by [anonymous]
1 min read27th Jun 201118 comments


Personal Blog

We've learned not to expect short inferential distances when explaining ideas we understand. We've also learned that leaping too far ahead when explaining ideas like transhumanism can freak people out.

I want to be really really good at explaining ideas. Does anyone have recommendations about how to figure out what the next inferential step is in another person's mind? 

Categories which are not answers themselves but are areas in which I expect to find answers:

  • Asking filter questions
  • Social contexts
  • Verbal cues
  • Body language
Ideally, I would like to find conversation-independent answers to this question; not just answers that fit for explaining transhumanism.


New Comment
18 comments, sorted by Click to highlight new comments since: Today at 5:03 PM

I don't have good tips for mind-reading in particular, but if you have a decent sense of how much the typical person already understands, these sub-skills I picked up in college should be able to help you, almost certainly not an exhaustive list:

  • Baby Steps - it's tempting to try to shoot straight for the thing you care about explaining. But if you try to describe it directly, even using very simple terms, you will be moving too fast even if the other person technically knows enough to eventually unpack what you're saying. If you do this your sentences will rot before they can be digested. Instead, spend at least a full sentence or two on each individual step from common knowledge toward your destination.

  • Stopping - this one is really hard, but after you've made each point, actually pause for a few seconds and look at the other person's facial expression and body language. Very few people are trained to explicitly ask with words, "can you unpack that a little?", but they will often say it with their face. This also gives them the chance to ask specific questions, or to process what you just said before you move onto the next thing. You can also ask explicitly whether you're being clear; wait for an answer before moving on.

  • One step forward, two steps back - be ready at any point to move back more than one step if it becomes evident that something hasn't been understood. If you really weren't understood the first time, then it's as if you didn't say it. If you find a different way to say it, it won't sound like you're repeating yourself; it will just sound like you're being clear.

  • Laconicism - if there is a detail that is not important, omit it. It doesn't matter if this means you're saying something that isn't strictly in every detail literally categorically true. (I still have trouble with this one.) All that matters is you move the other person as far as you can toward a true belief. Don't volunteer unimportant details unless they ask. (If they ask, then (a) telling the truth is the right thing to do, and (b) that means they understand the basics well enough to ask the question, which probably means they can handle the complicating detail.) There will always be aspects to a topic that interest you, and may well be important, that you should omit from a beginner's account.

Incidentally, I find Leonard Susskind is brilliant at all of these things. So, for a good example, his lectures on physics are well worth watching. Heck, they're worth watching even if you don't care about explaining things to people.

Wow, how did I miss this? This topic is one of big interest to me, since I seem to be much better at explaining things to others than getting a sufficient explanation out of others. It's a routine occurence for me to train people up to my level in a fraction of the time it took me to get to that level.

I strongly endorse fiddlemath's post, as it basically matches my approach, and I appreciate the link too. Thanks to jsalvati for mentioning my latest remarks on the matter.

So, here's my general strategy:

First, you have to want to explain it. (This doesn't seem like a problem in your case, but it's important in general.) It's tempting to maintain one's monopoly on knowledge and get divided between pursuing that, vs. actually trying to convey an understanding

Second, you need to have a deep understanding yourself. In this context, that means a Level 2 understanding, which means you can not only come up with the right answers in a domain, but have a model for that domain that deeply connects to other domains so that you can see how they're related, and what each implies for the other.

Third, you have to find the nearest point of common understanding ("nepocu"), which identifies the extent of your inferential distance. So, your filter questions should be aimed at identifying what common understanding you can draw from, so that you can take what you both know, and guide the listener stepwise to the parts that only you know. Build up every pre-requisite concept (sometimes several layers out), starting from this nepocu.

With all that in mind, here are some guidelines to follow, using a recent non-interactive explanation for examples:

Always be ready to "fall back a level" and explain the grounding concepts and prerequisites for what you're currently having trouble with. This is where the Level 2 understanding comes in: if your model for this domain is well-connected to your model for the rest of reality, there are arbitrarily many "inferential paths" you can take, and so you're always able to start from a domain closer to what the listener is already familiar with.

Note: for many people, the very idea that the body of human knowledge is mutually connected ("consilient") is novel, and their experience with the education system may have steered them away from even thinking like that, so they're used to looking at any topic as a bunch of random facts to remember. Connecting different areas will make things really "click".

