If you want to know how far a rock you throw will land (a prediction based on a model constructed based on previously performed experiments), you want your model to have the necessary predictive power. Whether it corresponds to some metaphysical concept of reality is quite secondary.
That doesn’t answer my question. To rephrase using your new example, if the prior experiments do not metaphorically “tap into reality,” why should I have any confidence that a model based on those experimental results will be useful in predicting future events?
Well, either the experimental result has predictive power, or it doesn’t. If certain kinds of experimental results prove useful for predicting the future, then I should have confidence in predictions based on (models based on) those results. Whether I call them “reality” or “a model” doesn’t really matter very much.
More generally, to my way of thinking; this whole “instrumentalists don’t believe in reality” business mostly seems like a distinction in how we use words rather than in what experiences we anticipate.
It would potentially make a difference, I suppose, if soi-disant instrumentalists didn’t actually expect the results of different experiments to be reconcilable with one another (under the principle that each experiment was operating on its own model, after all, and there’s no reason to expect those models to have any particular relationship to one another). But for the most part, that doesn’t seem to be the case.
There’s a bit of that when it comes to quirky quantum results, I gather, but to my mind that’s kind of an “instrumentalism of the gaps”… when past researchers have come up with a unified model we accept that unified model, but when current data doesn’t seem unified given our current understanding, rather than seeking a unified model we shrug our shoulders and accept the inconsistency, because hey, they’re just models, it’s not like there’s any real underlying territory.
Which in practice just means we wait for someone else to do the hard work of reconciling it all.
why should I have any confidence that a model based on those experimental results will be useful in predicting future events?
Because it has been experimentally confirmed before, and from experience we can assign a high probability that a model that has been working well in the past will continue to work in the similar circumstances in the future.
If you want to know how far a rock you throw will land (a prediction based on a model constructed based on previously performed experiments), you want your model to have the necessary predictive power. Whether it corresponds to some metaphysical concept of reality is quite secondary.
That doesn’t answer my question. To rephrase using your new example, if the prior experiments do not metaphorically “tap into reality,” why should I have any confidence that a model based on those experimental results will be useful in predicting future events?
Well, either the experimental result has predictive power, or it doesn’t. If certain kinds of experimental results prove useful for predicting the future, then I should have confidence in predictions based on (models based on) those results. Whether I call them “reality” or “a model” doesn’t really matter very much.
More generally, to my way of thinking; this whole “instrumentalists don’t believe in reality” business mostly seems like a distinction in how we use words rather than in what experiences we anticipate.
It would potentially make a difference, I suppose, if soi-disant instrumentalists didn’t actually expect the results of different experiments to be reconcilable with one another (under the principle that each experiment was operating on its own model, after all, and there’s no reason to expect those models to have any particular relationship to one another). But for the most part, that doesn’t seem to be the case.
There’s a bit of that when it comes to quirky quantum results, I gather, but to my mind that’s kind of an “instrumentalism of the gaps”… when past researchers have come up with a unified model we accept that unified model, but when current data doesn’t seem unified given our current understanding, rather than seeking a unified model we shrug our shoulders and accept the inconsistency, because hey, they’re just models, it’s not like there’s any real underlying territory.
Which in practice just means we wait for someone else to do the hard work of reconciling it all.
Because it has been experimentally confirmed before, and from experience we can assign a high probability that a model that has been working well in the past will continue to work in the similar circumstances in the future.