Hacker News Reader

Comment by emeril | Hacker News Reader

from the article: "Then the students were asked which was more likely: that Linda is a bank teller or that Linda is a bank teller and is active in the feminist movement. The vast majority went with bank teller and active feminist, which has to be the less likely choice because the probability of two conditions will always be less than the probability of either one."

Isn't that a bad question to ask, it suggests there are only two possible outcomes, wouldn't a better question include a third option of "not a bank teller and may or may not be a active feminist"?

View on HN

Yes. You've noticed that most psychological insights and paradoxes are actually linguistic ambiguities. But it's more fun to ignore that.

Wittgenstein says hello!

Maybe your (anyone's) system 1 might assume only two possible outcomes, but it's not in the question - it's which is more likely, and one option assuredly is, up to a less-than-or-equal sign.

I hope the actual study wasn't as "tricky" as it was referenced in the article re: the Linda example

I'd imagine there's enough stuff Kahneman identified with biases that have held up and don't involve artificial questions like this designed to trick the respondents whose real world applicability seem questionable at best...

further, in the supplied example, I'd argue that the prior probability of Linda being a feminist (based on her being an activist/etc.) is probably higher than her not being a feminist so, in a sense the respondents got it right (i.e., in that population, I'd argue there are more women who are bank tellers and feminists than just bank tellers)...

It’s mathematically impossible for the set of things who have both properties X and Y, to be larger than the set of things that have property X.

"Which is more likely" isn't a question that commonly has three options. It's a question that commonly has only two options, one of which is more likely than the other.

This aspect was investigated in the literature and broadly speaking, the biases still pop-up: https://journals.sagepub.com/doi/abs/10.1111/1467-9280.00350

- It was also touched on in the original paper that Tversky and Kahneman put out https://psycnet.apa.org/record/1984-03110-001

I read that as assuming that the "bank teller only" answer implied that the person was not a feminist, since most people would assume that being a bank teller does not make you a feminist. In any case, a bad phrasing to give for a survey.... These questions would be better off being as unambiguous as possible.

Why would that imply she is NOT a feminist? It just gives no information either way.

Because when you're talking to a real human that wants to hear your thoughts about Linda, it's much, much more likely that this is what they're asking and they didn't word it accurately.

That was the whole point of this particular episode… to highlight that ‘accurately’ involves hidden assumptions that you may or may not share with the listener. And then, try to identify if there is a systemic commonality in those hidden assumptions.

I don't think you're correct. I usually see it presented as evidence that people are modeling probability wrong.

> The conjunction fallacy (also known as the Linda problem) is an inference that a conjoint set of two or more specific conclusions is likelier than any single member of that same set, in violation of the laws of probability.

‘More likely’ to be true doesn’t imply they are the only two options. But pulling out these kinds of assumptions (that you heard that limit when it wasn’t explicitly there) is exactly the kinds of things they were trying to discover, partly to make sure to ask better questions.

Reply on news.ycombinator.com