...if caffeine were a new drug, I wonder what
Schedule it would be in and if people might
be even more leery of it than modafinil.
...
So I eventually got around to ordering another
thing of nicotine gum...
A second dose was similar, and the third dose
was at 10 PM before playing Ninja Gaiden II
seemed to stop the usual exhaustion I feel after
playing through a level or so.
Okay, I'm no longer part of the gwern fanclub. I could pick apart other aspects of this article, but I'll refrain, since it's not without a modicum of practical utility.
Yes, he writes very lengthy, detailed articles. Lovely. But they are anecdotes. And artisanal anecdotes at best.
To summarize gwern: A layman tries some stuff that is innately accessible to any layman, and writes haughtily aloof essays.
The article presents the following experimental premise:
“Linda is 31 years old, single, outspoken
and very bright. She majored in philosophy.
As a student, she was deeply concerned with
issues of discrimination and social justice,
and also participated in antinuclear
demonstrations.”
Then they asked the subjects which was more
probable:
(A) Linda is a bank teller
or
(B) Linda is a bank teller and
is active in the feminist
movement.
This example is bullshit, for several reasons.
We have several known personality attributes provided in the example, and all of them relate to the subjective opinions of an individual and are designed to provide belief in the potential for associations with similar political alignments.
Then we're provided with 2 choices, and each choice couples a previously unknown detail to the individual described, regarding occupation, with no opportunity to exclude the occupation.
We are asked to make an assumption about the individual, based on previously provided information, and guide the formation of our assumption with intuition.
According to The Letter Of The Law, the example asks the respondent to parse probability ONLY, and then penalizes according to the transitive property of equality, because technical interpretations of probability state that, when information has not been previously presented, an individual trait in isolation, is more probable, than a coupling of rare traits.
Which one is more probable? Oh wait, you're
wrong because you misinterpreted our
context-sensitive definition of the word
"probable." You lose.
According to The Spirit Of The Law, the example appears to present the respondent with a set of details, and prompt the respondent with a request to parse the details and demonstrate a display of reading comprehension, such that they show they've observed the relevant details, and drawn a conclusion by associating multiple cultural norms of common political alignment.
With the presentation of choice B, and
based on other personal details about
Linda, do you feel it likely that Linda
could be a feminist? Keep in mind that
we've offered clues to her political
alignment, and these may play a role
in the correct answer.
The example reads as:
Given: [0, A, 2, C, 4]
Is [E] likely?
or
Does [E, 6] make more sense?
But claims to present:
Given: [0, A, 2, C, 4]
Which is most probable?
> [$]
or
> [$, 99]
So, the example is an experiment in providing a loaded question, and then changing the context of expected interpretation, and then declaring proof that people are prone to misinterpretation.
The example is like asking someone if they were happy about who won The World Series, and then telling them you're not inviting them to a soccer game, because of their opinions on baseball.
The example the authors have provided is designed to promote assumptions, without providing adequate context for expectations, and that is fucking stupid.
Your reaction is exactly the reason the question was constructed that way. A lot of the available information would lead you to conclude that the provided details were important and that they support a specific option being more likely.
But if you rationally consider the options, it's apparent that option B can't possibly be more likely than option A, regardless of the information presented, because option B is by definition a subset of option A. It is not possible for Linda to fit option B but not option A, so option B can't possibly be more probable.
The fact that it's a leading question designed to promote assumptions is not a flaw; it's the whole point of the experiment. Even intelligent people are supposed to be led to the wrong conclusion because they try to analyze all the available information. But rational people are supposed to recognize that the presented information is irrelevant and that they can pick the right answer even if they don't know anything about Linda.
In the interest of full disclosure, I'll mention that I had exactly the same reaction regarding the quality of the question. It was only after some consideration that I realized this may have been intentional on the part of the people conducting the experiment.
Some amount of people probably assume that asking "What's more likely, A, or A and B" intends to ask for a comparison between A^~B and A^B, not simply A and A^B, in which case it would be an error in communication rather than an error in rationality.
The premise of the experiment is akin to considering whether or not people are prone to being swindled by a three card monte con-game.