Motivate each step in your explanation. An explanation is easier to follow if the listener knows why you're explaining it, and therefore what important things to look for. (This is why, in the linked blog post, I first explained what a signature must accomplish, before I describe the mechanics of public key signatures.) When explaining a process or method, it's confusing for someone to hear all the numerous steps, so it helps to start with, say, a naive, straightforward (but wrong) method and say something like, "But if you do it that way, it has this problem, so we do this instead" ... and gradually build up to the full thing.

Recall what it was like before you understood the topic and imagine what sort of things you wish someone would have told you. If something seems counterintuitive to a newcomer, acknowledge it from the beginning so they're not stuck wondering about it.

Hope that helps.

(An article I've been writing on this has been in development hell for a while now...)

Do you have tips for Recalling what it was like before you understood? I frequently notice that I don't know how to do that.

I super endorse Motivating each step, especially when it comes to math. I find I have a lot of trouble with advanced math textbooks that do not do this well (and that's most of them).

Do you have tips for Recalling what it was like before you understood?

Find a way to do the activity in a way that negates your previous knowledge and training.

For example, when I used to teach hooping, if I wanted to remember what it was like to try doing a move that you don't already "have", I would do it in my non-dominant direction. It would feel completely awkward, and I would catch myself making all the mistakes that first-timers to that move make, since I hadn't already trained it into muscle memory. Then when I taught a class, I would be prepared with what mistakes to look for, and already thought of ways to explain how to correct it.

As another example, in a Math Education course I took, they taught us how to do basic arithmetic in different bases (i.e. binary or hexxadecimal) in order to get rid of our intuitive understanding of those operations. That way, we could learn and explain it from a fundamental level, and we would remember how difficult it was at first, to learn.

Ooh! That's actually quite good. This might even make for a good post. Have you thought about writing one up? Even just putting this comment in a post would be good I think.

I wonder if it would be useful to do this with a bunch of skills that you've already mastered (handwriting with non dominant hand, etc.). It would be neat to make a list of ways to negate previous knowledge and training.

It would feel completely awkward, and I would catch myself making all the mistakes that first-timers to that move make, since I hadn't already trained it into muscle memory.

This is an awesome strategy!

One of the things I do to figure out how people can do stuff wrong (i.e. in swimming, which isn't something you can try doing in your non-dominant direction) is to break down the motion into tiny parts and do that tiny part while watching them, to figure out if that tiny part is the one they're getting wrong.

I also do a lot of trial and error, because sometimes someone's stroke will look intuitively wrong to me in a way I can't really explain to myself, so I make a guess, teach them how to correct that, and then watch again and see if my intuition is any happier with it.

Do you have tips for Recalling what it was like before you understood?

Good question. Unfortunately, it seems that for recall, it's a case of "you either do or you don't", and this is probably where I'm most unique. The best advice I can give is to use the "Remembrance of Things Past" trick: think about any memory related to that time in your life, and see if you can trigger an association that leads back to what you were thinking then.

I super endorse Motivating each step, especially when it comes to math. I find I have a lot of trouble with advanced math textbooks that do not do this well (and that's most of them).

Very true. A good example is the definition of a limit in calculus. If you try to just read the formal, standard definition, your brain will feel like mush, but if you learn what the definition is trying to accomplish, all the variables it introduces suddenly make sense.

I was a TA for a while and was taught to use Socratic methods. If people have some background knoweldge, you can get very far by asking people "what would happen in this situation? what about this situation? Now if you apply this situation back to this one, what happens?" It depends how much time you have, but lots of questioning interspersed with some explanations seemed to help most students the best.

If you are trying to teach someone, I think doing that is better than telling, because then when someone else has made the inferences it will become more a part of them and they will remember more. However, doing this repeatedly is also good for learning to explain, because you learn all the different things that people get hung up on---what inferential steps are most difficult to make.

Note that if you want to be really good, you'll have to do a LOT. After a year of 4 hours a week of face-to-face TAing, I was finally able to help people through most types of problems and had reached the 'competent/actually helpful to most people' stage.

Explaining things is also something where there are many paths to the same goal. Different explanations will work for different people. To be the best explainer, you'll have to learn many different explanations. Read many viewpoints on transhumanism by different people and as you read them, try to think of how you would explain it to someone else. Build up a repertoire.

Try to taboo all the relevant vocabulary so that you are forced to think about what the words and concepts mean in more precise language.

Interesting that it it takes so long. I am reminded of SilasBarta's claim that a lot of things that require lots of experience just require the right insights plus some practice.

