Oops, I actually misinterpreted one part—when you wrote “if placed under the microscope”, my brain interpreted this literally, as if you were talking about a hypothetical future version of “mind reading” that would include checking your neurons by a microscope and probably interpreting the results using an AI.
What I meant is that people usually think about these things in “yes or no” categories. For example, if you asked people whether existing lie detectors work, the most frequent answers would probably be “of course yes, why would they use them otherwise?” or “of course not, it’s total bunk”.
There I didn’t mean to make a statement about lie detectors per se, but about: this is how people think about technologies when you ask them. They think the correct answer is either “yes” or “no”, even if it is something complicated like “sometimes” or “yes, but with exceptions”. If the popular belief happens to be an unqualified “yes”, and you happen to be the exception, you are screwed.
I believe the current “mind-reading” techniques like Paul Ekman’s are hit and miss. That they probably often work with typical people in typical situations, but fail when something unusual happens. (Someone may be scratching their nose because they are lying, but sometimes the nose is just itchy for a completely unrelated reason. Or the person is lying, but in a different way than you assume. Or is just generally uncomfortable, maybe thinking “this is true, but they seem unlikely to believe me”.)
Practically, “films, television, and role models, books, music and lyrics that someone has absorbed” are an enormous amount of data, especially for people who consume a lot of these media. Maybe someone who reads one book in ten years and only watches the mainstream TV could be modeled this way. But if you asked me to give you a list of books I have read and the movies I have seen, I could probably remember only a small fraction of them. How specifically is Paul Ekman going to find out which movies I have downloaded from internet, watched in private, then deleted and forgot about them? Is he really going to read the LW Sequences, and Worm, and other extremely long texts written for tiny subcultures, just to get a better model of me? No way. -- However, with the help of an AI processing the tons of texts could become feasible. (The problem of figuring out who read what still remains.)
You probably could figure out some things, for example you could notice that people are, maybe unknowingly, repeating some sentences or argument structures from some book or movie heroes. Then it would make sense to assume that they somehow identify with the heroes and the values they represent. But even then, people sometimes adopt some values differently than they were originally meant. For example, someone could be a 99% Randian, but believe that Rand made a mistake about one specific topic. But maybe the topic is relevant to the thing the mind-reader is trying to figure out. Also it’s possible that someone was a Randian (sorry for repeating the same example) in the past, but is not anymore, but still keeps some verbal manners or something. In other words, you might figure out the sources that influenced the person… but not whether the person has adopted that perspective wholesale or just partially, and how their opinions developed over time.
This will probably work better for some kinds of people, and worse for others. For example, many people, if they join a movement, they adopt 100% of the movement’s beliefs, because they really strongly want to belong. But other people are just like “yeah, they make some really good points, but are mistaken about many things”. Modelling the latter would be much more difficult. I assume that being difficult to model will positively correlate with intelligence and autism. (As in: intelligent people have more complex and therefore less predictable models of the world; autists care less about being compatible with their group and therefore knowing which groups they belong to has smaller predictive power.)
Oops, I actually misinterpreted one part—when you wrote “if placed under the microscope”, my brain interpreted this literally, as if you were talking about a hypothetical future version of “mind reading” that would include checking your neurons by a microscope and probably interpreting the results using an AI.
What I meant is that people usually think about these things in “yes or no” categories. For example, if you asked people whether existing lie detectors work, the most frequent answers would probably be “of course yes, why would they use them otherwise?” or “of course not, it’s total bunk”.
There I didn’t mean to make a statement about lie detectors per se, but about: this is how people think about technologies when you ask them. They think the correct answer is either “yes” or “no”, even if it is something complicated like “sometimes” or “yes, but with exceptions”. If the popular belief happens to be an unqualified “yes”, and you happen to be the exception, you are screwed.
I believe the current “mind-reading” techniques like Paul Ekman’s are hit and miss. That they probably often work with typical people in typical situations, but fail when something unusual happens. (Someone may be scratching their nose because they are lying, but sometimes the nose is just itchy for a completely unrelated reason. Or the person is lying, but in a different way than you assume. Or is just generally uncomfortable, maybe thinking “this is true, but they seem unlikely to believe me”.)
Practically, “films, television, and role models, books, music and lyrics that someone has absorbed” are an enormous amount of data, especially for people who consume a lot of these media. Maybe someone who reads one book in ten years and only watches the mainstream TV could be modeled this way. But if you asked me to give you a list of books I have read and the movies I have seen, I could probably remember only a small fraction of them. How specifically is Paul Ekman going to find out which movies I have downloaded from internet, watched in private, then deleted and forgot about them? Is he really going to read the LW Sequences, and Worm, and other extremely long texts written for tiny subcultures, just to get a better model of me? No way. -- However, with the help of an AI processing the tons of texts could become feasible. (The problem of figuring out who read what still remains.)
You probably could figure out some things, for example you could notice that people are, maybe unknowingly, repeating some sentences or argument structures from some book or movie heroes. Then it would make sense to assume that they somehow identify with the heroes and the values they represent. But even then, people sometimes adopt some values differently than they were originally meant. For example, someone could be a 99% Randian, but believe that Rand made a mistake about one specific topic. But maybe the topic is relevant to the thing the mind-reader is trying to figure out. Also it’s possible that someone was a Randian (sorry for repeating the same example) in the past, but is not anymore, but still keeps some verbal manners or something. In other words, you might figure out the sources that influenced the person… but not whether the person has adopted that perspective wholesale or just partially, and how their opinions developed over time.
This will probably work better for some kinds of people, and worse for others. For example, many people, if they join a movement, they adopt 100% of the movement’s beliefs, because they really strongly want to belong. But other people are just like “yeah, they make some really good points, but are mistaken about many things”. Modelling the latter would be much more difficult. I assume that being difficult to model will positively correlate with intelligence and autism. (As in: intelligent people have more complex and therefore less predictable models of the world; autists care less about being compatible with their group and therefore knowing which groups they belong to has smaller predictive power.)