The premise of the example only demonstrates a susceptibility to a situation where the individual is not expecting to be judged based on technicalities.
Technically, in a three card monte game, on the street, you have no assurance that the dealer is operating the deck with integrity. Technically, on the street, you have no assurance that other players are not collaborating with the dealer.
Does this prove that humans are often innately irrational? Maybe insofar as any other parlour trick does.
The claim that a bystander should know that their own capacity for estimation of the likelihood that Linda's occupation may be bank teller shall be poor, is masked, in terms of relevance to the rest of the context of the presented scenario. A bystander's estimation of feminist alignment is anticipated and intended.
When the bystander chooses option (B), an assertion that they have no insight into whether Linda was a banker or not, suddenly becomes the defining aspect of the test.
So now we've proven that given an unexpected context, a bystander is surprised by a sudden ambush within that context.
While such nuances may be interesting on a much grander scale, in most cases, the experiment is not framed that way, as a design to misdirect the individual, and certainly, the authors of this op-ed article make the same mistake in pointing at the idea that a bystander should be expected to know that they have no way of knowing whether or not Linda might be a bank teller.
Or alternatively, the existence of (b) means that (a) is generally understood to mean "Linda is a bank teller, and is NOT active in the feminist movement", regardless of how literally the instructions say to take it.
Either way, it's more a failure of communication between the experimenters and the subjects, rather than a failure of rationality on the part of the subjects. Which is still interesting (and perhaps more interesting), but in a completely different way.
Is it? I didn't interpret it that way, and that isn't what the question says. And why would it be "understood" that is what is being asked? Logically, the answers would be "she is a feminist" and "she is not a feminist", there would be no reason to construct the answers with the information that she is a bank teller if that is what was being asked.
This seems like another detail that rational thinking picks up- if you stop and think about the answer format the "generally understood" interpretation doesn't make any sense.
The answers given don't make sense together, and don't fit the question. I am speculating that the usual way (for people who don't prefer interacting with machines to interacting with people, which would be most people) to read this is to assume someone was careless/sloppy, and to mentally "correct" the answers to the nearest thing that makes the most sense before answering.
I agree that it's a very misleading experiment. People usually don't need to answer questions of pure dry mathematical probability. What most people answer here is a differently interpreted question:
"Which option, do you think, is probably more descriptive of Linda?"
Or "Do you think a bank teller usually has such a life history? How about a feminist bank teller?"
----
I hate it when experimenters ask about probability explicitly. It just test how much probability theory you've learned in school. How familiar you are with the mathematical framework. But it doesn't test everyday reasoning. Good tests for probabilistic thinking shouldn't even mention the word 'probability', but set up some physical or other task where you have to use probabilistic reasoning to solve it effectively.
For example the Wason selection task (https://en.wikipedia.org/wiki/Wason_selection_task) is also an artificial question that is hard to answer, but as soon as a practical story is built around it (alcohol and age), people can solve it much easier.
It's not that people are irrational, they are just bad at abstract thinking devoid of any practicality. We like to think in stories, situations, intentions. Abstract, "robotic" thinking is harder.
In other communities, to circumvent asinine ploys at lock-in, someone would just brute force the pdf file with a burner facebook login, and repost it. But then, you'd still be stuck with a pdf.
Yay! I feel reassured that everything is going smoothly, just as planned, and 10% growth as usual.
One could quickly argue that PR articles that skip the details, and make designs to produce the sensation of graphs that "go up and to the right," are a clear indicator that pleasing investors produces a bland gruel of profiteering corporations as ineffective at doing anything useful besides "making money," as is flatlander management, at providing direction to employees (read: millennials) who can't adjust to the sensation of not being subordinate to micro-managing overlords.
Dark patterns ask you to sit in a chair, without telling you that you can get paid to stand. (masking benefit without restricting access to the benefit; misdirection but without total erosion of trust)
This is about exploiting the exposure of unique identifiers (phone number mapped to email), and an interloping tattle-tale ratting out their correlation to the same owner.
It's something more akin to a Prisoner's Dilema, except people aren't cognizant that "They Are The Product" so no one thinks of themselves as prisoners ratting out conspirators.
He may be forum sliding, creating crufty noise, to dilute communication.