You wrote “Counterfactuals are a specific type of conditional prediction”. I’m not really a fan of this phrasing. What happens is that we change the world model then make a prediction. This isn’t the same as conditioning on an observation where the world model is left intact (which is how EDT tries to generate counterfactuals)
You describe the sun producing heat as a logically necessary connection. I disagree, at most it’s a linguistic convention, rather than a matter of logical necessity.
“Conditional predictions are a specific type of causal claim, that is, one in which two items are linked by a correlation”—I’m confused by this statement as correlation != causation
“Since Orion doesn’t know __L__, how can Orion figure out what he needs to learn in order to know __K __U __L __and nothing more?”—I’m confused. Doesn’t Orion already know more than this in that they can narrow it down to the K region which is a subset of the union. Could you clarify what union means here?
I agree that the world model is changing. I should have specified that what makes a counterfactual a specific type is that it conditions upon a change in the world state. Thus later, when I wrote, “The causal claim contains a counterfactual,” I could improve the clarity by pointing out that the causal claim contains a ‘world state variable’ which when altered creates a counterfactual world.
It is certainly more than a linguistic convention. When one applies the concept of causality to their experience and knowledge about the big bright glowing orb and to their knowledge and experience of heat, eventually the referents of the two terms “sun” and “heat” are shown to have necessary connection, such that in any stable, univocal use the terms there is a logical connection. I probably should have stuck to my draft version of the sun-heat paragraph, which, inspired by a section from Kant, was also significantly longer.
Yes, correlation is not causation, but when two things are correlated, there is either no relation or some relation. But when I make a conditional prediction, I assume that the antecedent and the consequent have some sort of causal connection either directly, or through a backdoor, or through colliders. Now I realize, conditional predictions don’t actually require the assumption that there is a link. Yet, I am going to assume people are trying to make meaningful speech acts, and this context, a conditional prediction implies that there is a correlation, that is in some perhaps unknown way indicative of causal forces.
You make sense to say that if Orion knows that he only knows K, then he actually knows more than K by virtue of knowing K at the limit. I disagree with this. He knows K only. He does not know that Perpetua knows the union of K and L. Nor does he know that Perpetua knows only K and L and no more. Nor does he know that K and L exist as a union. I think Perpetua should be able to identify that Orion knows K and only K, and create a path for him to learn K U L. But I am skeptical there is any way for Orion to figure out the path from his knowledge state of K to her knowledge state of K U L.
A few thoughts:
You wrote “Counterfactuals are a specific type of conditional prediction”. I’m not really a fan of this phrasing. What happens is that we change the world model then make a prediction. This isn’t the same as conditioning on an observation where the world model is left intact (which is how EDT tries to generate counterfactuals)
You describe the sun producing heat as a logically necessary connection. I disagree, at most it’s a linguistic convention, rather than a matter of logical necessity.
“Conditional predictions are a specific type of causal claim, that is, one in which two items are linked by a correlation”—I’m confused by this statement as correlation != causation
“Since Orion doesn’t know __L__, how can Orion figure out what he needs to learn in order to know __K __U __L __and nothing more?”—I’m confused. Doesn’t Orion already know more than this in that they can narrow it down to the K region which is a subset of the union. Could you clarify what union means here?
I agree that the world model is changing. I should have specified that what makes a counterfactual a specific type is that it conditions upon a change in the world state. Thus later, when I wrote, “The causal claim contains a counterfactual,” I could improve the clarity by pointing out that the causal claim contains a ‘world state variable’ which when altered creates a counterfactual world.
It is certainly more than a linguistic convention. When one applies the concept of causality to their experience and knowledge about the big bright glowing orb and to their knowledge and experience of heat, eventually the referents of the two terms “sun” and “heat” are shown to have necessary connection, such that in any stable, univocal use the terms there is a logical connection. I probably should have stuck to my draft version of the sun-heat paragraph, which, inspired by a section from Kant, was also significantly longer.
Yes, correlation is not causation, but when two things are correlated, there is either no relation or some relation. But when I make a conditional prediction, I assume that the antecedent and the consequent have some sort of causal connection either directly, or through a backdoor, or through colliders. Now I realize, conditional predictions don’t actually require the assumption that there is a link. Yet, I am going to assume people are trying to make meaningful speech acts, and this context, a conditional prediction implies that there is a correlation, that is in some perhaps unknown way indicative of causal forces.
You make sense to say that if Orion knows that he only knows K, then he actually knows more than K by virtue of knowing K at the limit. I disagree with this. He knows K only. He does not know that Perpetua knows the union of K and L. Nor does he know that Perpetua knows only K and L and no more. Nor does he know that K and L exist as a union. I think Perpetua should be able to identify that Orion knows K and only K, and create a path for him to learn K U L. But I am skeptical there is any way for Orion to figure out the path from his knowledge state of K to her knowledge state of K U L.