It would be interesting to write down a bunch of student questions, and lists of questions from an experienced TA and a naive TA and then try to figure out how they differ.

The reason was probably that there was a large amount of material (the class included electricity, magnetism, circuits, and optics), so that I had to learn many different explanations. Each of these topics is a course by itself, and so and to explain things for an introductory course you often have to have a deeper level of knowledge. I hadn't taken any classes other than the intro class, so acquiring the deeper explanations was a long process.

I'll assume that, by "explain", you want to communicate a graph of inferences.

In fact, the way you've framed the question, you want to have a specific notional dependency DAG in mind. Which may be Step 0: Even before you're trying to explain an idea to somebody, break it down into its constituent ideas. Make sure you understand what ideas you expect to follow from what other ideas, all the way down to ideas that you're certain that either your audience already understands, or which you understand at a deep enough level to explain on the fly. (Probably a level 2 or level 3 understanding.)

I bet actually making this sort of diagram is a really good idea before trying to actually explain tricky things. In fact, I spend a fair amount of my time writing down explanations; I should actually try this.

Given that you have such a diagram, either in front of you or internalized, you can at least isolate one idea at a time. For each idea, things that you might try:

  1. Ask your interlocutor to explain specific parts; to put in their own words why a thing must be so, given the assumptions in play. If you worry that they might be guessing at passwords, ask them to explain those passwords, as well. Or:

  2. Ask why you don't get a slightly different result. When explaining a math theorem, you can ask for a counterexample to a stronger theorem or a slightly different theorem. For a physical phenomenon, you could ask instead what happens when the initial arrangement is varied. To test their understanding of an inference, you could ask what happens when the assumptions are a little different. This takes more thought and more time, but might also be instructive to both parties as a side effect.

  3. For certain kinds of ideas, you can ask your interlocutor to solve an example problem. Programming, math, and physics do this beautifully, but I'm straining just now to explain just what these have in common. In these cases, though, it's almost immediately obvious if your interlocutor understands the idea or not. If they understand, they'll start to think; if they don't, they'll start to panic.

I want to be really really good at explaining ideas.

Read textbooks on topics which you know extremely well, write Wikipedia articles, or best of all volunteer to teach math to underprivileged children.

Like many problems in self improvement I think this comes down to a two essential factors, Practice and Evaluation. Social activities like convincing people need extensive real world practice to effectively apply, even with the best advice. However, improvement can only be made if you know how effective you are actions are.

that much is obvious

As an initial suggestion: locate only groups on difficult but none controversial topics and practice convincing people of things, asking them to rate their knowledge, opinions and beliefs before and after

perhaps we could set up such a thing here on less wrong for people to practice on one another or does that seem to Dark arts to people?

perhaps we could set up such a thing here on less wrong for people to practice on one another

Yes, please.

does that seem to Dark arts to people?

If you go into it knowing that people are going to try to persuade you of things, then you can decide for yourself whether you're okay with the epistemic risk.

By "explain" do you mean "communicate" or "persuade"?

Everyone has seen cyborgs, androids, avatars and virtuality in the movies. That aspect of transhumanism has been communicated, as a set of possibilities. If you want to persuade people that they might want to commit tof that sort of thing themselves, then you have to understand them. Value-shifts always start with the values someone already actually has.

A nitpick: Seeing things in movies does not equate to knowing anything about them in real life. Conveying a real understanding of things requires the communication of large amounts of complex information. Communicating large amounts of information to someone in a way that allows them to grasp how that information informs your thought process is probably enough to establish a meaningful dialogue with them about the subject. The difficulty is in persuading them to listen to you intently for a long enough time to get enough information across to them to convey your point. If they are already super interested in what you have to say, this becomes relatively easy and you can assume the role of 'teacher'. If they are not terribly interested in what you have to say, then you have to be able to persuade them to listen to you.

If they think that they already know something because they've watched/read X, Y or Z, but they happen to be mistaken, then it can be very difficult to have a dialogue with them. If they have not thought about their epistemology, and do not practice epistemic cleanliness, then there is often no clear-cut way for approaching their mistaken beliefs without putting them on the defensive, so you have to build a rapport and then convey your message slowly over time; this is the essence of what lukeprog refers to as empathic metaethics.

So persuasion and communication are really inexorably linked in the communication of a complex topic, and persuasion in the sense that we wish to use here is just a means of effective presentation of information with the intent to hold the listener/reader's interest.

Try explaining it to yourself out loud as if you were 5 years old. You can always enlist the ears of a friend or family to play the 5 year old.