«But when your counterparty is Nature herself, does it make any sense to sharply distinguish between risk and uncertainty? Isn’t the claim that there is likely to lead you astray, perhaps very badly, as simply throwing up your hands and saying “it is uncertain” keeps you from using the information about the distribution of ϕ that you do have (and you always have some) to calculate odds and then act on them?»
Maybe I am mistaken, or maybe making this more complicated than necessary, but it seems to me that there are multiple issues here. First is what we know (or have good reason to believe), second is what we do with what we know, and third is whether what we do is by choice or necessity.
If I am reading Drezdner and Grimmer correctly, then they are essentially dealing with the first issue. That is, based on what we know about the data and the models, is there good reason to believe that the polling models are accurate, and are accurately representing risk and uncertainty? If Grimmer's analysis is correct, then the answer to this question is an unqualified 'no'. And the danger of overvaluing these models is that we accept their "guesses" at the quantification of uncertainty as reasonable measures of what is not (yet) quantifiable.
The second issue is what we might do based upon our knowledge or lack thereof, and why. It seems clear that making a bet with some counterparty is always a matter of choice. We can choose what odds and stakes we might accept, or even whether we accept at all. In some cases where uncertainty is high, we might conclude that the counterparty has knowledge that we lack, and choose not to accept any odds that they might offer. Alternatively, we might conclude that the counterparty is ignorant and offers odds that are unrealistically unfavourable to themselves, and choose to accept at some reasonable odds.
I don't think that has anything in particular to do with whether the matter has to do with nature or human action. For example, if there is a prediction that it will rain heavily at some location tomorrow, and that the rainfall with be between one and two centimeters, and someone offers 100-1 odds that it will be more than 1.5 centimeters, then one might choose to accept the bet. This seems to be a case of true uncertainty, with no known way that the counterparty could have additional relevant information.
But in any case, in choosing to make a bet with some counterparty, there is always at least the question of whether they know something that you do not, and whether there is additional information that removes some of the uncertainty. In choosing to make such a bet, one is not evaluating the uncertainty (per se), but the state of the counterparties knowledge.
At other times, the question is: "do we need to act?" If we do not - if, for example, we are engaged in pure research and simply seeking knowledge - then indeed we can simply throw up our hands and say something like "it looks like X, but the uncertainty is very high, so our reasons to believe that it is X are only very weak. We need more data." If we do need to act in some way - and this can be so whether or not our "counterparty is Nature herself" (unless by 'Nature herself' we mean something like "the world"), then we have no choice but to act based upon the best reasons we have, whether or not they are sufficiently "good". If we need to decide when we will plant our crops, and the only thing we can do is decide bases on the best data we have, recognizing that we may be wrong. So also for a political campaign's decision as to where to spend its advertising dollars or to hold additional campaign stops. In both cases, uncertainty may be high, but decision is required, so the only option is to use "the information about the distribution of ϕ that you do have" - however good it is - to make such a decision (again, recognizing that it is uncertain).
As such, it seems to me something of a mistake to conflate betting markets - where even participating is a choice - with forced decisions in cases of incomplete or uncertain information. And also to conflate what we know (or have good reason to believe) with what we choose to do (or must do) with what we know.
«But when your counterparty is Nature herself, does it make any sense to sharply distinguish between risk and uncertainty? Isn’t the claim that there is likely to lead you astray, perhaps very badly, as simply throwing up your hands and saying “it is uncertain” keeps you from using the information about the distribution of ϕ that you do have (and you always have some) to calculate odds and then act on them?»
Maybe I am mistaken, or maybe making this more complicated than necessary, but it seems to me that there are multiple issues here. First is what we know (or have good reason to believe), second is what we do with what we know, and third is whether what we do is by choice or necessity.
If I am reading Drezdner and Grimmer correctly, then they are essentially dealing with the first issue. That is, based on what we know about the data and the models, is there good reason to believe that the polling models are accurate, and are accurately representing risk and uncertainty? If Grimmer's analysis is correct, then the answer to this question is an unqualified 'no'. And the danger of overvaluing these models is that we accept their "guesses" at the quantification of uncertainty as reasonable measures of what is not (yet) quantifiable.
The second issue is what we might do based upon our knowledge or lack thereof, and why. It seems clear that making a bet with some counterparty is always a matter of choice. We can choose what odds and stakes we might accept, or even whether we accept at all. In some cases where uncertainty is high, we might conclude that the counterparty has knowledge that we lack, and choose not to accept any odds that they might offer. Alternatively, we might conclude that the counterparty is ignorant and offers odds that are unrealistically unfavourable to themselves, and choose to accept at some reasonable odds.
I don't think that has anything in particular to do with whether the matter has to do with nature or human action. For example, if there is a prediction that it will rain heavily at some location tomorrow, and that the rainfall with be between one and two centimeters, and someone offers 100-1 odds that it will be more than 1.5 centimeters, then one might choose to accept the bet. This seems to be a case of true uncertainty, with no known way that the counterparty could have additional relevant information.
But in any case, in choosing to make a bet with some counterparty, there is always at least the question of whether they know something that you do not, and whether there is additional information that removes some of the uncertainty. In choosing to make such a bet, one is not evaluating the uncertainty (per se), but the state of the counterparties knowledge.
At other times, the question is: "do we need to act?" If we do not - if, for example, we are engaged in pure research and simply seeking knowledge - then indeed we can simply throw up our hands and say something like "it looks like X, but the uncertainty is very high, so our reasons to believe that it is X are only very weak. We need more data." If we do need to act in some way - and this can be so whether or not our "counterparty is Nature herself" (unless by 'Nature herself' we mean something like "the world"), then we have no choice but to act based upon the best reasons we have, whether or not they are sufficiently "good". If we need to decide when we will plant our crops, and the only thing we can do is decide bases on the best data we have, recognizing that we may be wrong. So also for a political campaign's decision as to where to spend its advertising dollars or to hold additional campaign stops. In both cases, uncertainty may be high, but decision is required, so the only option is to use "the information about the distribution of ϕ that you do have" - however good it is - to make such a decision (again, recognizing that it is uncertain).
As such, it seems to me something of a mistake to conflate betting markets - where even participating is a choice - with forced decisions in cases of incomplete or uncertain information. And also to conflate what we know (or have good reason to believe) with what we choose to do (or must do) with what we know.