Yes, explanations are associated with predictions and it is often a bad sign when an explanation does not recommend a prediction. But no, an explanation is definitely not only as good as its predictive power.
What we’re (or at lease I’m) talking about is referred to in the literature as Hempel’s symmetry thesis: that every adequate explanation is potentially predictive and every adequate prediction is potentially explanatory. It may well be the case that every explanation is could have been a prediction but the second part, that every adequate prediction could be a successful explanation is certainly false. I don’t think any more than the above barometer example is needed to show this. But just in case: here is another. Say you see the shadow of a flag poll on the ground. You know what time it is and thus the location of the sun. Therefore, you can predict the height of the flag poll. But the facts about the shadow and the sun do not explain the height of the flag poll. Explanations just don’t have the same logical form as a predictions.
This is part of the reason why Hempel’s once dominant theory of explanation is no longer accepted. Scriven is the figure most associated with the criticisms.
This isn’t (necessarily) anything to freak out about. I’m not arguing for any wishy washy nihilism about explanations. There are probably some who do, but personally I think the best going theory of explanation is the causal theory of explanation, particularly one involving a manipulationist theory of causation like that of Judea Pearl or James Woodward (I’m dropping names so you can google if you want). Under this theory as explanation of X tells you what sorts of things you would have had to manipulate to change X. Experimentation replaces prediction as the central activity of science. This does not mean we stop predicting or that predictions are no longer important. I do not at all disagree that beliefs should pay rent. I am simply not comfortable saying that the metric by which the mysteriousness of an explanation is judged should be predictive power because predictive power is not the same as explanatory power.
Explanations just don’t have the same logical form as a predictions.
I never once even hinted at a claim that they did.
I said that an explanation is “only as good as its predictive power.” I never once mentioned anything about there being symmetry between explanations and predictions—my statements were entirely unilateral.
Explannations that lack predictive power are not useful. I could explain that the Gods of Ysgard cause storms by going bowling in the clouds after getting drunk. You can’t make any useful predictions from this but it’s a perfectly simple explanation, far simpler and comprehensive than any actually useful explanation of thunderstorms. We throw it out precisely because it is so useless.
Experimentation replaces prediction as the central activity of science
Sir, experimentation without prediction is impossible. Experimentation is meant to falsify predictions in order to validate hypothesis into theoretical models.
I do not at all disagree that beliefs should pay rent.
Then you shouldn’t argue against it.
I am simply not comfortable saying that the metric by which the mysteriousness of an explanation is judged should be predictive power because predictive power is not the same as explanatory power.
Explanations are not predictions, this is true, but nobody was claiming that they were except you. Regardless, no explanation is any more valuable than its predictive power.
Yes, explanations are associated with predictions and it is often a bad sign when an explanation does not recommend a prediction. But no, an explanation is definitely not only as good as its predictive power.
What we’re (or at lease I’m) talking about is referred to in the literature as Hempel’s symmetry thesis: that every adequate explanation is potentially predictive and every adequate prediction is potentially explanatory. It may well be the case that every explanation is could have been a prediction but the second part, that every adequate prediction could be a successful explanation is certainly false. I don’t think any more than the above barometer example is needed to show this. But just in case: here is another. Say you see the shadow of a flag poll on the ground. You know what time it is and thus the location of the sun. Therefore, you can predict the height of the flag poll. But the facts about the shadow and the sun do not explain the height of the flag poll. Explanations just don’t have the same logical form as a predictions.
This is part of the reason why Hempel’s once dominant theory of explanation is no longer accepted. Scriven is the figure most associated with the criticisms.
This isn’t (necessarily) anything to freak out about. I’m not arguing for any wishy washy nihilism about explanations. There are probably some who do, but personally I think the best going theory of explanation is the causal theory of explanation, particularly one involving a manipulationist theory of causation like that of Judea Pearl or James Woodward (I’m dropping names so you can google if you want). Under this theory as explanation of X tells you what sorts of things you would have had to manipulate to change X. Experimentation replaces prediction as the central activity of science. This does not mean we stop predicting or that predictions are no longer important. I do not at all disagree that beliefs should pay rent. I am simply not comfortable saying that the metric by which the mysteriousness of an explanation is judged should be predictive power because predictive power is not the same as explanatory power.
I never once even hinted at a claim that they did.
I said that an explanation is “only as good as its predictive power.” I never once mentioned anything about there being symmetry between explanations and predictions—my statements were entirely unilateral.
Explannations that lack predictive power are not useful. I could explain that the Gods of Ysgard cause storms by going bowling in the clouds after getting drunk. You can’t make any useful predictions from this but it’s a perfectly simple explanation, far simpler and comprehensive than any actually useful explanation of thunderstorms. We throw it out precisely because it is so useless.
Sir, experimentation without prediction is impossible. Experimentation is meant to falsify predictions in order to validate hypothesis into theoretical models.
Then you shouldn’t argue against it.
Explanations are not predictions, this is true, but nobody was claiming that they were except you. Regardless, no explanation is any more valuable than its predictive power.