meant for them to do.”
Me: “I agree. I’m working on figuring out how to avoid that, but in the mean time I think it would lead to the fewest miscommunications if you tried to remember that I probably just mean what I say and aren’t trying to communicate something wildly different via subtext.”
Them: “Ok, I’ll try to do that.”
***A little later, while discussing the previous conversation about [thing]:
Them: “Here’s why we don’t think [thing] is very much of a problem and aren’t taking action on it.”
Me: “Thanks for the explanation! I think that’s very reasonable. I just wanted to know what to tell people who asked me what to do when they see [thing] happening.”
Them: ***Seems to think this meant I didn’t understand why they aren’t taking action about [thing] and proceeds to explain to me several more times why they don’t think [thing] is a problem***
***A little later:
Them: “It took me several re-readings of our previous conversation to realize that you were speaking literally.”
Them: ***Proceeds to talk about how I was asking them to take action on [thing].***
Me: “No, that’s not what I was asking. I was asking for what your stance was on [thing].”
Them: “Well here’s why some people might interpret your statements differently.”
Me: “Yes I’m aware, we already went over that. You stated that I was trying to get you to take more action about [thing], and I was clarifying that the intention of my statements was not to communicate a request for action, but a request for information.”
Them: “Got it.”
Them: “So going back to our previous conversation where you were trying to get me to do more about [thing], I think it would be better if you had expressed that in this way...”
What I think is happening here is that this person’s day job is to play politics and gain favor with other people and other groups, and they’re very good at it. Those sorts of interpersonal games tend to include a lot of subtext, and as a result this acquaintance has adopted as their default that every statement has multiple buried layers of additional meaning, to the point where they were nearly incapable of taking things literally. This disconnect made it nearly impossible for the two of us to communicate to each other what we really meant.
And remember that this was a situation where we were both aware of the communication disconnect and legitimately trying to overcome it. Things get so much worse when the situation is adversarial and both parties are unconsciously looking for ways to conveniently misunderstand the other one.
For example, during the Joe Rogan-Spotify controversy, several of my acquaintances made claims like “the progress of science demands that we censor people who disagree with the current scientific consensus”. This made it pretty clear that they didn’t know what the scientific method actually is, so being someone who likes science, I made a Facebook post explaining why this made no sense. Several people decided that this meant I was a Joe Rogan fan. Rather than try and figure out the interpretation I intended, they took the most unfavorable interpretation and ran with it, since that’s the one that was most convenient for them.
(Sadly, it’s not just Twitter.)
One of them helpfully provided an elegant example of how to use multiple levels of communication to deliberately mislead people while maintaining plausible deniability that they weren’t trying to do that. They started messaging my acquaintances and saying “Isaac has been supporting arguments made by Joe Rogan”. This statement is, literally, true. But it strongly implies something like “Isaac supports most arguments made by Joe Rogan” or “Isaac supports particularly objectionable arguments made by Joe Rogan”. Listeners who aren’t looking for deception will likely interpret it that way. But if the sender gets accused of trying to mislead people, they can say “Woah woah woah, my statement was true! Isaac supports people drinking enough water, and Joe Rogan has also advocated for drinking water.” Or whatever.
I’ve also been accused of trying to “trap” people into contradicting themselves. This concept confused me for a long time; if they have inconsistent beliefs, how can that be my fault? But I think I’ve figured it out: This happens when we’re discussing a topic that they haven’t put much thought into, and they’re figuring out their beliefs as they go along. My approach to disagreeing with people is to get clear confirmation of where they stand on one thing before moving on to another thing. This approach removes their ability to silently retcon their beliefs if they realize that something they said previously leads to a conclusion other than the one they wanted to arrive at.
Restricting communication to only one level does not solve this problem. Multi-level communication is often used to communicate information to some listeners while hiding it from others, as in parody, innuendo, code words, dog whistles, euphemisms, weasel words, and Roman à clef. (Or even hiding it from the same person you’re trying to communicate with, as in the previous section.) This results in a coordination problem; if everyone always communicated literally then there would be far fewer miscommunications, and everyone would be better off. But that status quo would allow individual groups to gain an advantage over everyone else by switching back to multi-level communication.
Humans are creatures of habit. We tend to get into patterns of behavior and then continue applying those patterns even in undesired contexts.
For example, venting anger is often harmless in the moment. But it reinforces that your method of dealing with anger is “express it visibly in an uncontrolled manner”, and then you become more likely to resort to that behavior even when you need to be in control.
Communicating via deception and misdirection is subject to the same pressures. I used to communicate much more literally and directly than I do now. This didn’t make me a lot of friends. So I tried to learn how other people communicated and imitate it. But this started affecting my communication in ways I didn’t want it to.
One time, I was traveling to an event and arranged up front that someone would give me a place to stay the night before my 5 hour drive home. The event went great, and near the end, the person in question asked me if I still needed a place to stay. I instinctively said “oh no thanks, I’ll be fine”. A few seconds later I thought to myself “why on earth did I say that? That’s not remotely true.” Too embarrassed to go back and change my answer, I ended up sleeping in my car.
