But the task is exactly the same in both situations – only the symbols on the cards have less/more familiar meaning. I see this as being not able to generalize from known examples.
And why is your assessment of the ability to generalize limited to only this math representation which people don't use?
Does this mean that these should be treated differently?
a) "3 * 2 = 6"
b) "III + III = VI"
c) "tres et tres aequalis sex"
d) "ln(2) + ln(3) = ln(6)"
While symbols (and even patterns) are not the same, you could see that all these examples point to the same thing.
> And why is your assessment of the ability to generalize limited to only this math representation which people don't use?
I think people fail to generalize even in even simpler situations. I can't think of anything obvious that would apply to adults, but this is a nice example for children: https://en.wikipedia.org/wiki/Conservation_(psychology) & video demonstration: https://www.youtube.com/watch?v=gnArvcWaH6I
The question is, what does it tell you when people struggle with Roman LCX*ML more vs the more familiar first pattern besides the fact that training/familiarity makes it easier?
And it's telling you have to reach for the kids
The article you link to seems to support this view: "Cosmides and Tooby argued that experimenters have ruled out alternative explanations, such as that people learn the rules of social exchange through practice and find it easier to apply these familiar rules than less-familiar rules."
I now wonder if people tend to do better on abstract Wason tasks after being prompted by logically-equivalent social-relations questions, as opposed to being prompted with additional abstract problems (without being told whether they gave the correct responses to the prompts.)
My guess at what might be underlying these outcomes is that people clearly see the purpose of the social rules but not of the abstract ones, and our intuition is attuned to assessing how various social scenarios will play out, taking into account the motives of the actors (this might seem to be the same as what Cosmides and Tooby claim has been ruled out, and maybe it is, but I think what they are ruling out is us having learned various schema to be applied mechanically in the corresponding social situations.)
Alternatively, could it be something as simple as this: having an understanding of the social rule's purpose helps the subjects avoid confusing conditional and biconditional rules? At least in the English language, acceptable usage can be quite ambiguous in this regard.
There is a figure with 4 cards, showing: all orange, all blue, the number 2, the number 1.
The only condition is that an even number must have an all blue flipside. This condition is either irrelevant or satisfied for the blue card. For all other cards, we don't know. The card with "1" may have the number "4" on the flip side. So we need to check ask other cards.
TL;DR: only a blue face guarantees compliance with the rule, all other cards need to be flipped.
In the second case, the unstated assumption that the cards all have one side worth digits and one side with symbols is still unstated, but a lot stronger. (These are things to check for age-alcohol compliance, so logically they must contain both age and alcoholic state of consumption.)
Very good point, thank you. I have added an explicit statement that cards are consistent.
"To deal with hyper-planes in 14-dimensional space, visualize a 3-D space and say 'fourteen' to yourself very loudly. Everybody does it." — Geoffrey Hinton