The point isn't that people aren't capable of being rational. If they read the question carefully and think hard about it, of course they'll be able to figure it out. But that takes a lot of conscious effort.
The point is that people aren't often willing to put in that effort, so they default to intuition. What's more, they might not even realize they're doing this.
But there is a huge difference between 'intuition about conversation' and 'intuition about the state of reality'. The former is not tied to rationality as much as practice at fighting through obtuse language. The latter is much closer to telling you whether people are thinking rationally by default.
Compare https://news.ycombinator.com/item?id=12524799 this comment, where the kids already know which pile has more candy, but they don't understand what the weirdly-formatted question is asking.
Interesting experiment. Based on
thaumasiotes' description, I don't see any problem with the experimental design, only with the conclusion drawn from it. Something changed between the younger and older kids' ability to understand the question that was asked, and that in and of itself is interesting.
I object to the characterization of these experiments as "communicating badly" or relying on "obtuse language". The questions asked in both experiments are exceedingly straightforward and unambiguous. If "which is longer" and "which is more likely to be true" fall below the bar for designating a question as confusing, then so does every other question imaginable.
Conversation is an integral part of the "real world", and it's almost always far more ambiguous than examples we're examining. Yet we're still expected to parse and decipher conversation in order to survive in society: to thrive at your job, to vote for the best candidate, to get through school, etc.
It's in no way useful to design an experiment that attempts to avoid conversation in a world that runs on conversation. There exist people who can consistently pick the correct answer in all of these experiments, no matter how the question is worded. And there exist people who will get it wrong unless the question is worded perfectly. Who is more rational? Who is better at grasping "the state of reality", as you put it? Who would you rather have on your team at work?
There are interesting things you can ask about why people would understand or not understand a question.
But if they don't understand it, you can't conclude anything at all about whether they know the answer.
It doesn't matter if a question is formatted in a "straightforward" way. Nothing is very straightforward to a small child, and in normal conversation "A or A^B" is usually a mistake.
Honestly, people say things they don't mean all the time. A literal interpretation of conversation has a lot of potential to hurt you. It's also irrational. The rational goal is to figure out the most likely intended meaning(s). It's not to figure out what the meaning would be in a counterfactual world where people use language in a more logical way.
> There exist people who can consistently pick the correct answer in all of these experiments, no matter how the question is worded. And there exist people who will get it wrong unless the question is worded perfectly. Who is more rational?
If people figured out that the experimenter was asking a nonsensical and purely logical question, then that's a valuable skill. But it's not really connected to whether they are generally rational or irrational. It's also not connected to whether they understand basic probability.
> Who is better at grasping "the state of reality", as you put it?
In this experiment, you can't tell. When it comes to the probabilities about Linda, people's understanding is a total mystery if they didn't know what the question wanted.
You can figure out how they interpret questions on one axis, but you would need other tests to figure out why. They might be more rational, or they might be more literal, even to an irrational level.
I think we disagree on two big things, and the rest is fluff.
First, I still don't think it's fair to characterize the questions ("which is more likely" and "which has more M&Ms") as "nonsensical and purely logical." I wouldn't even call the questions atypical. We contend with simple questions like this all the time in society. The intent behind both questions is unambiguous, and there are no reasonable alternative interpretations, unless you assume the question asker is simply mistaken.
Second, we're on different pages about what the experiment is evaluating. I don't claim that the people who answer incorrectly are less capable of rationality, nor that they don't understand probability. What I believe is that they are less likely to recognize when it's best to apply their logical toolset. As a result, they're more likely to use intuition to answer questions that are best answered via careful analysis. This appears exactly like failing to understand the question itself.
To wit: The younger kids used the length of the line to answer "which has more M&Ms", but the older kids immediately recognized that it's better to use a more in-depth analytical tool: counting. I hypothesize that the younger kids are simply not as good (yet) at knowing when to apply this tool. Intuition is easier and less effortful, thus it's the default unless we make ourselves think harder. This pattern extends to adults, too. How many people think we should be tougher on crime because it "feels" like there is more violence today than ever before, yet don't even consider that they need to look at some actual numbers to justify this conclusion?
It's specifically a question of "which is more likely, A or A^B" that is weird. The typical "which is more likely" has non-overlapping categories. It doesn't even need to be a mistake. If I ask whether you want a sandwich or a burger, it's obvious that 'sandwich' actually means 'non-burger sandwich'.
I'm claiming that they are not using intuition instead of analysis. Or at least, you can't tell that from their answers. There are logical reasons to interpret the question as non-overlapping sets. They could be performing a very careful analysis and still pick the second option.
I only disagree with your last sentence. There's no way to perform a careful analysis of option A vs option A+B while maintaining the assumption that they are non-overlapping sets. Simply reading the answers proves that assumption wrong. Thus, it seems way more likely that the explanation is a lack of thinking/analysis... i.e. intuition. People are answering the question they expect instead of the question that's in front of them, because that's easier: http://lesswrong.com/lw/9l3/the_substitution_principle
But you are right, we can't know that from their answers alone. There are numerous possible reasons for picking the wrong answer, so further experimentation is required if we really want to know why.
The point is that people aren't often willing to put in that effort, so they default to intuition. What's more, they might not even realize they're doing this.