//Update: Zip version of my question using an analogy with physics:
In physics we have friction. Friction is a very complex process and creating a fully working model that works in all conditions is very complicated. But, in most cases, we can approximate friction with simple F=uN, and find "u" constant for common cases.
Natural language processing is very complicated. And creating a model from scratch for it is incredibly difficult. But we have already a working black box - our brain. Have someone tried to create a far simpler model, that only describes the relationship between text structure and "probability of understanding", and made some experiments on already existing brains to find "constants" for common cases? It can be useful to improve our communications, by finding wrong, nonworking patterns, for example. If someone has done it, I would like to have a link, or name of article/book, or smth like that.
I have a question - have someone heard about communication theory basing on the "filtering hypotheses" concept? I know that it sounds a bit like a mess, so let me tell a story to explain what I've meant.
One day I've been watching TED. There was a talk from one guy who was explaining how to draw simple comics characters. There were only a few lines in each picture, and after drawing every line, he was pausing.
I've created this animation to show you the process:
And because he was painting slow, I had time to analyze my thoughts. It was like:
*nose appears*- What is he doing? Nose can't be such big.
*eyes* - That 66 are even don't look like the eyes. What will he do next?
*ear* - That ear is too far. His head will be too big.
*hair* - Hm. Now it looks ok.
*mouth line* - Oh, he looks angry.
*line for smile*- And now he is fun.
I was excited about how fast my brain changes its mind about the picture on each step.
I also noticed that I was trying to predict what will happen with every next line appear on the frame. Like the brain has been trying to complete the picture. But there were too many possibilities of how it will end. For me, it looked like switching between future versions of images.
With thoughts like "It was an interesting observation" I closed the video.
The next day I've been trying to solve my PROBLEM. The problem has a simple description: I don't know how to communicate with people effectively. Previous days I decided to find someone who knows how to deal with it. I've been searching for an experienced specialist in this field.
It seemed like the person of that kind should, you know, communicate with people. I decided that it can be a psychotherapist or someone like that. At least, their work is to talk with you.
But there was another problem: I completely don't know how to choose a psychotherapist. Suddenly I found a cafe they meet up. I thought: "I can simply visit this place. And there I'll find people who can help me with my search!" I turned to my girlfriend (we talked about my problem that day) and said:
I can describe her face in one word:
It was clear that the inferential distance was too long. I've started giving her pieces of this puzzle. One by one, making little steps with pauses, to give her time to think. After a few hints, she said:
-So, you found their meeting place. It increases your chances of finding someone who will help you.
I did it. But this communication wasn't an effective one.
-Describe your thoughts - I asked.
And she said that after my "Eureka!" in the form of "Psychotherapists Cafe!", she was trying to understand what I meant. But there were too many possibilities of what I am trying to say. And she was switching between them, trying to choose the right one. And I thought that it sounds very familiar.
For a few days, I've been collecting observations about communications and trying to generalize them. What I've summarised:
- I receive a message.
- My brain makes some hypotheses about how to interpret this.
- Created hypotheses pass kind of filter. Part of them moves to the bin.
- I am "calculating interpretation results" for the rest of them.
- I am trying to associate interpretation results with other information in my brain.
- That produces new hypotheses and we go to step 3.
- I receive a second message. Now I can verify hypotheses previously created. Filter some of them and increase probabilities for others. And so on.
Step by step, one hypothesis becomes high-probable enough for converting to the working one.
And I've generated another idea, that seemed good to me:
If I can describe a process of communication for myself, maybe I can use this model for communicating with other people? If I know how the task looks like, I can choose the right data-structure for my message to optimize its processing. It's working for programming, why shouldn't it work for people? Maybe I can estimate how should I write text?
So I took my notebook and pen and started mapping my thoughts to math, drawing schemes, adding details: the capacity of working memory, patience resource, reward mechanism, etc.
I've been playing with it for a while. But then the other idea came to me:
I'll be a complete fool if I am reinventing the wheel. Or if my idea is completely crazy. I decided to check it.
I was googling for hours, but haven't succeeded. There were a lot of books and articles about filtering signals. I took some to read later. But none of them was not using signal processing theory to increase the effectiveness of human-human interaction.
I've tried to go another way. There should be theories of communication between people! I've found something named "Uncertainty reduction theory". But it wasn't "uncertainty" I've been looking for. I've checked a few more theories, none of them fit. Also, it hasn't seemed like people who wrote them were understanding how to explain something. Because, you know, their words were too abstract, and there were so many possibilities of what they could mean.
I failed my little research and ended up with three hypotheses:
- I have trouble verbalizing my thoughts, so I can't find a proper set of keywords for search. - Most probable one.
- My idea is dumb. - Less probable.
- I've made something new that worth researching. - I love that one. But let's be honest, there are 7 billion people around and much more lived previously. And they communicate. It's highly likely that someone invented an idea like mine.
I decided that if I can't find the right keywords to ask Google, at least, I can try to explain the idea to other people. Maybe someone will understand me, and he knows something related to my concept. Maybe someone wrote "Rationalist Guide to Communications With Other People", and I simply don't know about it?
So, I decided to share my thoughts. Then, I realized that there is one place where usage of words like "hypotheses" and "probability" is highly above average. And it also seems like people on this resource less likely will use arguments like: "You can't predict people behavior using math!", or "Communication is a subjective thing so it doesn't worth trying to describe it".
And here we are, and I want to ask you a few questions:
- Have you seen theory of communication that uses a similar idea to what I call "hypotheses filtering"? If so - please, leave a reference.
- Are there other communication models that work? I mean can be used for prediction, can be falsified and so on?
- In your opinion, is the idea of estimating text "readability" viable? I mean, if we have some model that approximately describes how does reading works, maybe we can use it to predict how likely our message will be interpreted correctly? And more, maybe it will help us to determine the wrong communication patterns?