My approach to communication had shifted from “say things that are true and relevant” to “say things that will make other people react positively”. In many circumstances this leads to the desired outcome, but sometimes it fails spectacularly.
In another situation, I was talking with my partner about which of two personal policies we should institute. We were talking about whether policy A or policy B would lead to the least of outcome C. I thought of two reasonable-seeming arguments; one for why policy A would decrease outcome C, and one for why policy B would decrease it. But I only told her about the first argument, making it seem like policy A was the clear winner.
Again, my approach to communication had been corrupted by social norms. For reasons unrelated to outcome C, I would have preferred policy A to be the one we picked. So rather than answer the question truthfully, I gave the answer that was most likely to lead to the outcome I wanted. Once I realized what had happened the next day I had to go apologize for lying to her.
Not to mention how instinctive deception of others will make it harder to detect instinctive deception of yourself.
“To tell the truth is the same as to be a good tailor, or to be a good farmer, or to write beautifully. To be good at any activity requires practice: no matter how hard you try, you cannot do naturally what you have not done repeatedly. In order to get accustomed to speaking the truth, you should tell only the truth, even in the smallest of things.”
My current approach is to try to avoid falsehoods and misrepresentation unless the alternative is significantly worse. I’m still willing to selectively omit information and imply things that I don’t believe, but I look for any reasonable alternative first. A lot of social norms can in fact be satisfied without too much deception, just emphasis of some things over others.
I generally communicate more towards the literal end of the spectrum. I use sarcasm from time to time, I sometimes get annoyed and let that color my phrasing, and I keep an eye out for hidden meanings in other people’s statements. If I think I’ve found one, I’ll usually ask them for clarification rather than assume I know what they meant.
I prefer online communication when possible, since there’s less pressure to respond immediately. This gives me time to compose my thoughts and avoid the failure mode of saying something instinctively that I don’t actually mean. When forced to respond in-person I’m trying to get into the habit of saying “let me think about that” when necessary.
And I try to clearly signal that I mean things literally most of the time, like with this blog post. Many of my acquaintances have picked up on that and the rate of miscommunications with them has gone down.
I’m definitely not perfect at any of these. I often imply something or dodge a question without intending to do so, sometimes without even noticing afterwards. I find it very valuable when someone asks me “You said X. Does this also mean Y?” Often the answer is no, but sometimes it’s yes, and this lets me know that I failed at being appropriately clear and explicit, and need to adjust.
Possibly it’s also seen as shallow or needy, and disincentivized for that reason as well.
And it’s worse than this because other people exist. Even if both of us are ok with honesty, they risk someone else overhearing the communication, not knowing that I’m ok with it, and perceiving them as rude.
This dynamic and my willingness to defy it has led to me becoming “the feedback guy” in my professional social circle. Whenever someone wants in-depth, critical feedback about something they’re working on, they often come to me, even if we barely know each other.
This conversation is obviously paraphrased and seen through my lens of what information I wanted and how I perceived their responses. I’m sure they would have a different interpretation.
There also may have been an element of “I don’t want to answer the question and me repeatedly changing the topic is supposed to be a clear signal of that, I guess Isaac is just being really rude by continuing to hound me for an answer.”
For example it’s possible that my request to take me literally was interpreted as “I want to be able to use subtext around you without you ever acting on it.”
I’ve noticed that people who dislike me tend to have much less accurate understandings of what I mean whenever I say something. Of course I can’t say for sure that the causal chain goes in that direction; it could also be that they always had a harder time understanding non-figurative speech and those misunderstandings led to their dislike of me. Realistically it’s probably a combination of the two.
If any of them had bothered to ask me whether I supported Joe Rogan or had ever listened to a single episode of his podcast, they would have gotten a resounding “no”. What I actually believed was largely irrelevant to them; they saw an opportunity to discredit someone they disliked, so they took it.
When I first heard the phrase “Netflix and Chill”, I thought it meant what it seemed to on the surface; some friends getting together to chill out and watch Netflix. Luckily I didn’t have enough friends at the time for this to result in an awkward situation.
It’s also possible that I’ve always been this way and simply have gotten better at noticing when it happens.
I considered whether including this quote might dishonestly imply that I had actually read the book rather than found it in some forum post, and settled for including this footnote.
Though that article makes it out as being surprisingly less bad than I’d have expected.
Did you know that the Christian bible says you should never lie under any circumstances?
For example, if I have both positive and negative feedback about someone’s actions, if I spend more words describing the positive aspects than the negative aspects, they’ll come away from the interaction with a more positive impression, even if I accurately conveyed all my concerns.
One time a friend who lived far away invited me to come visit. I thought this was probably intended to convey that I could stay with them, but I wasn’t sure, so I asked. Turned out that was not a part of the offer, and it would have been really awkward for me to just show up at their door with a suitcase. Glad I asked!