> In other words, it appears to make the model more complicated and less aesthetically pleasing.
There is this perpetuating religious dogma in physics that physicists are uncovering the "elegant" rules of god, rather than creating models that match their observations. Models make predictions, and although they do reveal some things about the nature of reality, they surely do not define what it is. The next model might define it to be something completely different. Yes, we generally find that most physical phenomena can be very elegantly modeled, but that does not make a simple model better than a complex model that matches observation better.
That's true, but that statement was in the context of the desire that some people (including, it would seem, the author of the article) have to get rid of the cosmological constant. The only reasons I am aware of that anyone dislikes the cosmological constant are aesthetic. It's an extra parameter that makes things a little more complicated, but there's no a priori reason it should be zero. Thus, I assert, if one wants to get rid of it, you shouldn't be replacing it with a _more_ complicated model that's no more predictive.
Good point in that the aesthetics of a model are in the eye of the beholder - although:
> that does not make a simple model better than a complex model that matches observation better
Perhaps this is less of an issue in physics than some other fields because the observations are less noisy (?) -- but if getting the best match for existing observations is all you care about from a model, you could end up with something grossly over-fitted. I guess it goes without saying that any proposed physical laws need to pass the equivalent of cross-validation :)
Comments
> In other words, it appears to make the model more complicated and less aesthetically pleasing.
There is this perpetuating religious dogma in physics that physicists are uncovering the "elegant" rules of god, rather than creating models that match their observations. Models make predictions, and although they do reveal some things about the nature of reality, they surely do not define what it is. The next model might define it to be something completely different. Yes, we generally find that most physical phenomena can be very elegantly modeled, but that does not make a simple model better than a complex model that matches observation better.
That's true, but that statement was in the context of the desire that some people (including, it would seem, the author of the article) have to get rid of the cosmological constant. The only reasons I am aware of that anyone dislikes the cosmological constant are aesthetic. It's an extra parameter that makes things a little more complicated, but there's no a priori reason it should be zero. Thus, I assert, if one wants to get rid of it, you shouldn't be replacing it with a _more_ complicated model that's no more predictive.
Good point in that the aesthetics of a model are in the eye of the beholder - although:
> that does not make a simple model better than a complex model that matches observation better
Perhaps this is less of an issue in physics than some other fields because the observations are less noisy (?) -- but if getting the best match for existing observations is all you care about from a model, you could end up with something grossly over-fitted. I guess it goes without saying that any proposed physical laws need to pass the equivalent of cross-validation :)
Except that in all cases so far, the correct model turned out to be exceedingly simple.