Has anyone read "How We Reason" by Philip Johnson-Laird? He and others in his field (the "model theory" of psychology/cognitive science) claim that their studies refute the naive claim that human brains often operate in terms of logic or Bayesian reasoning (probablistic logic). I gather they'd say that we are not Jaynes' perfect Bayesian reasoning robot or even something resembling a computationally bounded approximation to it.
I'm intrigued by this recommendation:
... formal logic cannot be the basis for human reason. Johnson-Laird reviews evidence to this effect. For example, there are many valid conclusions that we never bother to draw because they are of no practical use to us. We also make systematic errors in reasoning that we would not make using logic. The content of logic problems used in research studies greatly affects their difficulty; it would not if logic were the primary process. We use knowledge to help us imagine possibilities and then evaluate the possibilities for consistency with other evidence.
Constrained by the span of short term memory, the strength of our general intellectual abilities, and our level of expertise, we construct and manipulate mental models of the problems we reason about. "...[F]rom the meanings of sentences in connected discourse, the listener implicitly sets up a much abbreviated and not especially linguistic model of the narrative ... Where the model is incomplete, material may even be unwittingly invented to render the memory more meaningful or more plausible." We can manipulate these models in a number of ways, including updating them with new information, combining two models when appropriate, searching for confirming evidence or information, and using counterexamples to challenge a model's validity.
Mental model theory explains a number of systematic errors human beings make when reasoning. For example, the difficulty of reasoning problems is related to the number of models that must be held simultaneously in memory to work through them. And we exhibit a recurring bias to use a single model to reason about situations that have more possibilities than we can keep track of. We oversimplify. Consistent with model theory, we have difficulty reasoning with information about what is false about a situation.The real key to human rationality is our ability to recognize and grasp the implications of counterexamples
It seems like an interesting read, but I'd like to know if the research field is a scientific one, i.e. that their stories aren't just pleasing, but can predict, or at least explain real phenomena.
In the Google books preview, I see the author spends some time claiming that we build iconic visual/spatial representations and that a lot of our thinking isn't verbal or available to verbal introspection (fairly uncontroversial to me).
I liked the two related imagination-puzzles:
1. I have thousands and thousands of very thin needles, which I hold in a bundle in my hands. I throw them up into the air, imparting a random force to each of them. They fall to the ground, but, before any of them hits the ground, I stop them by magic in mid-air. Many of the needles are horizontal or nearly so, and many of them are vertical or nearly so. Are there likely to be more needles in the first category, more needles in the second category, or do the two categories have roughly equal numbers?
[and the same thing but for very thin circular disks - let's assume they're also dense, so the air isn't a factor]
2. I have thousands and thousands of very thin circular disks, which I hold in a bundle in my hands. I throw them up into the air, imparting a random force to each of them. They fall to the ground, but, before any of them hits the ground, I stop them by magic in mid-air. Many of the disks are horizontal or nearly so, and many of them are vertical or nearly so. Are there likely to be more disks in the first category, more disks in the second category, or do the two categories have roughly equal numbers?
He claims that for spatial propositions where we can imagine a picture that's more or less equivalent ("the cabinet is behind the piano" [as we face the keys]), the negation of that proposition can't be so pictured (in direct correspondence) because ... where would you put the cabinet? You could imagine all the alternative places it could be (presupposing that there is a specific piano and specific cabinet). You could imagine something "not cabinet" behind the piano (a cabinet with a red x, a cabinet repelling field?). He suggests an or(p1,p2,....pn) of images where we imagine the cabinet to be (that aren't behind the piano). I'm not sure what we can conclude from this. We already know that negation is tricky - linguistically, and mentally. Maybe I like to imagine someone telling me "no, you're wrong to say X" - to use a non-visual system.
He explains that inferences about (written) non-spatial visual relations (light/dark clean/dirty) take longer to process than spatial ones, that the spatial and visual word inference word problems had different fMRI hot spots, that congenitally blind people weren't faster on spatial queries (i.e. were slower on average than non-blind, but didn't suffer any additional penalty on the visual ones). I suppose this could be taken as weak evidence that we can perform "logical" inferences with some sort of spatial-relationship processing, and that perhaps non-spatial attributes take longer to translate (even though they refer to visual qualities like light/dark).
I'm leaning toward buying the book, since the writing is pleasant. But I thought first I'd ask if anyone could recommend for the quality of research in this field.
The key is that there are three different dimensions being collapsed into two "orientations". With a needle, two dimensions are "horizontal", and one is "vertical". With a disk, two are "vertical", and one is "horizontal". Part of the issue with this problem is that people don't generally have an explicit definition of "vertical" and "horizontal". Ask them whether they know what they mean, and they'll say "sure", but ask them to give a definition, and they'll flounder.